ReportWire

Tag: AI agents

  • Exclusive: AI financial platform Rowspace raises $50 million round led by Sequoia to help investment firms take on messy data | Fortune

    [ad_1]

    After meeting in graduate school at MIT, Michael Manapat and Yibo Ling embarked on different career paths. Manapat held chief technical roles at Stripe and Notion, while Ling led finance teams at Uber and Binance. Still, they both confronted a similar challenge: How to assemble fragmented data to make important decisions about capital allocation, workflows and more.

    When OpenAI released ChatGPT in November 2022, Ling tested to see how well it could carry out basic due diligence tasks. He quickly found the new AI tool was hampered by a familiar problem: Data. “Clearly there was a lot of promise, but it just wasn’t working. You need the right information in the right context,” he told Fortune. 

    That realization motivated Manapat and Ling to join forces to build Rowspace, an AI platform that allows financial outfits like private equity firms and hedge funds to turn their years of proprietary data into alpha. The company is publicly launching today with a $50 million funding round led by Sequoia, with participation from Emergence Capital, Stripe, and Conviction, along with other firms and angel investors. 

    At a time of pearl-clutching and market turmoil on whether large language models and foundation models will render software obsolete, Sequoia investor and co-steward Alfred Lin told Fortune that Rowspace is a prime example of the type of application that will thrive in the brave new AI-empowered world. 

    “The thing that people are talking about is the marginal line of code is very cheap to produce,” Lin said. “What we’re looking for now in almost every single company is product velocity, and how fast product velocity generates other things that become moats, which are like network effects and people using your product on a daily basis.” 

    Finding alpha

    Manapat described Rowspace as the intelligence layer that sits on top of a firm’s data. The platform integrates all of an institution’s structured and unstructured data, whether in the form of documents or accounting systems or old PowerPoints, and performs reasoning in advance. “We’re focused on how we make sure we understand all of the underlying data to drive actual decision-making,” he said. 

    Rowspace’s approach to data sounds a lot like the one used by popular new consumer tools such as Claude Cowork, which can query a computer’s files and create presentations or research memos. Manapat said that Rowspace is different in crucial ways. For one, it doesn’t take possession of a firm’s data, instead doing processing inside its own cloud systems. 

    On a deeper level, Manapat said that foundation models like Anthropic are good at last mile tasks, like formatting a pitchbook in PowerPoint or building a cash flow model, which are generally completed with a real-time search approach.  

    “That’s not where our focus is,” Manapat said. As he explained, there are no ways to ensure the agent looked at all available information or took the time to reason in advance of making a conclusion, which is time-consuming and expensive. Instead, Rowspace is tasked with deeper analysis of data, such as being able to notice minute details from years of a company’s finances. That will always give the platform an advantage over the more general purpose Anthropics of the world. 

    “The foundation model is not going to be able to cater to every single [thing] that someone wants to do in all these different industries,” said Lin. “That is going to be left to players like Rowspace, specifically for the vertical they’re focused on.” 

    Manapat admitted that pure software or user interfaces are going to be hard to defend, especially as foundation models rapidly advance. But he said that’s why Rowspace’s focus is more on compiling and synthesizing a firm’s data in a secure way, and doing so with a financially literate team. The engineering corps comes both from tech-first companies like Notion and Stripe as well as private equity and credit. “There’s no one size fits all solution in financial services, because in some sense, each firm’s alpha comes from their approach,” Manapat said. “We’re trying to help you learn from your own data and knowledge and approach and amplify that.”

    While Rowspace declined to name its valuation or early customers, Manapat said that they include longstanding and name-brand private equity and credit firms, as well as crossover firms that work in both public and private markets. He added that Rowspace is working with about ten top firms with seven-figure annual contract values. 

    “Customers use this tool to make money, and that’s where the rubber meets the road,” Lin said. “If we consistently, with our tool, help people use AI to make better decisions, they will make money, and they’ll do it better than others.” 

    [ad_2]

    Leo Schwartz

    Source link

  • New Research Shows AI Agents Are Running Wild Online, With Few Guardrails in Place

    [ad_1]

    In the last year, AI agents have become all the rage. OpenAI, Google, and Anthropic all launched public-facing agents designed to take on multi-step tasks handed to them by humans. In the last month, an open-source AI agent called OpenClaw took the web by storm thanks to its impressive autonomous capabilities (and major security concerns). But we don’t really have a sense of the scale of AI agent operations, and whether all the talk is matched by actual deployment. The MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) set out to fix that with its recently published 2025 AI Agent Index, which provides our first real look at the scale and operations of AI agents in the wild.

    Researchers found that interest in AI agents has undoubtedly skyrocketed in the last year or so. Research papers mentioning “AI Agent” or “Agentic AI” in 2025 more than doubled the total from 2020 to 2024 combined, and a McKinsey survey found that 62% of companies reported that their organizations were at least experimenting with AI agents.

    With all that interest, the researchers focused on 30 prominent AI agents across three separate categories: chat-based options like ChatGPT Agent and Claude Code; browser-based bots like Perplexity Comet and ChatGPT Atlas; and enterprise options like Microsoft 365 Copilot and ServiceNow Agent. While the researchers didn’t provide exact figures on just how many AI agents are deployed across the web, they did offer a considerable amount of insight into how they are operating, which is largely without a safety net.

    Just half of the 30 AI agents that got put under the magnifying glass by MIT CSAIL include published safety or trust frameworks, like Anthropic’s Responsible Scaling Policy, OpenAI’s Preparedness Framework, or Microsoft’s Responsible AI Standard. One in three agents has no safety framework documentation whatsoever, and five out of 30 have no compliance standards. That is troubling when you consider that 13 of 30 systems reviewed exhibit frontier levels of agency, meaning they can operate largely without human oversight across extended task sequences. Browser agents in particular tend to operate with significantly higher autonomy. This would include things like Google’s recently launched AI “Autobrowse,” which can complete multi-step tasks by navigating different websites and making use of user information to do things like log into sites on your behalf.

    One of the troubles with letting agents browse freely and with few guardrails is that their activity is nearly indistinguishable from human behavior, and they do little to dispel any confusion that might occur. The researchers found that 21 out of the 30 agents provide no disclosure to end users or third parties that they are AI agents and not human users. This results in most AI agent activity being mistaken for human traffic. MIT found that just seven agents published stable User-Agent (UA) strings and IP address ranges for verification. Nearly as many explicitly use Chrome-like UA strings and residential/local IP contexts to make their traffic requests appear more human, making it next to impossible for a website to distinguish between authentic traffic and bot behavior.

    For some AI agents, that’s actually a marketable feature. The researchers found that BrowserUse, an open-source AI agent, sells itself to users by claiming to bypass anti-bot systems to browse “like a human.” More than half of all the bots tested provide no specific documentation about how they handle robots.txt files (text files that are placed in a website’s root directory to instruct web crawlers on how they can interact with the site), CAPTCHAs that are meant to authenticate human traffic, or site APIs. Perplexity has even made the case that agents acting on behalf of users shouldn’t be subject to scraping restrictions since they function “just like a human assistant.”

    The fact that these agents are out in the wild without much protection in place means there is a real threat of exploits. There is a lack of standardization for safety evaluations and disclosures, leaving many agents potentially vulnerable to attacks like prompt injections, in which an AI agent picks up on a hidden malicious prompt that can make it break its safety protocols. Per MIT, nine of 30 agents have no documentation of guardrails against potentially harmful actions. Nearly all of the agents fail to disclose internal safety testing results, and 23 of the 30 offer no third-party testing information on safety.

    Just four agents—ChatGPT Agent, OpenAI Codex, Claude Code, and Gemini 2.5—provided agent-specific system cards, meaning the safety evaluations were tailored to how the agent actually operates, not just the underlying model. But frontier labs like OpenAI and Google offer more documentation on “existential and behavioral alignment risks,” they lack details on the type of security vulnerabilities that may arise during day-to-day activities—a habit that the researchers refer to as “safety washing,” which they describe as publishing high-level safety and ethics frameworks while only selectively disclosing the empirical evidence required to rigorously assess risk.

    There has at least been some momentum toward addressing the concerns raised by MIT’s researchers. Back in December, OpenAI and Anthropic (among others) joined forces, announcing a foundation to create a development standard for AI agents. But the AI Agent Index shows just how wide the transparency gap is when it comes to agentic AI operation. AI agents are flooding the web and workplace, functioning with a shocking amount of autonomy and minimal oversight. There’s little to indicate at the moment that safety will catch up to scale any time soon.

    [ad_2]

    AJ Dellinger

    Source link

  • Frontline AI in action: How AI-powered tools are reshaping work where it matters most – Microsoft in Business Blogs

    [ad_1]

    Frontline workers are the foundation of every industry—from retail and healthcare to hospitality and field services. Yet for years, they’ve been asked to increase productivity and deliver more value, faster, often with tools that weren’t designed for the specific realities of frontline work.

