There is some chatter about a recent report "Thousands of CEOs just admitted AI had no impact on employment or productivity"[1].
This is being interpreted as AI not having the economic effect the pundits predicted. However, having had one career in enterprise and another in startups, I look at this differently. When people talk about AI making them "more productive", they're usually conflating three very different things. It's worth pulling them apart:
- Faster Horses: Doing the same thing, just quicker.
- Zero Sum Games: Doing it because everyone else is.
- Actual Productivity: Building real output faster or with fewer inputs.
The thing is, the first two result in no net productivity gains. In fact, you might see an overall reduction in some cases.
Faster Horses[2]
This is improving what you already do. Sometimes great, but not everything we do is necessarily productive.
For example: managers use AI to generate Performance Reviews. HR uses AI to summarize. Much faster, but is it more productive? Arguably (very probably), no. My co-founder Joris posted about "Workslop" recently..
Many processes in business are very inefficient, and sometimes that friction has evolved for good reason. Maybe it's really just not that important? Clearing your product backlog sounds great... unless they're all terrible ideas. Which leads to the question: why are these tickets languishing in the backlog in the first place?
To put it another way, if your business, industry, or processes already have fundamental efficiency issues, speeding it up isn't going to lead to a productivity boon.
Upshot: In these cases, throwing AI at it isn't the answer. A total reboot is the revolution. Do you need a backlog any more? Do you need performance reviews? Maybe we don't need Pull Requests (or all the time)? How can this new tech remove the need for things rather than bolster them?
Zero Sum
This is where you need to adopt something simply to stay in the race.
For example: AI recruiting might give early adopters a significant edge, leading to an arms race where everyone ends up using the same tools. Or AI-generated marketing content; once your competitors are flooding every channel, you have to match them just to stay visible.
But talent and attention are finite resources. No net "productivity" is gained across the industry, yet you're at a severe disadvantage if you're not investing in these tools.
Upshot: Fast-following is really effective here. If you're on the bleeding edge the investment is big, but the payoff isn't necessarily higher than someone coming in later.
Actual Productivity
This is what we're actually striving for. You can use AI to do things that were previously impractical, or at a scale that was unachievable.
Examples:
- Small teams shipping products that previously required large organizations. A 3-person startup building what used to need 30 engineers.
- Real-time translation enabling companies to serve markets they previously couldn't enter at all.
- Analyzing customer support transcripts at scale to surface product issues that would have gone unnoticed for months.
- Drug discovery and protein folding; problems that were computationally intractable before.
Upshot: The real question isn't "how much faster are we?" It's "what can we do now that we couldn't before?"
Closing out
This is why I believe those CEOs reported no productivity gains. Most AI investment right now is going into faster horses and zero-sum arms races — categories that don't move the needle.
The trap is that faster horses is the easiest to measure and the easiest to sell. "We generated 40% more Pull Requests this quarter." But if those PRs are patching that you could have just ignored, you haven't gained much. For HR it might be Performance Reviews. There is busy-work everywhere.
The real question is whether we need these things at all, or whether AI lets us rethink them entirely.