The AI Implementation Report is a fortnightly newsletter for leaders navigating AI strategy. Each issue highlights the research, data and trends that matter most for implementation, not the hype, not the doom, just what's actually working and what isn't. Curated by Mark Phelps and Drew Horton. A Mark Phelps LLC production.
In Issue #2 we cited Glean's survey of 6,000 digital workers: 11 hours a week saved, and only 13% of their firms seeing performance improve. We left something out. The rest of that study explains where the 11 hours actually goes, and it isn't back into the business.
THE ONE TO READ
If you only read one AI implementation article before your next leadership meeting, make it this one.
Workers save 11 hours a week with AI. They spend 6.4 of them getting the output usable. The gap is not a training problem
The AI Productivity Trap Hiding Behind Every Time-Saved Stat — UNLEASH (2026) | FREE
87% of digital workers now use AI. 75% say it makes them more productive, and they're not wrong. Glean's Work AI Institute measured roughly 11 hours a week saved, more than a quarter of the working week.
Then they measured what happens next. 6.4 of those hours go to botsitting: the waiting, re-prompting, correcting and re-running required before AI output is fit to use. More than half the dividend, spent supervising the thing that was supposed to save the time.
The number that should stop you is further down. 69% admit to shipping work they haven't verified, don't fully understand or couldn't confidently stand behind. 41% say they sometimes deliver work they couldn't explain if someone asked them to walk through it.
Four in ten! Read it again with your own org in mind.
Rebecca Hinds, who runs the Work AI Institute, puts it plainly. If the productivity dividend gets spent this way, the work wasn't eliminated. Firms have just built "a new layer of overhead." Worth noting Glean sells enterprise AI search, so they have commercial skin here. It cuts the other way though. A vendor publishing data that undercuts its own category pitch is usually telling you something real.
The detail that stayed with us: that last behaviour, which Glean calls botshitting, runs highest at organizations that cut headcount and named AI as the reason. Which tracks if you think about it. Tell a team the tool is why their colleagues are gone, then hand them the tool and the departed workload. The rational move is to ship whatever the model produces and hope nobody looks too closely. That isn't a character flaw or the team’s fault. It's what the incentive structure asks for.
So here's the thing, and it's the whole point of this issue.
The instinct is to fix the output. Add a review gate, hire a checker, write a policy. That treats botsitting as a quality problem and just relocates the overhead to a different line on the org chart.
The study points somewhere else. 53% of workers say the information they need simply isn't accessible through their AI systems. They're not correcting the model because it's dumb. They're correcting it because it's working blind. And when organizations actually supply that context, botsitting drops about 9% and workers become 31% less likely to ship work they can't stand behind.
That's the finding worth carrying into a meeting. Most firms have spent two years optimizing the prompt. That was never where the gain was. It was in what the model knew before anyone typed anything.
The dividend is real. The question is whether anyone has decided what it's for.
LEADERSHIP LOAD OUT: BRING THIS TO THE MEETING
Your team is probably saving real hours. Ask where those hours went. If nobody can answer, you didn't buy productivity, you bought overhead with better branding.
ALSO ON OUR RADAR
What the market rewards, who's watching, and what's quietly leaving.
Are the Careers of Older Workers Being Cut Short by AI? — Geoffrey Sanzenbacher, Center for Retirement Research at Boston College (Jun 2026) — Workers 55 and older in AI-exposed occupations became 25% more likely to exit employment after ChatGPT's release. The capability walking out the door is the kind that took decades to build. | FREE
Are AI Certifications Worth It? — Revelio Labs (2026) — AI certifications sat at 1–2% of all professional certifications for five years. They're now approaching 30%, a twentyfold jump since ChatGPT. Workers priced this skill long before most employers got around to it. | FREE
More Than Half of Employees Are Using Unapproved AI Tools at Work — KnowBe4 via TechRadar (Jun 2026) — 55% of UK employees are using unapproved AI tools and only 16% think their organization manages AI safely. 27% are sourcing their own on top of what they've been given, which is less a policy failure than a verdict on the sanctioned option. | FREE
PARTING THOUGHT
Every organization running AI right now has a number like Glean's sitting inside it. Hours saved, hours spent babysitting, work shipped that nobody could defend. The only real difference between firms is whether anyone has gone looking for it.
Know someone navigating AI implementation? Forward this to them.
Hit reply and tell me what you think. I read every response.
Published for AI implementation leaders in the US and UK.
Mark Phelps is the founder of Mark Phelps LLC, an AI implementation consulting practice and Fractional VP of Product Design service based in the Boston metro area, serving product, design and technology leaders in the US and UK. He brings more than 20 years of Fortune 100 experience at GE, Fidelity, TIAA, Bank of America and Marsh McLennan, and for the last three years has held fractional VP roles at Rockefeller Capital Management and See All AI. He holds a Stanford AI Certificate earned with Distinction and an MBA in Digital Strategy from Suffolk University (Beta Gamma Sigma). He is a named inventor on a pending design patent for an AI-powered intraoperative 3D surgical navigation UI and a United States Marine Corps veteran.
20 Portsmouth Ave., Suite One #161, Stratham, NH 03885