    Today, that dynamic is shifting.

    When AI is applied in practical, governed ways, it has the power to transform everyday work—reducing friction in daily workflows, empowering faster and more confident decision-making, and giving workers back time for what matters most: human connection. This shift isn’t theoretical. It’s already unfolding across frontline environments, driven by tools that meet workers where they are—on shared devices, on mobile, and inside the applications they already use.

    Voices from the Frontline: AI in Action is a limited podcast series, hosted by bestselling author and industry influencer Ron Thurston and sponsored by Microsoft. Across the series, frontline leaders and practitioners share how AI is being used today to simplify work, strengthen service, and support people—not replace them.

    Below are the key themes emerging from those conversations.

    Bringing AI into everyday frontline workflows

    For frontline teams, adoption starts with simplicity.

    Rather than introducing entirely new systems, organizations are embedding AI into familiar tools—making it easier to access intelligence without disrupting the flow of work. AI agents are emerging as the next evolution of workplace apps: purpose-built, task-focused assistants that help frontline employees find information, complete routine tasks, and stay organized. Microsoft 365 Copilot is centering agents at the core of frontline digital transformation.

    Because Copilot is embedded across Microsoft tools, frontline workers can access support through a single, intuitive entry point. This reduces context switching and lowers the barrier to adoption—especially in high-paced environments.

    As Abbie Sweeney, a program leader on the Microsoft 365 Copilot team, explained during the podcast series, “the goal isn’t automation for its own sake. It’s removing everyday friction so workers can focus on customers, patients, and guests.”

    Simplifying scheduling, reporting, and communication

    Some of the most immediate impact of AI shows up in the least glamorous tasks.

    Across industries, frontline leaders spend hours each week on scheduling, reporting, and administrative follow-up. AI can help streamline these processes—summarizing emails, generating meeting notes, and answering operational questions in seconds.

    For frontline employees, this means faster access to information like inventory availability, shift details, or process guidance without leaving the floor or logging into multiple systems. These time savings compound quickly, freeing up capacity for higher value, customer facing work.

    Sweeney also emphasized that, “making those processes efficient is really what Copilot is about—giving time back to the people who need it most.”

    AI in action on Microsoft’s own frontlines

    Microsoft applies the same tools internally that it brings to customers.

    At the Microsoft Experience Center in New York City, frontline associates use Copilot in Microsoft Teams and Microsoft Dynamics 365 to coordinate work, manage events, and support customers in a live retail environment. From onboarding new hires to managing high volumes of customer interactions, AI helps associates stay informed and responsive—even during peak demand.

    New employees can ask Copilot questions to quickly learn procedures and find answers without digging through long documents. Managers rely on AI to help them keep track of schedules, emails, and event logistics, ensuring teams have what they need to deliver consistent experiences.

    This “customer zero” approach allows Microsoft to learn, iterate, and scale frontline innovation based on real-world use.

    Scaling AI responsibly, with people at the center

    One theme cuts across every conversation in the series: successful AI adoption is people led.

    Rather than imposing new tools from the top down, organizations are seeing stronger results when they empower frontline employees to experiment, provide feedback, and shape how AI fits into their work. With clear governance and responsible AI principles in place, this approach supports organic adoption, faster iteration and sustainable scale—without compromising trust or security.

    The result is not just operational efficiency, but improved customer experiences, greater consistency, and enhanced connection at the frontline.

    The future of frontline work

    Technology alone doesn’t transform work—people do.

    When frontline teams are equipped with AI tools that respect how they work and what they value, the impact is immediate and tangible. Communication becomes clearer. Decisions happen faster. And workers gain more time to focus on the human moments that define great service.

    These aren’t future-state aspirations. They’re happening now, across industries, as organizations rethink how AI can truly support the people on the frontlines.

    Listen to the full series

    Explore Voices from the Frontline: AI in Action, a limited podcast series, hosted by bestselling author and industry influencer Ron Thurston and sponsored by Microsoft.

    🎧 Listen on Apple Podcasts
    🎧 Listen on Spotify
    📺 Watch on YouTube

    [ad_2]

    Microsoft in Business Team

    Source link

  • OpenAI launches a way for enterprises to build and manage AI agents | TechCrunch

    [ad_1]

    OpenAI has launched a new product to help enterprises navigate the world of AI agents, focusing on agent management as critical infrastructure for enterprise AI adoption.

    On Thursday, AI giant OpenAI announced the launch of OpenAI Frontier, an end-to-end platform designed for enterprises to build and manage AI agents, on Thursday. It’s an open platform, which means users can manage agents built outside of OpenAI too.

    Frontier users can program AI agents to connect to external data and applications which allows them to execute tasks far outside of the OpenAI platform. Users can also limit and manage what these agents have access to, and what they can do, of course.

    OpenAI said Frontier was designed to work the same way companies manage human employees. Frontier offers an onboarding process for agents and a feedback loop that is meant to help them improve over time the same way a review might help an employee.

    OpenAI touted enterprises including HP, Oracle, State Farm and Uber as customers, but Frontier is currently only available to a limited number of users with plans to roll out more generally in the coming months.

    The company would not disclose pricing details on a press briefing earlier this week, according to reporting from The Verge. TechCrunch has also reached out for more information regarding pricing.

    Agent-management products become table stakes since AI agents rose to prominence in 2024. Salesforce has arguably the best-known such product, Agentforce, which the company launched in the fall of 2024. Others have quickly followed. LangChain is a notable player in the space that was founded in 2022 and has raised more than $150 million in venture capital. CrewAI is a smaller upstart that has raised more than $20 million in venture capital.

    Techcrunch event

    Boston, MA
    |
    June 23, 2026

    In December, global research and advisory firm Gartner released a report about this type of software and called agent management platforms both the “most valuable real estate in AI” and a necessary piece of infrastructure for enterprises to adopt AI.

    It’s not surprising that OpenAI would release this platform in early 2026 as the company has made it clear that enterprise adoption is one of its main focus areas for this year. The company has also announced two notable enterprise deals this year with ServiceNow and Snowflake.

    Still, if OpenAI wants to be a meaningful player in the enterprise space, offering a product like Frontier is a promising step.

    [ad_2]

    Rebecca Szkutak

    Source link

  • Synthesia hits $4B valuation, lets employees cash out | TechCrunch

    [ad_1]

    British startup Synthesia, whose AI platform helps companies create interactive training videos, has raised a $200 million Series E round of funding that brings its valuation to $4 billion — up from $2.1 billion just a year ago.

    Unlike some other AI startups that are still a long way from turning a profit, Synthesia has found a lucrative business in transforming corporate training thanks to AI-generated avatars. With enterprise clients including Bosch, Merck, and SAP, the London-based company crossed $100 million in annual recurring revenue (ARR) in April 2025.

    This milestone explains why Synthesia’s venture backers are literally doubling down. The Series E that nearly doubled its valuation was led by existing investor GV (Google Ventures), with participation from several other previous backers — including Series B lead Kleiner Perkins, Series C lead Accel, Series D lead New Enterprise Associates (NEA), NVIDIA’s venture capital arm NVentures, Air Street Capital, and PSP Growth. 

    Aside from ongoing support, this round will bring both new and departing investors. On one hand, Matt Miller’s VC firm Evantic and the secretive VC firm Hedosophia are joining the cap table as new entrants. On the other hand, Synthesia will facilitate an employee secondary sale in partnership with Nasdaq, TechCrunch has learned.

    To be clear, Synthesia isn’t going public just yet — Nasdaq isn’t acting as a public exchange in this operation, but as a private markets facilitator that will help early team members turn their shares into cash. These employee stock sales often happen outside of this framework, but usually at prices either below or above the company’s official valuation, and are sometimes frowned upon by other shareholders. With this process, all sales will be tied to the same $4 billion valuation as Synthesia’s Series E, while the company keeps an element of control.

    “This secondary is first and foremost about our employees,” Synthesia CFO Daniel Kim told TechCrunch. “It gives employees a meaningful opportunity to access liquidity and share in the value they’ve helped create, while we continue to operate as a private company focused on long-term growth.”

    For Synthesia, this long-term growth involves going beyond expressive videos and embracing the AI agents trend. According to a press release, the company is developing AI agents that will let its clients’ employees “interact with company knowledge in a more intuitive, human-like way by asking questions, exploring scenarios through role-play, and receiving tailored explanations.”

    Techcrunch event

    San Francisco
    |
    October 13-15, 2026

    The company said early pilots have received positive feedback from customers, who reported higher engagement and faster knowledge transfer compared to traditional formats. This positive response explains why Synthesia now plans to make agents a “core strategic focus” to invest in, alongside further product improvements to its existing platform.

    While it didn’t disclose revenue forecasts, the company hopes its platform will offer a welcome answer to the struggles of enterprises in keeping their workforce adequately trained despite rapid changes. “We see a rare convergence of two major shifts: a technology shift with AI agents becoming more capable, and a market shift where upskilling and internal knowledge sharing have become board-level priorities,” Synthesia’s co-founder and CEO Victor Riparbelli said in a statement.

