The Adoption Gap is a newsletter for leaders who want AI tools that work, published periodically. Each issue cuts to the research, data and trends that matter, not the hype, not the doom, just what closes the gap between AI deployed and AI adopted. Curated by Mark Phelps and Drew Horton. A Mark Phelps production.
Last issue I laid out the five axes. This week I want to take the idea that sits behind the whole newsletter: AI tools that work are designed before they’re deployed.
This issue: ADOPT · Design
THE ONE TO READ
If you only read one AI article before your next leadership meeting, make it this one.
Deployed means the tool is installed. Adopted means someone opens it on a Tuesday without being asked. I’d argue most of the gap between the two is decided before launch.
The GenAI Divide: State of AI in Business 2025 by Aditya Challapally, Chris Pease, Ramesh Raskar and Pradyumna Chari, MIT NANDA (Jul 2025) | FREE (PDF)
That’s the position I’d like to focus on. Whether an AI initiative or tool gets adopted by its intended users mostly comes down to choices made before it ships: which workflow it changes, who owns the result, what the people doing the work are asked to do and what they’re asked to trust. Product, IT or a vendor can install the tool. Someone has to make those four choices on purpose.
The noisiest number behind this is the NANDA report’s claim that 95 percent of organizations are getting zero return from their generative AI initiatives. Treat it with a grain of salt. It’s a preliminary report built on a small sample (interviews with 52 organizations, surveys of 153 senior leaders and a review of more than 300 public initiatives), and the authors themselves call the figures directional, with success judged by what users and executives reported. I’d read it as a signal, not an exact measurement.
What holds up is the explanation. It says the divide doesn’t seem to be about model quality or regulation, and points to the approach instead. It names organizational design as the dominant barrier and ties success to clear ownership and tools that fit the workflow they land in. Which confirms the thesis behind this newsletter: adoption is a design problem, and the design work happens before the rollout, not after it.
In addition, WalkMe, which sells digital adoption software, surveyed 3,750 executives and employees this year. Fifty-four percent of workers said they’d bypassed AI tools and done the task by hand at least once in the past 30 days, and only 9 percent trust AI tools with complex, business-critical decisions. That compares with 61 percent of executives.
Both of these studies rest on what people told researchers, and neither can show cause. For instance, a tool that gets bypassed may have been badly designed, or its users may have been poorly briefed, or it may have been the wrong tool for the wrong workflow. NANDA’s authors make the same point about their own build-versus-buy numbers.
Still, I’m placing my bet on design.
LEADERSHIP LOAD OUT: BRING THIS TO THE MEETING
Take one AI initiative that’s about to ship, or one that already has. Can you say in a sentence each which workflow it changes, who owns the result, what the people doing the work are being asked to do and what they’re being asked to trust? Any answer that takes longer than a sentence is design work that’s still open.
ALSO ON OUR RADAR
Forum AI CEO Says Most Companies Start With the Wrong AI Strategy. Her Fix Begins With 1 Problem by Chloe Aiello, Inc. (Oct 19, 2026) | FREE
At the Inc. 5000 Conference, Campbell Brown of Forum AI said she keeps seeing companies choose the plan, the model and the vendor before they know their most important use case. Her advice: “Stop and figure out one use case.” That’s the Alignment question, asked before the vendor call. Forum AI sells AI evaluation.
New Study Shows C-Suite Leaders Highly Confident in AI ROI Even as 58% Claim There’s No Clear Ownership of AI and 75% Lack AI Governance by Larridin (Feb 3, 2026) | FREE
In a January 2026 survey of 365 senior leaders at companies with 1,000 or more employees, about 58 percent named unclear or fragmented ownership as the main barrier to measuring how their AI tools are performing. Ninety-two percent of the C-suite respondents still said they were fully confident in the impact of their AI tools. Confidence is high. Clarity about who owns the result is not. Larridin sells AI measurement software.
KPMG 2026 CEO Outlook by KPMG UK (Oct 15, 2026) | FREE
Among 150 UK CEOs in KPMG’s 2026 global survey of 1,400, 85 percent said they’re redesigning roles, career paths and team structures for human and AI collaboration, up from 71 percent in 2025. That’s stated intent, not results, so I’d treat it as a direction of travel. KPMG sells AI services.
PARTING THOUGHT
Deployment is a date on the calendar. Adoption is a habit, and habits don’t form by accident or by deadline. Someone designs the conditions, or an expensive tool just might sit there.
Know someone rolling out AI tools at their company? Forward this to them.
Hit reply and tell me what you think. We read every response.
And if there's 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'll give you props when yours lands.
Published for leaders putting AI tools to work in the US, Canada and UK.
Mark Phelps
A Partner for Product & AI Integration Design
Mark Phelps has more than 20 years leading product design and process improvement at GE, Fidelity, Bank of America, TIAA and Marsh McLennan. He works as a fractional product design and AI integration partner, most recently at Rockefeller Capital Management and on a regulated AI-powered healthcare software product. Stanford AI Certificate, UI/UX Design for AI Products, earned with distinction. MBA with Beta Gamma Sigma. Named inventor on a pending design patent for an AI-powered medical device interface. Marine Corps veteran.