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 is actually working and what is not. Curated by Mark Phelps and Drew Horton. A Mark Phelps LLC production.
Last issue the labor adjustment hiding in the pay data. This issue is about something hiding in plain sight on every adoption dashboard in the building. Passive AI use produces output that looks identical to active collaboration. The metrics can't tell them apart. The people using the tool can.
THE ONE TO READ
If you only read one AI implementation article before your next leadership meeting, make it this one.
This is what capability erosion looks like when you measure it.
Relying on AI at work reduces self-efficacy, ownership, and meaning while active collaboration mitigates the effects — Elena Hayoung Lee, Yidan Yin, Nan Jia and Cheryl J. Wakslak, Scientific Reports (Mar 15, 2026) | FREE
So much early research on AI at work focuses on adoption rates. This study joins a growing shift toward asking how it actually gets used. The finding: the same tool produces opposite outcomes depending on who does the thinking first. 269 working professionals; analysts, managers and HR people, ran through short writing tasks. Some used no AI, others copied whatever the model handed them and shipped it. A third group drafted first and let AI clean it up.=
The people who used no AI held their sense of ownership and meaning, no surprise there. The ones who drafted first and let AI edit came out the same, no different from using no AI at all. But the people who copied and pasted lost ground on all three metrics and were left disconnected and without confidence in the output. Same tool, opposite outcomes. The critical difference was whether they engaged in the work or outsourced it to the tool.
Here is the part worth taking into the meeting. The study did not stop at how people felt in the moment. Everyone did a second task with no AI at all, and that is where the copy-paste group showed something the others did not. Their confidence stayed low even after the tool was gone, and their sense that they could do the work on their own did not come back. The effect outlasted the task that caused it.
We know that adoption goals tend to reward the fastest method, which is also the one that erodes the skill. (A note to keep in mind; these were ten-minute writing tasks, not careers.) Still, the fix is cheap and it fits on a Post-It: write a bad draft first, then let the AI make it better.
LEADERSHIP LOAD OUT: BRING THIS TO THE MEETING
Pull a week of work from your team and compare the drafts to the final output. If the honest pattern is prompt, copy, ship, then your adoption metric is rising while the ability to catch errors is falling. Your dashboards can’t see this trade and the people moving fastest may be seeing the least. Efficiency and judgment are parting ways.
ALSO ON OUR RADAR
“Comfortable lie,” “comforting belief,” “comforting structure.” Organizations aren’t blind to AI’s risk; they’re choosing comfort and building systems that make the choice feel rational. Aviation made the same choice decades ago: optimize for automation, then discover what had disappeared when it mattered most.
The Comfortable Lie We’re Telling About AI Fluency — Drew Horton, Medium (Apr 30, 2026) | FREE
From our own Drew Horton, and it is the organizational frame under everything else here. Passive AI use can look like productivity right up until you notice it has thinned out the self-efficacy and meaning an organization actually runs on, and the push to maximize adoption selects against the exact people best placed to catch what is going wrong. Read it as the diagnosis the three links below are all symptoms of.
Who Am I Without AI? How To Recognize AI Erosion and Lead Without Losing Yourself — Jeffrey Yip, Center for Creative Leadership (Jul 2, 2026)| FREE
The leadership consequence of everything in the lead. AI expands what a leader can do while quietly eroding judgment, listening and conviction, which are the three things that cannot be delegated to anyone, machine or otherwise. The practical response is unglamorous: protected thinking time, hard boundaries around consequential decisions, and evaluation that rewards judgment instead of output volume.
ICAO Urges Balance Between Automation and Pilot Skills — Mexico Business News (May 20, 2026) | FREE
The precedent, and it is the one worth borrowing. Aviation did not lose manual flying skill in a single decision. It optimized for automation until the missing capability showed up only in abnormal conditions, which is the worst possible moment to find out. ICAO safety officials are now saying out loud that manual competence has to be maintained deliberately, because it does not maintain itself. Substitute your own profession for the cockpit.
AI and the Judgment Trap: How Artificial Intelligence Is Reshaping Human Judgment, Leadership and Decision-Making — Open Magazine (May 25, 2026) | FREE, signup required
The belief underneath all of it: machines take the tasks and humans keep the judgment. This piece argues the line is dissolving, and not because the machines got wiser. It gets built through difficult engagements and slow feedback, and those are precisely what gets compressed away first. Worth the signup, which is why we broke our own free-access rule for it.
PARTING THOUGHT
Every dashboard in the building can see whether people are using AI. Not one of them can see whether the people using it still believe they could do the work without it. The first number goes up on its own, the second only moves if somebody decides it matters.
Know someone navigating AI implementation? Forward this to them.
Hit reply and tell me what you think. We read every response.
And if there is something you think we should cover, send it over. We plan these issues as one connected argument, but the best suggestions come from readers, and we will give you props when yours lands.
Published for AI implementation leaders in the US and UK.
Mark Phelps is a product and design leader with more than 20 years of Fortune 100 experience at GE, Fidelity, TIAA, Bank of America and Marsh McLennan. As the founder of Mark Phelps LLC, he serves as a Fractional VP of Product Design with a focus on AI-driven environments.
For more than three years he has worked as a fractional product design leader, first at Rockefeller Capital Management, where he was engaged to provide critical process improvements, and then on an AI-powered healthcare software product, brought in for urgent user interface deliverables. Both engagements were extended well beyond their original scope. He is a named inventor on a pending design patent for the product’s interface.
He holds a Stanford AI Certificate earned with Distinction and an MBA in Digital Strategy from Suffolk University (Beta Gamma Sigma). He is a United States Marine Corps veteran.
20 Portsmouth Ave., Suite One #161, Stratham, NH 03885