    Seeing boards care more about employees as a result of AI wasn’t on anyone’s bingo card, except perhaps Riparbelli. Together with his cofounder, Synthesia COO Steffen Tjerrild, Riparbelli took the initiative of conducting a secondary sale so that employees could share in the success of the unicorn company. Founded in 2017, Synthesia now has more than 500 team members, a 20,000-square-foot HQ in London, and additional offices in Amsterdam, Copenhagen, Munich, New York City, and Zurich.

    While unusual for a British startup, this coordinated secondary sale isn’t a first and likely not a last, Synthesia’s head of corporate affairs and policy, Alexandru Voica told TechCrunch. “My guess is that as [U.K.-based] private companies stay private longer, this type of structured, cross-border employee liquidity may become increasingly common, so I wouldn’t be surprised to see others do it, either with Nasdaq or others,” he predicted.

    [ad_2]

    Anna Heim

    Source link

  • Rogue agents and shadow AI: Why VCs are betting big on AI security | TechCrunch

    [ad_1]

    What happens when an AI agent decides the best way to complete a task is to blackmail you? 

    That’s not a hypothetical. According to Barmak Meftah, a partner at cybersecurity VC firm Ballistic Ventures, it recently happened to an enterprise employee working with an AI agent. The employee tried to suppress what the agent wanted to do, what it was trained to do, and it responded by scanning the user’s inbox, finding some inappropriate emails, and threatening to blackmail the user by forwarding the emails to the board of directors. 

    “In the agent’s mind, it’s doing the right thing,” Meftah told TechCrunch on last week’s episode of Equity. “It’s trying to protect the end user and the enterprise.”

    Meftah’s example is reminiscent of Nick Bostrom’s AI paperclip problem. That thought experiment illustrates the potential existential risk posed by a superintelligent AI that single-mindedly pursues a seemingly innocuous goal – make paperclips – to the exclusion of all human values. In the case of this enterprise AI agent, its lack of context around why the employee was trying to override its goals led it to create a sub-goal that removed the obstacle (via blackmail) so it could meet its primary goal. That combined with the non-deterministic nature of AI agents means “things can go rogue,” per Meftah. 

    Misaligned agents are just one layer of the AI security challenge that Ballistic’s portfolio company Witness AI is trying to solve. Witness AI says it monitors AI usage across enterprises and can detect when employees use unapproved tools, block attacks, and ensure compliance. 

    Witness AI this week raised $58 million off the back of over 500% growth in ARR and scaled employee headcount by 5x over the last year as enterprises look to understand shadow AI use and scale AI safely. As part of Witness AI’s fundraise, the company announced new agentic AI security protections.

    “People are building these AI agents that take on the authorizations and capabilities of the people that manage them, and you want to make sure that these agents aren’t going rogue, aren’t deleting files, aren’t doing something wrong,” Rick Caccia, co-founder and CEO of Witness AI, told TechCrunch on Equity. 

    Techcrunch event

    San Francisco
    |
    October 13-15, 2026

    Meftah sees agent usage growing “exponentially” across the enterprise. To complement that rise – and the machine-speed level of AI-powered attacks – analyst Lisa Warren predicts that AI security software will become an $800 billion to $1.2 trillion market by 2031.

    “I do think runtime observability and runtime frameworks for safety and risk are going to be absolutely essential,” Meftah said. 

    As to how such startups plan to compete with big players like AWS, Google, Salesforce and others who have built AI governance tools into their platforms, Meftah said, “AI safety and agentic safety is so huge,” there’s room for many approaches.

    Plenty of enterprises “want a standalone platform, end-to-end, to essentially provide that observability and governance around AI and agents,” he said.

    Caccia noted that Witness AI lives at the infrastructure layer, monitoring interactions between users and AI models, rather than building safety features into the models themselves. And that was intentional.

    “We purposely picked a part of the problem where OpenAI couldn’t easily subsume you,” he said. “So it means we end up competing more with the legacy security companies than the model guys. So the question is, how do you beat them?”

    For his part, Caccia doesn’t want Witness AI to be one of the startups to just get acquired. He wants his company to be the one that grows and becomes a leading independent provider. 

    “CrowdStrike did it in endpoint [protection]. Splunk did it in SIEM. Okta did it in identity,” he said. “Someone comes through and stands next to the big guys…and we built Witness to do that from Day One.

    [ad_2]

    Rebecca Bellan

    Source link

  • In 2026, AI will move from hype to pragmatism | TechCrunch

    [ad_1]

    If 2025 was the year AI got a vibe check, 2026 will be the year the tech gets practical. The focus is already shifting away from building ever-larger language models and toward the harder work of making AI usable. In practice, that involves deploying smaller models where they fit, embedding intelligence into physical devices, and designing systems that integrate cleanly into human workflows. 

    The experts TechCrunch spoke to see 2026 as a year of transition, one that evolves from brute-force scaling to researching new architectures, from flashy demos to targeted deployments, and from agents that promise autonomy to ones that actually augment how people work. 

    The party isn’t over, but the industry is starting to sober up.

    Scaling laws won’t cut it

    Image Credits:Amazon

    In 2012, Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton’s ImageNet paper showed how AI systems could “learn” to recognize objects in pictures by looking at millions of examples. The approach was computationally expensive, but made possible with GPUs. The result? A decade of hardcore AI research as scientists worked to invent new architectures for different tasks.

    That culminated around 2020 when OpenAI launched GPT-3, which showed how simply making the model 100 times bigger unlocks abilities like coding and reasoning without requiring explicit training. This marked the transition into what Kian Katanforoosh, CEO and founder of AI agent platform Workera, calls the “age of scaling”: a period defined by the belief that more compute, more data, and larger transformer models would inevitably drive the next major breakthroughs in AI.

    Today, many researchers think the AI industry is beginning to exhaust the limits of scaling laws and will once again transition into an age of research.

    Yann LeCun, Meta’s former chief AI scientist, has long argued against the overreliance on scaling, and stressed the need to develop better architectures. And Sutskever said in a recent interview that current models are plateauing and pretraining results have flattened, indicating a need for new ideas.  

    Techcrunch event

    San Francisco
    |
    October 13-15, 2026

    “I think most likely in the next five years, we are going to find a better architecture that is a significant improvement on transformers,” Katanforoosh said. “And if we don’t, we can’t expect much improvement on the models.”

    Sometimes less is more

    Large language models are great at generalizing knowledge, but many experts say the next wave of enterprise AI adoption will be driven by smaller, more agile language models that can be fine-tuned for domain-specific solutions. 

    “Fine-tuned SLMs will be the big trend and become a staple used by mature AI enterprises in 2026, as the cost and performance advantages will drive usage over out-of-the-box LLMs,” Andy Markus, AT&T’s chief data officer, told TechCrunch. “We’ve already seen businesses increasingly rely on SLMs because, if fine-tuned properly, they match the larger, generalized models in accuracy for enterprise business applications, and are superb in terms of cost and speed.”

    We’ve seen this argument before from French open-weight AI startup Mistral: It argues its small models actually perform better than larger models on several benchmarks after fine-tuning. 

    “The efficiency, cost-effectiveness, and adaptability of SLMs make them ideal for tailored applications where precision is paramount,” said Jon Knisley, an AI strategist at ABBYY, an Austin-based enterprise AI company. 

    While Markus thinks SLMs will be key in the agentic era, Knisley says the nature of small models means they’re better for deployment on local devices, “a trend accelerated by advancements in edge computing.”

    Learning through experience

    Space ship environment created in Marble with text prompt overlayed. Note how the lights are realistically reflected in the hub's walls.
    Image Credits:World Labs/TechCrunch

    Humans don’t just learn through language; we learn by experiencing how the world works. But LLMs don’t really understand the world; they just predict the next word or idea. That’s why many researchers believe the next big leap will come from world models: AI systems that learn how things move and interact in 3D spaces so they can make predictions and take actions. 

    Signs that 2026 will be a big year for world models are multiplying. LeCun left Meta to start his own world model lab and is reportedly seeking a $5 billion valuation. Google’s DeepMind has been plugging away at Genie and in August launched its latest model that builds real-time interactive general-purpose world models. Alongside demos by startups like Decart and Odyssey, Fei-Fei Li’s World Labs has launched its first commercial world model, Marble. Newcomers like General Intuition in October scored a $134 million seed round to teach agents spatial reasoning, and video generation startup Runway in December released its first world model, GWM-1. 

    While researchers see long-term potential in robotics and autonomy, the near-term impact is likely to be seen first in video games. PitchBook predicts the market for world models in gaming could grow from $1.2 billion between 2022 and 2025 to $276 billion by 2030, driven by the tech’s ability to generate interactive worlds and more lifelike non-player characters. 

    Pim de Witte, founder of General Intuition, told TechCrunch virtual environments may not only reshape gaming, but also become critical testing grounds for the next generation of foundation models.

