Skip to content

Knowledge base

Adoption gaps and resistance

AI pilots stall when the organization does not change with the tool. Usage spreads unevenly and shallowly, nobody owns the risk of AI-generated output, and engineers fear for their expertise more than their jobs, so adoption needs structural and cultural work alongside the rollout.

Written by

Updated:

Getting Started 2

Our position. We read a stalled adoption as a leadership failure, not a technology failure. Leadership has three jobs: unlock usage, make adoption visible and reduce fear (opens in a new tab). Put the worries on the agenda before the tools, and have someone from your own organization answer them.

The gap opens between individual speed and shared capability. The tools make each engineer faster and make collaboration harder.

How do you close the adoption gap?

  • Ask what changed because you spent the tokens, not whether people use AI.
  • Name the worry about positioning to the person, before it hardens into objections about tooling or quality.
  • Win the doubtful optimists first, since they are the ones who move the skeptics.

Why do technically successful AI pilots stall?

Individual proficiency does not automatically become team capability. Rotational Labs observed that the tools accelerate individuals while making collaboration harder, forcing the team to add deliberate synchronization: more co-working sessions, shared experiment artifacts, and purpose-built internal tooling (Rotational Labs - What We’ve Learned From 3 Months Living with AI Coding Tools (opens in a new tab)). One practitioner essay describes the organizational version of the same gap as the “messy middle” of adoption: usage is everywhere, uneven, and partially hidden, and the useful question shifts from “are people using AI?” to “what changed because we spent those tokens?”, from token-to-output to token-to-learning (When everyone has AI and the company still learns nothing (opens in a new tab)); the corollary is to separate deep, powerful use from shallow theatre, since any raw metric of “AI usage” invites gaming (video essay on the AI-native talent market (opens in a new tab), 13 January 2026).

Field data shows how easily that step is missed: Accenture reports that 88% of AI pilots never scale to production, 62% of companies say they lack the skills, 70% lack the change-management capability to make a deployed tool actually get used, and 89% see no clear return yet (WeAreDevelopers – The Technology Revolution: Mastering the Challenges of Radical Change (opens in a new tab), 3 June 2025); German field data agrees from the other side, with only about 27% of companies yet harvesting real benefits from their AI investment (FAZ Pro – Wie Unternehmen mit GenAI ihre IT-Organisation neu erfinden (opens in a new tab), November 2025). A large comment analysis adds a pointed lesson for anyone rolling AI out to engineers: their loudest question has moved from “does it work?” to “can we even tell if it works?”, so a pitch built on “this makes you faster” answers a question the audience has stopped asking, and enablement lands better when it speaks to trust and discipline than to raw speed (What 30,000 YouTube Comments Tell Us About AI and Software Engineering (opens in a new tab), 1 May 2026).

A founder whose company measures AI systems for a living puts the general form of the gap sharply: the bottleneck is almost never model capability but institutional readiness, and model capability is compounding faster than organizational adaptation, so a pilot that succeeds technically in a week can still stall for months over who owns risk acceptance for AI-generated output (A Country Full of Geniuses (opens in a new tab), 13 February 2026). The raw numbers of who is even using the tools are humbler than the discourse suggests: only about one employee in five has any employer-provided chatbot at all, two-thirds of people who start using AI drop it within a month, and most users have never once watched a model think for longer than a minute, so the visible adoption curve is mostly shallow experimentation, not the deep use the gains require (Beyond the AI Hype (OMR) (opens in a new tab), 7 May 2026).

One LinkedIn essay reframes why the deep version stalls: most AI initiatives are run as technology deployments and fail for reasons that have nothing to do with technology, and the engine of failure is organizational drag, the structural resistance already present before the first move, so the leader’s job is to reduce drag before adding force (The Drag Problem in AI Adoption (opens in a new tab), 2 June 2026). Serious adoption hits four dimensions at once, technological, structural, cultural, and strategic, and companies that staff for the first and ignore the other three misclassify a transformation as a tool rollout; the underdiagnosed resistance is not fear of losing the job but of positioning, the quieter question of whether one’s expertise will still matter and in what shape, which surfaces as vague objections about tooling or quality and has to be named directly, to the person, before pressure converts it into active resistance (same source).

A survey puts a number on the drag. Roland Berger’s 2022 survey of 50 board-level executives found 69% saying their employees could not collaborate with other departments, especially when it came to IT, and some of them attributed this to the IT department being seen as just a service provider rather than a partner for change; it predates generative AI, but it shows the drag as a matter of how departments see each other, not only of skills (Roland Berger - The digital dilemma (opens in a new tab), 14 July 2022). A German magazine review of the resistance literature adds what to do with the people: the group that sets an organization’s direction is the doubtful optimists, critical thinkers who are nonetheless positive about the technology, and they are the ones who move the skeptics; perceived control matters for wellbeing even where real control is absent, and an organization raises it by making what is coming predictable and explainable, which is the mechanism behind transparent communication. The studies it cites also found that higher awareness of automation technologies correlates with lower commitment and career satisfaction and higher turnover intention, and that employees felt informed about AI but not by their employer, so silence from the top gets filled from elsewhere. The narrative is part of it: hybrid intelligence that augments the workforce, as against an artificial intelligence that outranks it (codecentric - Der Softwerker Vol. 24: Who’s afraid of AI? (opens in a new tab), German).

Resources

Articles

Papers and reports

  • The digital dilemma (opens in a new tab) - Roland Berger survey of board-level executives on business-IT misalignment, separating missing collaboration capacity from contradicting incentives; predates generative AI.

Posts

Talks and videos