of enterprise AI pilots fail to reach sustained production
MIT-backed analysis · Fast Co · Apr 2026
of executives who laid off staff for AI now report regret about how they planned the implementation
Forrester 2026 Future of Work
Put the two numbers above together and you get a story common enough now to have a name: AI Regret.
A hastily implemented rollout stalls, the people cut to fund it are already gone and the promised savings rarely show up. Eighteen months ago that reversal took the better part of a year to surface, lately it runs closer to six months. The more useful statistic is what the same research reveals about the companies it doesn't happen to.
Gartner studied 350 executives and found no meaningful correlation between how much workforce a company cut and how much ROI the newly implemented AI delivered. Only 21% report significant returns, and the ones who do aren't running better tools, they're running better implementations (DataCamp, 2026). The companies that cut hardest before redesigning the work are the ones now rehiring, two-thirds of them already in the process, with 31% spending more on the rehires than they saved on the original cuts (Careerminds, Feb 2026).
“Businesses cannot unlock AI-driven growth by buying technology alone. The real bottleneck is the severe shortage of professionals capable of integrating, governing, and scaling these systems.”
— Sander van't Noordende, CEO Randstad · May 19, 2026 · analysis of 50 million job postings
The implementations that avoid AI Regret share three things:
Workflow redesign comes before technology deployment. Tools dropped on top of an unchanged process and an unprepared team produce the 95% failure rate. Workflow first, technology second, every time.
AI should make your team better, not smaller. The 21% who see significant returns use it to free their people for higher-value work. The 55% who regret their approach replaced their people instead, and most are rehiring at higher cost than the layoffs saved.
The team is part of the redesign, not the audience for the rollout. The people who do the work help design how the work is done, and humans stay in the loop at the failure points where verification matters most.
See how this works in practice → Services
Mark Phelps LLC was founded to help the executives and product leaders who own these decisions implement AI without the stalled rollout or the regret that follows.
Twenty years of Fortune 100 transformation and product design leadership at GE, Fidelity, TIAA, Bank of America and Marsh McLennan. The last three years in fractional VP roles at Rockefeller Capital Management and See All AI. Same playbook each time: workflow first, cycles short, teams in the loop. The technology stops being the hard part, and the regret doesn't follow.
What if AI made your team better, not smaller?
ROI without Redundancy.
Here because you're wondering how AI affects your career, not just your company's bottom line? That's a different conversation, one worth having. → notobsolete.co