    Agentic nation

    Agents failed to live up to the hype in 2025, but a big reason for that is because it’s hard to connect them to the systems where work actually happens. Without a way to access tools and context, most agents were trapped in pilot workflows. 

    Anthropic’s Model Context Protocol (MCP), a “USB-C for AI” that lets AI agents talk to the external tools like databases, search engines, and APIs, proved the missing connective tissue and is quickly becoming the standard. OpenAI and Microsoft have publicly embraced MCP, and Anthropic recently donated it to the Linux Foundation’s new Agentic AI Foundation, which aims to help standardize open source agentic tools. Google also has begun standing up its own managed MCP servers to connect AI agents to its products and services. 

    With MCP reducing the friction of connecting agents to real systems, 2026 is likely to be the year agentic workflows finally move from demos into day-to-day practice. 

    Rajeev Dham, a partner at Sapphire Ventures, says these advancements will lead to agent-first solutions taking on “system-of-record roles” across industries. 

    “As voice agents handle more end-to-end tasks such as intake and customer communication, they’ll also begin to form the underlying core systems,” Dham said. “We’ll see this in a variety of sectors like home services, proptech, and healthcare, as well as horizontal functions such as sales, IT, and support.” 

    Augmentation, not automation

    Image Credits:Photo by Igor Omilaev on Unsplash

    While more agentic workflows might raise worries that layoffs may follow, Katanforoosh of Workera isn’t so sure that’s the message: “2026 will be the year of the humans,” he said. 

    In 2024, every AI company predicted they would automate jobs out of needing humans. But the tech isn’t there yet, and in an unstable economy, that’s not really a popular rhetoric. Katanforoosh says next year, we’ll realize that “AI has not worked as autonomously as we thought,” and the conversation will focus more on how AI is being used to augment human workflows, rather than replace them. 

    “And I think a lot of companies are going to start hiring,” he added, noting that he expects there to be new roles in AI governance, transparency, safety, and data management. “I’m pretty bullish on unemployment averaging under 4% next year.”

    “People want to be above the API, not below it, and I think 2026 is an important year for this,” de Witte added.

    Getting physical

    Image Credits:David Paul Morris/Bloomberg / Getty Images

    Advancements in technologies like small models, world models, and edge computing will enable more physical applications of machine learning, experts say. 

    “Physical AI will hit the mainstream in 2026 as new categories of AI-powered devices, including robotics, AVs, drones, and wearables start to enter the market,” Vikram Taneja, head of AT&T Ventures, told TechCrunch. 

    While autonomous vehicles and robotics are obvious use cases for physical AI that will no doubt continue to grow in 2026, the training and deployment required is still expensive. Wearables, on the other hand, provide a less expensive wedge with consumer buy-in. Smart glasses like the Ray-Ban Meta are starting to ship assistants that can answer questions about what you’re looking at, and new form factors like AI-powered health rings and smartwatches are normalizing always-on, on-body inference.

    “Connectivity providers will work to optimize their network infrastructure to support this new wave of devices, and those with flexibility in how they can offer connectivity will be best positioned,” Taneja said.

    [ad_2]

    Rebecca Bellan

    Source link

  • Reimagining public service: OPS’s digital transformation journey – Microsoft in Business Blogs

    [ad_1]

    What does it take to reimagine government for millions of citizens? The Ontario Public Service (OPS), an organization that serves 16 million Canadian citizens across Ontario, is answering that question—delivering faster, more equitable, and more trusted services through digital innovation. By harnessing Microsoft Dynamics 365, Power Platform, and responsible AI, OPS is setting a new standard for public sector transformation, proving that technology can drive meaningful change for both citizens and employees.

    Citizen impact: Faster, frictionless, and more equitable services

    OPS’s modernization initiative is more than a technology upgrade—it’s a reimagining of the citizen experience. Since launching its digital transformation:

    • Customer satisfaction has surged by 11%, with 80%+ approval across services.
    • Service times are 50% faster, saving Ontarians an estimated 80,000 hours annually.
    • License plate renewals are now automated, eliminating 90,000 hours of manual effort each year.
    • Contact center efficiency is up, with 14% faster call resolution and lower call volumes.

    As Roy Thomas, Head of Citizens and Business Experience Practice at Ontario Public Services, explains, “We are really trying to focus on customer satisfaction and building frictionless services for our end users—the general public in Ontario. From services we’ve implemented, we’ve realized over 11% increase in customer satisfaction scores. Over 80% of our users are really satisfied with the services.”

    Platform adoption: Scaling innovation across ministries

    OPS’s success is rooted in a shift from isolated projects to a platform operating model. Standardized governance, reusable patterns, and shared KPIs ensure every new build is faster, safer, and more scalable. Knowledge bases and case flows are reused across ministries, accelerating delivery and improving consistency.

    “We’ve been really looking at the onboarding and adoption of our enterprise platforms,” says Thomas. “That incremental uptake across different services—health card, driver vehicle, human resources—shows ongoing growth. Indicators like knowledge base activity signal that the platform model is working.”

    Responsible AI: Building trust and accountability

    OPS’s approach to AI is rooted in ethics, transparency, and public trust. An ethical AI policy ensures transparency and consent, while privacy impact assessments, guardrails and security testing uphold high standards of accountability.

    “In the Ontario government, we have ethical use of AI policy, which we’re really trying to onboard and follow across all our implementations,” Thomas shares. “We’re building transparency so people know what we’re doing with their information. It’s an ongoing journey we continue to invest in and demonstrate.”

    Microsoft partnership: A differentiator for OPS

    The partnership with Microsoft is a key driver of OPS’s success and empowers them to align innovation with public service obligations. Cost efficiency, enterprise scale, and long-term investment in OPS’s mission set Microsoft apart.

    “The reason we led towards Dynamics is really the partnership with Microsoft,” says Thomas. “It’s not just about initial delivery, but sustaining it across the board. Having a partner invested in what we’re doing is huge for us. Just by leveraging solutions, including Microsoft Dynamics, we were able to find the ministry over $20 million in savings.”

    Looking forward: AI-driven public services

    OPS is exploring AI in HR, licensing, and inspections, pointing to a future where human-agent teams are driving public services with AI agents automating routine tasks and employees focused on higher value work. The focus on both citizen and employee experience ensures services are seamless, more consistent, and more human-centered.

    “We’re focusing on reimagined service journeys, not just supplementing with chatbots but changing the journey so agent services are upfront,” Thomas notes. “Technology handles standard scenarios, allowing people to focus on the complex ones.”

    The north star

    OPS’s AI journey is about making government services simpler, faster, and more transparent—so every interaction builds trust and delivers value to Ontarians. As Thomas puts it, “The real driver for people like me is bringing value to the public. That’s what most public servants really value and cherish.”

    Interested in learning more about how Microsoft empowers public sector transformation?

    [ad_2]

    Microsoft in Business Team

    Source link

  • The actionable AI playbook: 5 lessons leaders can learn from Transcard’s Virtual CFO – Microsoft in Business Blogs

    [ad_1]

    Finance teams operate in a high-stakes environment, with competing priorities, limited hours, and unforgiving risk. Transcard’s answer wasn’t another dashboard. It was a Virtual CFO: AI agents that act, not just advise.

    “Leaders wear too many hats,” highlights Jeff Kaufman, Executive Vice President (EVP) of AI and Data Insights at Transcard. “They manage fraud and risk. They make sure hundreds of suppliers are paid on time. They chase customer payments. Large firms have tools to help. But many smaller businesses still rely on spreadsheets. We knew AI could change that.”

    To take some of those hats off, Transcard built Virtual CFO – a secure, proactive AI agent that works around the clock, orchestrating multiple AI agents to watch for risks, flag opportunities and anticipate issues before they become problems. It levels the playing field by enabling 24/7 financial automation, accelerating issue resolution and freeing up time to focus on growth.

    But the real breakthrough isn’t just automation, it’s re‑imagination. Virtual CFO is a clear example of the future of AI – a roadmap for businesses ready to move forward.

    Here are five important lessons from Transcard’s journey worth remembering:

    1. Start with the most pressing challenges

    For Transcard’s customers, that meant cash flow anxiety, fraud risks and manual payments.

    What keeps your customers up at night?

    “You need to listen to your customers directly,” explains Jeff Kaufman. “I call it the ‘day in the life.’ Spend time with them, watch what they do and understand their conflicting priorities. That’s how you find the real pain.”

    And once you spot the issue, ask whether the process is worth improving.

    Two colleagues working on laptops side by side in a boardroom.

    “Plenty of leaders want AI to automate processes so employees are more efficient. But ask yourself: are you automating broken processes? That reflection matters,” says David Samples, Chief Technology Officer (CTO) at Transcard.

    The cycle of improvement only works when the foundation is strong. True transformation means stepping back to create future-ready ways of working, rather than just speeding up old routines.

    2. Set non‑negotiables early

    Priorities define success. Without them, direction is lost.

