Tag: you don’t need AI

If you’re absolutely sick and tired of hearing about how much you need AI, you’re not alone.

But the AI hype and the AI FOMO is far from over. In fact, you’re in for another battle within the next few months. And you need to stand your ground.

The AI sales cycle is shifting from hawking the productivity magic of generative AI to hawking the practical magic of autonomous and predictive AI – real AI, if you will. I’ve been waiting for this moment, because this is my kind of AI. Useful AI, active AI, valuable AI.

If… and it’s a very big “if” that I’ll get into in a minute.

That said, if the last three years have taught us anything about how “AI” gets positioned to an already profit-focused and innovation-starved tech industry, I’m not super psyched about the case that’s going to be made for this new, mandatory, company-saving, necessary flavor of AI.

So as someone who has been building AI since before AI was cool, I want to be the first to raise my hand and let you in on a little secret.

You still don’t need AI.

A Workforce Beaten Over the Head With LLMs

The peak of the generative AI hype cycle is in our rear-view, but it was only maybe six months ago when we were still all hearing:

• You’re going to be replaced by AI.
• You’re going to be replaced by the guy who knows how to use AI.
• You’re going to be replaced by an AI who can use AI far more efficiently than you.

In fact, I’d venture a guess that a lot of you are still hearing this. 

Then, like a needle scratch on a party record, studies like the infamous MIT “95 percent of AI pilots saw no ROI” put the official nerd stamp on what a lot of us were already feeling and even saying out loud: These are tools, not “thinking machines.” The increase in productivity that they bring to an organization is directly parallel to the level of experience and talent of the human using them.

But these nascent tools were sold as magic boxes for which you could plop anyone in front of their now-magic keyboards – people like completely inexperienced junior developers or completely non-technical executive team members – and get expert-tier results for any initiative, across any discipline, for any industry.

Dude. You could literally “vibe” code the next Shopify. Literally. Code. With vibes.

So that’s what Corporate America did. They right-sized centuries of combined institutional knowledge from the org to become more nimble. They burned billions of dollars on “infrastructure” for chatbots and AI-first cultures. And then they waited to be propelled to the top of their market – ready to “leave in the dust” all their competition, whom they were told weren’t using AI, but who were actually using the exact same AI they were using. 

When the ROI didn’t materialize, a lot of AI hucksters disappeared, and a lot of AI marketing language was “softened” to reflect reality. 

And here we are.

My Expert Conclusion: You Don’t Need AI

Before we go any further, I want this to be perfectly clear: 

AI, even generative AI, is a great advancement in technology. I’ve been working with it professionally since 2010, when I helped invent the first commercially available generative AI platform, and I’ve been using concepts of AI since the last freaking century, when I was building some of the first recommendation engines. I’ve since adapted to the advanced generative AI tools – LLMs and GPTS – and I’m maximizing my productivity with them. Yes, there is a Joebot. No, you can’t use it.

But my main focus since 2015 has been on automation and predictive AI. If generative AI is the one that speaks knowledge, autonomous and predictive AI are the ones that act on that knowledge.

Over the last year, I’ve gone full time consulting on all the kinds of AI, and I’ve noticed over the last six months or so that I’ve again had to resort to the same expert answer I was giving back in 2021, when everyone from VC-backed startups to PE-backed scale-ups to publicly traded tech firms were knocking on my door, asking for my help and experience integrating AI the “right way.”

That answer: “Well, here’s the thing. You don’t need AI implemented the ‘right way.’ In fact, you don’t need AI at all.”

You’d be surprised how shocked these folks were. Some of them even got mad at me. 

It’s about to happen all over again. Stand firm, soldiers.

The Snake Oil Sales Pitch for AI Automation

Earlier when I said that Generative AI “speaks” knowledge and autonomous and predictive AI “act” on that knowledge, notice I didn’t say that either kind of AI “was” knowledge, or that “knowledge” was included for free like a large fries in a value meal. 

Knowledge has nothing to do with the science of AI. Knowledge is institutional. 

The only reason modern generative AI can “speak” knowledge is because that knowledge was ripped from the entirety of unstructured data housed on the internet. When I was working with the first generative AI models back in 2010, we limited our “knowledge” to our client’s proprietary, institutional data, and we worked with their most senior people to train the models.

Thus, we couldn’t say much – like we didn’t produce sycophantic “you’re the greatest” drivel, although we did make some decent jokes every now and then. No, we couldn’t blather on, but everything we said was accurate and knowledgeable, which meant our clients could take action based on our “spoken” knowledge of their institutional data.

That’s valuable.

But when they come for your wallets with autonomous and predictive AI, they’re going to skip over the institutional knowledge, and “free large fries” their way into your workflows.

“Don’t worry, the AI will learn on its own.”

You Need Data, Knowledge, and Automation

Here’s that “if” I promised earlier.

Autonomous and predictive AI without a base of institutional knowledge is like a robot with a machine gun that hasn’t been taught how to aim. 

That’s what’s about to be let loose on your workforce.

You need data. Mounds of it. When we built for Yahoo, we had millions of fantasy outcomes to analyze each week. When we built for the Associated Press, we had decades of quarterly earnings reports data to sift through and learn from. At Spiffy, where we did mobile vehicle maintenance, our techs used a proprietary app which generated thousands of data points for every service. 

This is the AI candy that lets loose the AI kid in the AI candy store.

But not all data is structured. So you need to be able to merge structured and unstructured data into a single store, or warehouse, or lake, or ocean, or whatever you want to call it. You need systems that constantly update that data, learn from it, and make it available immediately. 

You need to quantify and digitize processes and actions into repeatable workflows, catching every outlier whether that outlier has already happened or it might someday.

And you need sharp, experienced people with the institutional knowledge to bring all of this together. Then, they need to automate all that.

Then with all of that in place, only then should you start listening to pitches about new AI, and how much you need to dive into it before all your competitors or that guy in the next cubicle gets there first.

Please join the rebel alliance of over 10K tech professionals on my email list. I’ll keep bringing the institutional knowledge gleaned from a lifetime wrestling with the business side of tech.