Most of the leaders Dr. Zdenka Cumano and I work with at AI Leader Edge are not short on AI pilots. They are short on adoption. A small team gets excited. It runs a pilot. The pilot works. Everyone applauds. Then nothing happens. Six months later there are five more pilots and the same nothing.
I have seen this pattern before, and I paid for it twice. It is the chasm, and it lives inside the company.
I got the curve wrong twice.
In 1999, at Internet Venture Works, we built the dot-com versions of businesses owned by private equity firms. Those firms were visionaries. But the people who had to use those businesses were ordinary shoppers, and most of them were not ready to buy a heavy bag of dog food from a website. We had enthusiastic customers and were still stuck before the chasm. The right side of the curve did not arrive in time.
An AI pilot repeats that mistake inside a company. The executive who sponsors it is the visionary. The operating team that has to live with it is the pragmatist. Applause from the sponsor is not adoption by the team.
Lycos Clubs taught me the other half. We launched before Yahoo. We had the audience. We built it for power users, and it did not get much traction. Yahoo Clubs came six months later, did less and took off. Geoffrey Moore had a name for what Yahoo shipped: the whole product. Most pilots are Lycos Clubs. They impress the people who asked for them and confuse everyone else.
A pilot proves the technology works for the volunteers. It proves nothing to the pragmatists.
The core barrier to scaling is not infrastructure, regulation, or talent. It is learning.
The whole curve, under one roof.
Every organization contains the whole adoption curve. The innovators are already using AI at home and sneaking it into their work. The early adopters volunteer for the pilot. The large pragmatic middle runs the business. They adopt a new tool only when it is proven, complete and already working for people like them.
MIT NANDA’s 2025 report found that only 5 percent of custom enterprise AI tools reach production. Roy Amara explains part of why. We tend to overestimate the effect of a technology in the short run and underestimate it in the long run. Leaders expect this quarter’s pilot to transform the company. They underinvest in the slow work: redesigning roles, training people and building the data and rules the systems will need.
Win one beachhead.
Moore’s prescription works inside a company the same way it works in a market. Pick one beachhead. One workflow, one team, one painful problem. Deliver the whole product there, not just the tool. Running a unit of 350 people at Lycos taught me that what is not written down does not scale. The whole product is mostly the part a pilot never writes down.
- The data
- What the system needs, in the shape it needs.
- The training
- So the team works without calling the pilot team.
- The rules
- What the system may decide, and what it may not.
- The measurement
- Proof, in the team’s own numbers.
- The owner
- A person who owns the result.
Win that beachhead completely. Then move to the team next door, the one that shares the most with the first. That is your bowling alley. When a leader asks me why their AI program is stuck, I ask three questions. Where is each team on the curve? What is your beachhead, and have you delivered the whole product there? Are you planning in straight lines for a technology that compounds?
Find your internal beachhead.
- Draw your curveDraw the adoption curve for your organization. Place each team on it.
- Circle the beachheadCircle the team with the most painful problem and the most willingness to change.
- Write the whole productWrite down everything that team would need to succeed without calling you. That list is your whole product.
- Name the next pinName the team that shares the most with your beachhead. Go there next.
Quick answers.
Why do AI pilots fail to scale?
Rob Frasca argues that every organization contains the whole adoption curve. A pilot proves the technology works for volunteers, the early adopters, but proves nothing to the pragmatic operating teams. He learned the pattern at Internet Venture Works, where visionary buyers could not carry pragmatic users across the chasm. Crossing it takes one beachhead team and a whole product: data, training, decision rules, measurement and an owner.
What is an internal beachhead for AI adoption?
In Rob Frasca’s framework, it is one workflow, one team and one painful problem where AI is delivered completely, before moving to the team next door that shares the most with the first.
Sources.
- Rob Frasca, The Creator’s Journey (forthcoming), Chapter XIII, Becoming Super.
- Aditya Challapally, Chris Pease, Ramesh Raskar and Pradyumna Chari, The GenAI Divide: State of AI in Business 2025 (MIT NANDA, July 2025). Copy of the report as hosted by Webber Wentzel (secondary host). Source
- Geoffrey A. Moore, Crossing the Chasm (HarperBusiness, 1991).