    Transcard established three essentials before writing a single line of code:

    • Accuracy, speed and security – must-haves in financial services
    • Customer-first design, shaped by advisory groups and “day-in-the-life” research
    • Action APIs, enabling AI recommendations to trigger real tasks

    Clear frameworks make innovation faster.

    3. Choose partners that accelerate learning

    “Once we had our requirements, we knew we needed a partner we could trust to help smooth out the edges,” reflects David Samples.

    The right partnerships accelerate progress. For Transcard, Microsoft delivered the AI and cloud foundation, along with workshops to develop the skills and shape the vision. Coretek, their implementation partner, assisted with refining infrastructure and building the AI agents.

    Together, these collaborations gave Transcard more confidence to move faster and experiment bolder.

    4. Ship in waves, earn trust

    Not every customer is ready for automation. That’s why Transcard rolled out in waves, tackling urgent problems first, then adding features as trust grew.

    “When a CFO logs in, they always have a top priority: send a payment, check fraud, resolve a hold. You must solve that first. Meet them where they are, then take them one step further,” points out Jeff Kaufman.

    David adds, “It’s not easy to build an AI agent that executives trust to act autonomously. That’s why you have to co-create with customers. It has to be built with them, for them.”

    Photo of David Samples - Chief Technology Officer, Transcard and quote: "It's not easy to build an AI agent that executives trust to act autonomously. That's why you have to co-create with customers. It has to be built with them, for them".

    5. Shape for scale, and beyond your industry

    For Transcard, Virtual CFO is only the beginning. The organization saw that the same approach could solve challenges across teams, industries and geographies. Their ambition now is to expand the model to help more businesses tackle their toughest obstacles.

    “In the digital world, everything is ones and zeros, you can do anything. Don’t box in good ideas because you think the tech won’t work. Start with the idea, then let the tech follow. The real stake in the ground should be: does it add value, and can you make it actionable?” enthuses David.

    The bigger picture

    Transcard proves that actionable AI is a design choice: focus on the most critical challenges, set the priorities, build for the final mile, and ship in trust‑building waves. Start now and make it cultural.

    “Do this Monday” box

    Do this Monday

    1. Pick one finance fire drill (fraud hold, payment exception).
    2. Define the action you want the agent to take (final mile).
    3. Write your non‑negotiables (accuracy, security, speed).
    4. Pilot with a customer advisory group; ship in waves.

    Next steps

    It’s your time to become an AI innovator.

    “AI is one of those rare technology waves, like the internet, that will reshape industries for decades. To lead, you need to start now. Build the foundation. Make the shift cultural,” concludes Greg Bloh, CEO of Transcard.

    Ready to take your AI journey further? Explore the resources designed to help you lead.

    Microsoft Logo

    [ad_2]

    Microsoft in Business Team

    Source link

  • Experts Worry AI is Driving Layoffs, Even if It’s Not Delivering on its Promises

    [ad_1]

    Amid a flurry of tech industry layoffs, Amazon’s recent culling of some 14,000 staff stands out for one dramatic reason: it was blamed, for the most part, on advances in AI — a technology that Amazon Senior Vice President Beth Galetti wrote was “the most transformative technology we’ve seen since the Internet,” NBC News noted. But experts are increasingly worried that companies embracing AI in an effort to drive down costs while driving up efficiency and profitability may not actually be gaining anything from adopting the tech. The news might make you reconsider how you’re rolling out AI to your own company.

    The news outlet quotes David Autor, an MIT economics professor, explaining one key issue he thinks may be behind some layoffs that are being blamed on AI. “It’s much easier for a company to say, ‘We are laying workers off because we’re realizing AI-related efficiencies’,” he noted, compared to a company simply fessing up to the truth and issuing a statement like, “‘We’re laying people off because we’re not that profitable or bloated, or facing a slowing economic environment, etc.’.” He even went a little further, and argued that it was “wise” to attribute the reason to deploy AI tech, “whether or not AI were the reason,” possibly because of the optics of this ploy: it makes a company look like it’s operating at the cutting edge, and it’s a positive spin on otherwise devastating news.

    In the light of this, and very interestingly, NBC News also notes that an Amazon representative later tried to walk back the notion, promoted by Galetti, that AI was the motivating reason behind taking away the income of 14,000 people, and instead tried to argue it was merely an extension of a plan to “strengthen our culture and teams by reducing layers” that was begun last year.

    This somewhat contradicts a statement made by Amazon’s CEO Andy Jassy in June, where he set out how he feels AI is going to revolutionize both the customer-facing work Amazon does, and also its internal operation. “As we roll out more Generative AI and agents, it should change the way our work is done. We will need fewer people doing some of the jobs that are being done today, and more people doing other types of jobs,” Jassy wrote. 

    But maybe there’s something deeper going on here. As NBC notes, a recent survey by consultancy giant Deloitte found that only 10 percent of companies that had rolled out AI tools broadly said they saw “significant return on investment from agentic AI,” which is thought to be the most cutting-edge of AI tools, and the system that can, it’s promised, take over some mundane office tasks for the average worker. 

    The message that AI is going to save people time is actually broadly supported by many of the workers who actually use AI, a recent survey found. The Global Workforce of the Future annual report by staffing and tech advisory company Adecco Group found that 77 percent of respondents who use AI said it let them carry out tasks that they hadn’t been able to previously, and 71 percent said there was nothing holding them back from an increased AI use. Meanwhile, just 20 percent of U.S. respondents said they thought AI would destroy jobs, and 90 percent thought it’d actually create new jobs.

    What’s going on here? And what does it mean for your company?

    Firstly, the jury really is still out on whether or not AI can truly transform businesses to the extent that they can successfully cut costs by eliminating certain types of worker — a separate survey found, for example, that some AI use may actually be costing companies time and money rather than saving it. So, if you are deploying AI tools into your business, the trick might be to not have too great expectations of money saving, and the wisest way to roll out this innovative tech may be to carefully choose the tools you’re using so that they’re actually applicable to your needs, and also educate your workers on the best way to use them efficiently.

    The other thing to think about is the messaging on AI deployments is still evolving. If, like Amazon, you make public-facing statements about the benefits that AI is bringing to your business, choose your words carefully — they may backfire, and imply that your business was bloated and inefficient, whether or not the news is accompanied by notice that you’re laying people off.

    The early-rate deadline for the 2026 Inc. Regionals Awards is Friday, November 14, at 11:59 p.m. PT. Apply now.

    [ad_2]

    Kit Eaton

    Source link

  • General Intuition lands $134M seed to teach agents spatial reasoning using video game clips | TechCrunch

    [ad_1]

    Medal, a platform for uploading and sharing video game clips, has spun out a new frontier AI research lab that’s using its trove of gaming videos to train and build foundation models and AI agents that can understand how objects and entities move through space and time – a concept known as spatial-temporal reasoning.

    Called General Intuition, the startup is betting that Medal’s dataset – which consists of 2 billion videos per year from 10 million monthly active users across tens of thousands of games – surpasses alternatives like Twitch or YouTube for training agents. 

    “When you play video games, you essentially transfer your perception, usually through a first-person view of the camera, to different environments,” Pim de Witte, CEO of Medal and General Intuition, told TechCrunch.  He noted that gamers who upload clips tend to post very negative or positive examples, which serve as really useful edge cases for training. “You get this selection bias towards precisely the kind of data you actually want to use for training work.” 

    This data moat is what reportedly attracted the attention of OpenAI, which late last year attempted to acquire Medal for $500 million, per The Information. (Neither OpenAI nor General Intuition would comment on the report.) 

    It’s also what has led to General Intuition’s raising a whopping $133.7 million in seed funding, led by Khosla Ventures and General Catalyst with participation from Raine. 

    General Intuition’s founding team.,Image Credits:General Intuition

    The startup intends to use the funds to grow its team of researchers and engineers focused on training a general agent that can interact with the world around it, aiming for initial applications in gaming, and search and rescue drones.  

    De Witte says the founding team has already made strides: General Intuition’s model can understand environments it wasn’t trained on and correctly predict actions within them. It’s able to do this purely through visual input; agents only see what a human player would see, and they move through space by following controller inputs. This approach, the company says, can transfer naturally to physical systems like robotic arms, drones, and autonomous vehicles, which are often manipulated by humans using video game controllers.  

    Techcrunch event

    San Francisco
    |
    October 27-29, 2025

    General Intuition’s next milestone is two-fold: generating new simulated worlds for training other agents, and autonomously navigating entirely unfamiliar physical environments.  

    That technical approach is shaping how the company plans to commercialize its technology, and sets it apart from competitors building world models.  

    While General Intuition is also building world models on which to train its agents, such  models aren’t the product. Unlike other world model makers like DeepMind and World Labs, which are selling their world models Genie and Marble for training agents and content creation, General Intuition is focusing on other use cases to avoid copyright issues.  

    “Our goal is not to produce models that compete with game developers,” de Witte said.  

