Tag: productivity theater

There are two trends percolating in the tech workplace that are about to collide head on in the worst way possible. 

The first trend is something you might be familiar with, and I call it “AI productivity theater.”

It works like this: When a company adopts AI out of FOMO or for marketability (“Now powered by AI!”), the mandate from leadership usually boils down to the very basic “use more AI and be more productive!” Not too long after those mandates, savvy employees figured out that all they really had to do was fire up a chatbot and “look more productive.”

“Oh wait, you can make a presentation for me? Cool. Let’s do that! What do I want to present? I don’t know, you tell me!”

You’re either laughing with me or you just got seen.

The second trend is a good trend. It appears that finally, anecdotally, and ever-so-slightly, AI adoption is actually starting to cause some good old-fashioned productivity – the measurable kind with dollar signs attached to it.

How did that happen? Well, here’s my theory. Remember way back a few weeks ago when those eggheads at MIT told us that 95 percent of enterprise AI pilots had zero measurable ROI? I think a few clever folks, contrarians like myself, took that news the other way.

“Wait. You mean to tell me that 5 percent of enterprise AI pilots had measurable ROI? What the hell did they do?”

“Oh wait. AI is a tool. Like spreadsheets. I can use spreadsheets!”

Yeah! It’s good news. The problem is, most companies can’t tell the difference between real AI productivity and AI productivity theater. And that’s going to waste another spectacular amount of money.

What Is AI Productivity Theater?

You know it when you see it. Actually, you know when you’re told that productivity is super-happening but not how much or why or how it can be measured and replicated.

Look, a lot of AI mandates came from boards, who were scrutinizing ugly company balance sheets against perceived AI productivity promises. Those mandates were filtered through leadership, down to management, and like a game of password, by the time it got to the rank-and-file employee it had become, “You need to use this shit.”

And most of “this shit” was chatbots.

I’ve talked to those board members, those leaders, those managers, and a load of those rank-and-file employees. I’ve consulted for many of those companies. I don’t want to out them, but I do love the mental image of so many company leaders reading this, thinking I’m talking about them, and casting the stinkeye at their cubicle farms for snitching. So here we go:

“By the end of 2024, we hit 95 percent AI adoption across the organization.”

That’s real. That was a leadership message to one of the rank-and-files I know. And according to her, that meant that almost the entire company was using ChatGPT at more than the free level.

The initial metrics were bogus and a lot of them stayed that way. Whether it’s the amount of time spent in a chat, the number of conversations, the lines of code generated, the ham-handed integration into existing workflows, or even the building of an “Ai-first culture,” the metric to show productivity gains from AI was disguised as a metric to show the rate of AI adoption.

There’s even an IBM commercial mocking this. And when IBM is mocking you, you’re doing tech wrong.

But what about the 5 percent who were actually generating real productivity gains?

The Solution Is Clear

It’s right there in the MIT study. Right in the executive summary!

“This divide does not seem to be driven by model quality or regulation, but seems to be determined by approach.”

“These early results suggest that learning-capable systems, when targeted at specific processes, can deliver real value, even without major organizational restructuring.”

Oops. We did the other thing. And fired everybody. 

Well, OK, if we know the approach was wrong, then what’s the right approach? Again, all we need to do is go back to that tool thing. 

In the right hands, used the right way for the right use cases – and by “right” I mean “specific,” not 95 percent of the org – AI can be an amazing productivity booster. 

We Have a Second Chance At the Approach

If we’re going to harvest real productivity gains, the first thing we need to do is re-establish trust between leadership and employee as it concerns the adoption and productivity expectations that should be coming from these tools.

Then we need to understand this unflinching truth: The more experienced the employee, the more productive they’re going to be with AI tools. Giving a junior employee ChatPGT and telling them to be more productive is like giving them a blank spreadsheet and telling them to come up with next fiscal year’s budget. 

And then we need to start hiring and training more junior employees before we run out of senior employees, and they take all the expertise to create gains from these tools out the door with them. 

The next phase of mass AI adoption – which looks to be prescriptive, predictive, autonomous, and agentic AI – this phase is going to make chatbots look even more like spreadsheets.

If we take the AI productivity theater gains as real, without measurable validation, which is tempting because there are more “gains” to report across more parts of the entire company, we’re going to reward the wrong behavior, and give some very good actors some very rich new material. 

Then, when autonomous AI is making decisions instead of suggestions, the difference between real AI productivity and AI productivity theater isn’t just wasted budgets. It’s systems that auto-approve the wrong invoices, supply chains that optimize for the wrong variables, customer service that escalates the wrong issues, and predictive models that predict the wrong things because they were trained on data derived from feelgood theater.

The data is finally showing up. The productivity gains are real. If we enable AI productivity theater now, we’re screwed when autonomous AI starts making decisions based on a thumbs up and a smile.

If you found this to be honest and actionable, please join my email list, because I’m always honest and sometimes – not all the time – I’m actionable.