I use AI tools every day, and I’m not writing this to tell anyone to stop. They save me hours on dull work. So I came to two reports this week expecting the usual cheerleading. What I found was more useful. The people closest to the work are saying the payoff hasn’t shown up yet.
That’s the claim I’ll defend. The AI productivity story is running well ahead of the evidence.
Not slightly ahead. Well ahead.
Ask who is actually saving time
Bloomberg Law asked more than 1,000 lawyers about AI for its latest AI in Practice report. Lawyers were supposed to be first in line. Contracts, discovery, research memos, all of it text, all of it supposedly ripe for a machine. Yet 76% of them worry about AI going rogue, and a quarter fear it’ll take their job. That isn’t how a profession talks when it’s quietly banking big gains.
Bloomberg Law’s analysts saw this coming a year ago. In an earlier survey only 23% of in-house lawyers used AI tools daily, and a third of regular users said it saved them less than 30 minutes a day. Half an hour, at best. Then some of that half hour goes on checking the output, because one invented case citation can end a career.
Look at what employees admit
Visier sells workforce AI software. It has every reason to talk the market up. Its 2027 trends report, out this week, opens with the words “Beyond the AI Hype”. Its own research found that 48% of employees admit to exaggerating how much they use AI or how good they are at it, and 59% say their employer has no clear plan for how AI will change their jobs, or they don’t know of one.
Think about what that does to every adoption survey you’ve read. If half the workforce overstates its AI use to look current, the adoption figures are inflated at the source. So is any return-on-investment math stacked on top of them.
Visier also expects that putting AI agents on the org chart will backfire, and that companies will rethink AI-driven layoffs. Firms that rehired people they’d let go paid them about 3% more to come back. That’s the bill for a bet that didn’t land.
What I still don’t buy
None of this makes the tools useless. I don’t buy the opposite story either, the one where AI is a bubble with nothing inside. Lawyers are using it. Companies are rebuilding work around it. Something real is going on.
But the gains so far are small and uneven, and they cost money to verify. Executives who promised their boards a step change in 2026 are about to explain why it looks more like a gentle slope.
That’s fine. It just isn’t what they sold.