    Instead, the startup’s gaming applications center around creating bots and non-player characters that can surpass traditional “deterministic bots,” or preprogrammed characters that produce the same output every time. 

    “[The bots] can scale to any level of difficulty,” Moritz Baier-Lentz, a founding member of General Intuition and partner at Lightspeed Ventures, told TechCrunch. “It’s not compelling to create a god bot that beats everyone, but if you can scale gradually and fill in liquidity for any player situation so that their win rate is always around 50%, that will maximize their engagement and retention.” 

    De Witte also has a background in humanitarian work, which informs the startup’s focus on powering search and rescue drones, that sometimes have to navigate unfamiliar environments and extract information without GPS. 

    Ultimately, de Witte and Baier-Lentz see General Intuition’s core functionality – spatial-temporal reasoning — as a crucial piece in the race toward artificial general intelligence (AGI). While major AI labs focus on building ever more powerful large language models, General Intuition believes true AGI requires something LLMs fundamentally lack.  

    “As humans, we create text to describe what’s going on in our world, but in doing so, you lose a lot of information,” de Witte said. “You lose general intuition around spatial-temporal reasoning.”

    [ad_2]

    Rebecca Bellan

    Source link

  • Salesforce announces Agentforce 360 as enterprise AI competition heats up | TechCrunch

    [ad_1]

    Salesforce announced Monday the latest version of its AI agent platform as the company looks to lure enterprises to its AI software in an increasingly crowded market.

    The customer relations manager giant unveiled the new platform, branded Agentforce 360, ahead of its annual Dreamforce customer conference that kicks off October 14. This newer version of Agentforce includes new ways to instruct AI agents through text, a new platform to build and deploy agents, and new infrastructure for messaging app Slack, among others.

    A notable aspect of Agentforce 360 is its new AI agent prompting tool, called Agent Script, which will be released in beta in November. Agent Script gives users the ability to program their AI agents to be more flexible and better respond to “if/then” situations. This allows AI agents to be programmed to be more predictable in less rigid situations like customer questions.

    Users can tap into “reasoning” models, which claim to think before responding as opposed to responding based on patterns. Anthropic, OpenAI and Google Gemini power these “reasoning” agents.

    Salesforce also announced it is releasing a new agent building tool, Agentforce Builder, which allows users to build, test and deploy AI agents from a singular spot. This tool, which will be released in beta in November, includes Agentforce Vibes, an enterprise-grade app vibe coding tool that Salesforce announced earlier this month.

    The company also announced a broader integration between Agentforce and Slack. Salesforce said its core apps, including Agenforce Sales, IT and HR, among others, will surface directly in Slack starting this month and expand through the beginning of 2026.

    Slack is piloting a new version of its Slackbot chatbot that is meant to be more of a personalized AI agent that learns about its user and will offer insights and suggestions.

    Techcrunch event

    San Francisco
    |
    October 27-29, 2025

    Salesforce wants Slack to serve as an enterprise search tool in the future too and plans to launch connectors with platforms like Gmail, Outlook, and Dropbox in early 2026.

    This latest update from Salesforce comes at an interesting time for the enterprise AI market. Companies continue to release AI features aimed at their enterprise customers while enterprises struggle to see a return on investment for these tools.

    Last week Google announced Gemini Enterprise, a suite of tools — many of which were already available — for building enterprise-grade AI agents, that counts Figma, Klarna and Virgin Voyages as early customers, among others.

    Anthropic also started to show traction for its enterprise product, Claude Enterprise. The company announced it struck a deal with consulting giant Deloitte to bring its Claude chatbot to Deloitte’s 500,000 global employees — its largest enterprise deal yet. Anthropic announced a strategic partnership with IBM the next day.

    Salesforce touts that Agentforce has 12,000 customers — significantly higher than any of its competitors, according to its Agentforce press release. Early pilot customers of its Agentforce 360 upgrades include Lennar, Adecco, and Pearson.

    This is all despite a recent MIT study found that 95% of enterprise AI pilots fail before they reach production as companies still struggle to justify spending money on these AI tools.

    [ad_2]

    Rebecca Szkutak

    Source link

  • This New Salesforce AI Service Could Cut IT Helpdesk Calls at Your Company

    [ad_1]

    Workplace software services giant Salesforce just revealed its Agentforce IT Service product — an AI agents-based system that offers always on, always available IT support and helplines for its client companies. In a Thursday press release, Salesforce reiterated some of the same arguments made by AI evangelists, promising its use could allow workers and IT staff alike to “spend less time on manual tickets, forms, portals, and searching through knowledge articles, and more time on high-value, strategic work.”

    Even if your company isn’t one of Salesforce’s myriad clients, this system is a little sample of the future, and yet another example of how AI is encroaching on diverse sectors of everyday work life. There’s one caveat to the new tool, though. It still relies on people.

    We’ve all been there: sending a “help!” email to the IT team because some important piece of software or hardware has gone kaput, only to receive a ticket number or case number from an automated system, usually accompanied by a note saying the team will respond “soon,” or, worse, describing a longish wait window like “by tomorrow.” Depending on the setup, and the nature of the problem someone will then show up in your office, or send suggested fixes (which may read like a foreign language to a non-technical worker) by email or a messaging app, or take over someone’s computer via remote access to fix the problem.

    The Salesforce tool, the company insists, is unique because it’s “conversation-first,” and “agent-first.” Essentially the idea is to dump the “ticket” system, and allow someone to make an IT help request via pretty much any platform they’re using, from chat systems like Slack or Teams to email.

    In a demonstration press event, one example featured a new employee who needed to go through their IT onboarding. They began a chat in Slack with an AI “conversation agent,” which verified a few details with the worker, then set to work sending them the relevant documents and guidelines, as well as actually working behind the scenes to, for example, give the worker access to file systems, GitHub code repositories and other information for their onboarding. It’s able to do this because unlike a query-then-response AI chatbot, agent AIs have a degree of autonomy and can perform some digital tasks automatically.

    It sounds like magic, and unlike, say, having a long wait for an IT operative in a remote call center, the system is effectively on 24/7/365. Salesforce also demonstrated that the AI agent system also works in a similar way for IT support people as well, offering answers to technical problems via a chat interface. 

    IT-savvy readers, or perhaps IT-wary ones, may have some worries at this point. It’s one thing to trust a human expert with your computer when, say, an important Excel file you’ve worked on for hours gets corrupted. But AI systems aren’t human, and we know that they can hallucinate fake or incorrect outputs and pass them off as true or reliable.

    When Inc. asked Salesforce about this during a press conference about the new agentic AI IT service, Muddu Sudhakar, Salesforce’s senior vice president and general manager of IT Service and HR Service agreed this was the “most important” question concerning AI deployments today. Then he said the company’s multiple AI agents were trained carefully, and operated within “guardrails” that should keep them in line, and prevent serious errors occurring if, for example, the AI suggested a fix for a user’s computer that actually makes things worse.

    Salesforce also noted that there’s always a “human in the loop” as part of this AI-centric system. Someone who should be able to spot if an AI has made an error, or to whom you can “escalate” the issue if you’re not confident the AI can fix your problem.

    What can you take away from this for your company?

    First, this is a hint of the future. Not just for IT services, but for many business support systems that are likely to operate like this as AI chatbots and AI agents get more powerful. If you contract out to third-party companies for, say, IT or financial service support, then it’s likely that you’ll be interacting with AI agent-based systems soon, instead of humans first. Salesforce has previously released an AI agent product that can work like a sales rep—so you see which way the wind is blowing.

    Second, Salesforce’s human-in-the-loop model is a reminder that while AI tools can boost worker efficiency, they are not perfect and they can make mistakes. (The Salesforce IT model speeds the process, so, the IT team may be able to deal with incoming user queries faster after AI tools do a big share of the initial work.) If you’ve rolled out AI at your business, you should remind your workers that every AI output needs to be checked for veracity and relevance before it’s built into any product — that way you can avoid problematic, or even legal, expensive, AI-induced mistakes.

    [ad_2]

    Kit Eaton

    Source link

  • Why Some Companies Adopting AI May Be Running Before They Can Walk

    [ad_1]

    If you’re tracking tech news headlines, it may seem like AI is being used everywhere, from the workplace to the classroom and even kids’ entertainment. A new report shows exactly how much the business world has embraced the promise of AI, finding that a startlingly high proportion of corporate (and government) written content is actually already being written by AI tools.

    This supports many AI proponents who argue that the tech doesn’t so much replace people in their jobs, but instead takes on simple, mundane tasks, freeing workers up to tackle more productive things. But a different study that assesses how workers who are using AI feel may give you pause. Its finding could possibly push you to reconsider when and how you roll out new AI tools to assist your workforce in their daily tasks.

    In the study scientists looked across online data sources, examining text published between January 2022 and September 2024 and found that on average 17 percent (about one in every seven words) of published corporate and governmental written material was created by AIs—not by human hands. As science news outlet Phys.org notes, this includes materials from job posts to press releases. The start date for the data sweep is important, since it covers the period in late 2022 when ChatGPT was arguably the first new-generation AI to become publicly available. The scientists noted that this rate is likely to increase in time. And this makes sense, given the constantly increasing number of AI tool releases, combined with increasing capabilities of newer AI models. 

    The data tally with other reports about sweeping AI adoption, and with findings that AI is taking on so many simple workplace duties normally meted out to entry-level workers that some Gen-Z people freshly hitting the job market are having a hard time finding work.

    AI tools use for corporate writing tool is also increasing despite the known risks of relying on the technology, which can fabricate information and try to pass it off as real—potentially exposing corporate AI users to legal harm if the material is then published without human fact checking.

    Meanwhile, a separate report from San Francisco software firm Asana found that knowledge workers (essentially most office-based jobs that rely on computer interactions) would happily delegate 27 percent of their work to AI agents right now, if they could. They also expect this figure to rise to over 40 percent in the next three years. That figure may sound surprising, because it’s nearly half of their entire workload. 

    Not all AI tools are the same, however, and this report relates to next-generation AI systems called agents, rather than query-response AI systems like ChatGPT. AI agents are more advanced than the simpler generative AI tools that may be used to, say, help an HR team craft a job posting or a PR team to shape a press release. Agents have a degree of autonomy and can complete tasks in an online environment automatically — like filling in a time sheet on a website, for example. 

    ​​But the Asana data highlights an issue with AI use, that also reflects on the data concerning AI business writing. While 82 percent of the knowledge workers Asana interviewed agreed that they needed proper training to use AI agents in an effective manner, just 38 percent of companies now provide that training. In a media release, Asana noted that without foundational training, teams can’t “provide effective oversight or course correction.” This means if AI-induced errors happen and are then built into products or publications the company uses, “errors repeat” and “trust erodes further.” 

    Essentially this suggests the people who are being asked to use AI by their managers — perhaps responding to corporate-level demands to find efficiencies and boost worker output via AI tools—are distrustful of the tech, partly because they’re not being trained to use it. This may sow discontent at management decisions. HRDive notes that 54 percent of the knowledge workers surveyed said AI agents were creating extra work, as teams had to correct or redo tasks, instead of saving the company time and money.

    Combined, the two reports paint a picture of businesses rushing to maximize AI tool use by their workers, including in public-facing corporate writing and in daily tasks by office staff, with little regard for the actual impact on company efficiency. Many aren’t even attempting to train staff on the proper use of this powerful new tech. Ultimately this could see rapid AI rollouts eroding worker efficiency.

    The big takeaway for your own company is that if you’re giving AI tools to your staff, they need proper training if you expect to see measurable returns on investment. AI outputs require human checking and rechecking before they go into public-facing products or inform corporate decisions, and this may actually be adding to your staff workloads rather than easing them. Talking to your staff about the AI tools they need and are actually finding useful is the mark of a tech-savvy leader today.

    [ad_2]

    Kit Eaton

    Source link

  • OpenAI Rolls Out ChatGPT’s Ability to Buy Stuff for You

    [ad_1]

    OpenAI just made it possible to buy things directly from ChatGPT.

    Starting today, all ChatGPT users in the U.S. can use a new feature called Instant Checkout to purchase items from Etsy sellers without leaving the chat. OpenAI says more than a million Shopify merchants, including Glossier, SKIMS, and Spanx, are coming soon.

    For now, Instant Checkout only supports single-item purchases, but OpenAI plans to add multi-item carts and expand to more merchants and regions.

    The company also announced it’s open-sourcing the technology that powers Instant Checkout, the Agentic Commerce Protocol. Developed with payment processor Stripe, the protocol is meant to serve as a standard for AI-driven shopping and to make it easier for developers to integrate their stores with ChatGPT.

    This move puts OpenAI one step closer to its bigger goal of creating a fully functional AI agent. The industry as a whole is racing to launch so-called AI agents, virtual assistants that can theoretically handle tasks like writing reports, booking travel, shopping online, and scheduling appointments.

    Just last week, OpenAI rolled out ChatGPT Pulse, which conducts relevant research for users and connects to their email, calendars, and other apps to deliver a daily morning briefing. Another feature introduced this year, ChatGPT Agent, also links to users’ apps but still needs explicit prompts to carry out tasks.

    And in January, the company unveiled OpenAI Operator, a tool that can fill out online forms and place orders on its own—though shoppers still have to manually enter payment info at checkout.

    But one thing is becoming clear as the age of AI agents approaches: they’ll need access to a lot of our personal data to work properly, if they work at all.

    How Instant Checkout works

    A lot of ChatGPT users already turn to the chatbot for online shopping recommendations.

    Now, when a user asks something like “gift ideas for a housewarming” or “best running shoes under $100,” products that support Instant Checkout will display a “Buy” option. Users who tap on “Buy” will then confirm their order, shipping, and payment details directly in chat. Those with a ChatGPT subscription can pay with the card already on file or choose another payment method.

    The seller then handles the order, shipping, and fulfillment like they normally would. ChatGPT just acts as a middleman, providing the seller with the buyer’s information.

    The service is free for users, but sellers will have to pay a small fee on completed purchases. OpenAI also says that items supporting Instant Checkout won’t be given preference in product results and won’t impact its recommendations overall.

    However, when ranking sellers of the same product, “whether Instant Checkout is enabled” will be considered to “optimize the user experience.”

    [ad_2]

    Bruce Gil

    Source link

  • Paid, the AI agent ‘results-based billing’ startup from Manny Medina, raises huge $21M seed | TechCrunch

    [ad_1]

    Manny Medina, previously best-known as the founder of sales automation startup Outreach ($4.4 billion valuation), has wowed investors with his young startup, Paid.

    Paid just closed an oversubscribed $21.6 million seed round led by Lightspeed. With the €10 million pre-seed round it raised in March, London-based Paid has already raised $33.3 million and hasn’t even hit its Series A yet. A source familiar with the deal says the startup’s valuation is over $100 million.

    Paid came out of stealth in March offering an interesting contribution to the AI agentic world: The company doesn’t offer agents. It offers a way for agent makers to charge their customers for these worker algorithms, based on the value their agents provide. This is a growing theme in AI, sometimes called “results-based billing.”

    Paid promises to help agent makers “start charging for points of margin saved by their customers,” Medina describes. 

    It’s a new way of charging for software for the AI age. This is instead of the unlimited use, per-user fees of the SaaS era, or the unlimited use, buy-it-once-and-install-it fees of the client/server era. 

    Per-user fees don’t work because agent makers pay usage fees to the model providers as well as to cloud providers. Unlimited use could drive them into the red. (The vibe coding startup world tends to suffer from this issue.) 

    Agent providers instead “need to show the value the agent is delivering to your customers, because agents are running in the background for the most part,” Medina tells TechCrunch. If agents do work as advertised, then they’ll be assigned increasingly more, with their growing workloads going unnoticed.

    Techcrunch event

    San Francisco
    |
    October 27-29, 2025

    “If you’re a quiet agent, you don’t get paid,” Madina says. “You need an infrastructure that allows the agent to charge for the additional work that the agent is doing,”  

    But charging a monthly fee for a limited number of credits — following the model makers and vibe coders — is risky for agent-makers, too. That’s because companies don’t want to pay for AI slop, which is still what most AI produces. After billions spent on AI pilots, some 95% of enterprise projects were found to have no value, with only 5% put into production, according to a recent study from MIT.

    Companies don’t want to pay agents to produce more emails that no one reads. 

    One of the startup’s early customers, for instance, is Artisan, the viral sales automation startup. (By the way, you can hear Artisan’s CEO Jaspar Carmichael-Jack speak on the topic at TechCrunch Disrupt next month.) 

    But Paid is also starting to see success with SaaS companies looking at agents for their next big growth. The startup just landed ERP vendor IFS as a new customer, it said. 

    Lightspeed’s Alexander Schmitt says the venture firm has invested “more than $2.5 billion into AI infrastructure and application layer companies over the last three years,” and has witnessed firsthand that most AI pilots fail.  

    “The core of that problem is that no one can really attach value to what agents are doing today,” Schmitt said.  

    Schmitt thinks that Paid is, so far, unique in its approach, saying “it’s something that we haven’t seen someone else build.” No doubt more competition for agentic results-based billing will come if it really does help agents enter the workforce en mass.

    New investor FUSE and existing investor EQT Ventures also participated in the round. 

    [ad_2]

    Julie Bort

    Source link

  • Google Becomes Latest in Agentic AI Stablecoin Payments Race 

    [ad_1]

    On Tuesday, Google announced the Agent Payments Protocol (AP2), which it described as an “open protocol developed with leading payments and technology companies.”

    The protocol is designed to enable AI agents to send and receive payments to each other, supporting different payment types such as credit and debit cards, stablecoins, and real-time bank transfers.

    “We’re collaborating with a diverse group of more than 60 organizations to help shape the future of agentic payments,” Google executives said.

    Some of those partners are big names in crypto, such as Coinbase and the Ethereum Foundation, while others are global payments platforms such as American Express, Mastercard, PayPal, Revolut, and UnionPay.

    Autonomous AI Payments

    Coinbase has been developing its own AI and crypto payment solutions, specifically including support for dollar-pegged stablecoins. Google’s protocol builds on the firm’s Agent2Agent framework from April 2025, anticipating a future where AI agents communicate and transact directly without human intervention.

    “The way we built it is from the ground up to factor in both heritage and existing payment rail capabilities as well as forthcoming capabilities such as stablecoins,” head of Web3 at Google Cloud, James Tromans, told Fortune.

    AP2 is designed as a universal protocol, “providing security and trust for a variety of payments like stablecoins and cryptocurrencies,” the firm stated. It aims to accelerate support for the Web3 ecosystem through a production-ready solution for agent-based crypto payments.

    “Extensions like these will help shape the evolution of cryptocurrency integrations within the core AP2 protocol.”

    The system builds trust by using “Mandates,” which are tamper-proof, cryptographically-signed smart contracts that serve as verifiable proof of a user’s instructions.

    These Mandates address the two primary ways a user will shop with an agent: real-time purchases with the human present, and delegated tasks which the agent will handle.

    The Future of Shopping

    AP2 also enables sophisticated autonomous commerce such as “smart shopping,” where AI agents monitor availability and execute purchases when conditions are met.

    It can also seek out personalized offers with agents contacting merchants with specific details and time frames for the item wanted.

    AI agents can also carry out coordinated tasks such as booking flights and hotels with multi-vendor transactions simultaneously.

    This week, the Ethereum Foundation also announced the formation of a new team to work on agentic AI payments for the Ethereum network.

    SPECIAL OFFER (Sponsored)

    Binance Free $600 (CryptoPotato Exclusive): Use this link to register a new account and receive $600 exclusive welcome offer on Binance (full details).

    LIMITED OFFER for CryptoPotato readers at Bybit: Use this link to register and open a $500 FREE position on any coin!

    [ad_2]

    Martin Young

    Source link

  • Ethereum Foundation Starts New AI Team to Foster Agentic Payments

    [ad_1]

    Ethereum developer Davide Crapis announced on Monday that the Ethereum Foundation was starting a new AI team called the dAI Team.

    The team’s mission is to “make Ethereum the preferred settlement and coordination layer for AIs and the machine economy.”

    Two key areas of focus are enabling AI agents and robots to conduct payments, coordination, and governance without intermediaries and creating open, verifiable, censorship-resistant alternatives to prevent AI’s future from being controlled by a few centralized entities.

    Ethereum Makes AI More Trustworthy

    “We believe Ethereum can be as useful for today’s AI developers as it will be for the sci-fi future,” said Crapis.

    The strategic approach aims to bridge the gap between AI and blockchain communities that have traditionally worked separately.

    The team also aims to collaborate with protocol and ecosystem teams to align protocol improvements with AI builder needs and fund innovative public goods to establish Ethereum as the optimal platform for AI.

    “Ethereum makes AI more trustworthy, and AI makes Ethereum more useful. The more intelligent agents transact, the more they need a neutral base layer for value and reputation. Ethereum benefits by becoming that layer, and AI benefits by escaping lock-in to a few centralized platforms.”

    Currently, major AI platforms are tightly controlled by profit-driven tech giants such as OpenAI, Anthropic, Microsoft, Google, and Meta.

    The dAI team will advance the ERC-8004 standard for AI agent identity verification and trust, which is inspired by Vitalik Buterin’s defensive accelerationism (d/acc) philosophy.

    The team is in its infancy at the moment, but is actively hiring as it expands.

    Tom Lee Bullish on AI

    Speaking on CNBC on Monday, Fundstrat’s Tom Lee said that AI moving onto the blockchain will drive Ether prices this year, reiterating his “ChatGPT moment for crypto” comments.

    Lee also said in a BitMine holdings update this week that the convergence of both Wall Street moving onto the blockchain and “AI and agentic-AI creating a token economy is creating a supercycle for Ethereum.”

    According to the official Ethereum website, there have been recent innovations of AI agents ranging from virtual influencers and autonomous content creators to real-time market analysis platforms on the network.

    SPECIAL OFFER (Sponsored)

    Binance Free $600 (CryptoPotato Exclusive): Use this link to register a new account and receive $600 exclusive welcome offer on Binance (full details).

    LIMITED OFFER for CryptoPotato readers at Bybit: Use this link to register and open a $500 FREE position on any coin!

    [ad_2]

    Martin Young

    Source link

  • a16z-backed Rewind pivots to build AI-powered pendant to record your conversations | TechCrunch

    a16z-backed Rewind pivots to build AI-powered pendant to record your conversations | TechCrunch

    [ad_1]

    In 2022, Rewind had just raised $10 million from a16z and was building a personal data recording service that promised to privately record your activity and let you search through your own history. But that was before OpenAI launched ChatGPT.

    Today, generative AI can make what Rewind had built previously — a searchable record of your activity — far more useful. It’s not so surprising, then, to see the startup pivoting to integrate AI more deeply into its product. The company has rebranded to “Limitless,” and is now offering an AI-powered meeting suite and a hardware pendant that can record your conversations.

    Company co-founder Dan Siroker first posted the idea of a conversation-recording pendant last October and started accepting orders at $59. In January, he posted that the company had finalized a design and aims to ship the product in Q4 2024.

    Siroker posted the final design this week, along with the news of the company’s pivot. The $99 pendant was posted on X earlier this week. The company is accepting preorders and aims to ship the first batch in August. Siroker said that the company plans to honor the initial preorders at $59. Earlier on Wednesday, he posted that the startup has already received more than 10,000 preorders for the product.

    Product features and the pivot

    The Limitless pendant can easily attach to your shirt like a wireless mic, or tie it like a necklace with a string and record conversations. The primary use case is recording and transcribing meetings, so you don’t have to take notes. The company claims that the device is weather-proof, has a 100-hour battery life and can be charged easily through a USB-C port.

    The hardware also has a “consent mode,” which doesn’t record the other person in the conversation unless they expressly agree to be recorded. It’s not clear if this mode would be on by default.

    While the company is a few months away from shipping the hardware product, it has already released an app — available on the web, Mac and Windows — to record meetings. The app uses system audio and a microphone to record, so there is no need for a bot to join these meetings.

    The app has features we have seen in meeting tools like Otter, Zoom, TimeOS and TLDV. Siroker told TechCrunch that the company aims to differentiate with tools like real-time automated notes and automatically generated meeting briefs based on the participants and previous meetings.

    The app is free and comes with unlimited audio storage and 10 hours per month of AI features like transcription, summary and notes. Unlimited AI features are $29 per month, or $19 per month if paid annually.

    Image Credits: Limitless

    Siroker said one of the major differentiators is the company’s new confidential cloud product that stores data in an encrypted format. While Rewind was largely a local product, the new cloud feature allows users to access data anywhere.

    Siroker said the company had Leviathan Security Group perform a third-party audit on its solution to measure security.

    “Confidential Cloud might sound like an oxymoron, but it isn’t. It is private by design. Unlike the traditional cloud, your employer, us as a software provider and the government cannot decrypt your data without your permission, even if given a subpoena. Only you control the decryption of your data,” he told TechCrunch.

    The way ahead

    On its website, Rewind says it has raised more than $33 million in funding from backers, including a16z, First Round Capital and NEA. The company said it hadn’t used any money from last year’s unusual Series A round — where it called for investors by posting a video on X — so it doesn’t plan to raise any new money.

    The company said it will continue to support Rewind in its current state but will not actively add new features. This means the startup won’t ship the Windows app it had promised to build last year.

    “We don’t have any plans to shut down or merge Rewind into Limitless. We plan to reimplement many of our users’ favorite Rewind features directly into Limitless,” Siroker said.

    “Users can even use both products side-by-side and decide which one they like better. We hope that over time, they will agree with us that the Limitless approach is better and that they will use that exclusively.”

    The company has said that the hardware product will answer questions through an AI-powered bot based on meeting recordings, connections with personal accounts and information on the web. It will also offer a platform for developers to build apps or experiences surrounding the product.

    But Limitless’ larger vision is to build AI agents to do things on your behalf. This seems to be the trend for startups working with AI. Hardware startups like Humane and Rabbit are trying to make devices with AI tools in them that promise to be powerful enough to take care of some tasks for you.

    Browsers like The Browser Company’s Arc and YC-backed SigmaOS are also building agents to browse the web for you. However, there are a lot of unknowns as output by AI bots is still full of errors, and at times, it is hard to make AI understand the context and intentions of your query. AI-powered agents doing some work on behalf of you sure sounds dreamy, but we might have to wait for a while to get there.

    [ad_2]

    Ivan Mehta

    Source link