Skip to content
Maroš Jančo
← All writing

Are you AI-transitioning your company? Awareness & availability ≠ adoption!

Across your teams, can you actually say how regularly people use AI tools, on what work, what the leverage is, how those tools spread from one person to the next, and whether anyone has had a real chance to explore and show what changed? If those questions have no answers, "we're adopting AI" is a hope, not a status. Adoption has to be structural — not "we bought the licences."

Why it stalls

Teams are measured on what gets done — the milestone, the ticket — and less on how. Plenty of delivery methodologies exist, but which of them account for the leverage and use of AI tools? AI tooling usually stays optional — and optional against a deadline means later. It's an exploration-versus-exploitation problem, which is policy-based, not motivational: exploiting the known workflow ships predictably this sprint; exploring a coding agent or an automation has an uncertain payoff and a certain cost, paid from the same hours you owe the roadmap. Make exploration optional and you have decided against it.

This is not the team being incurious. You may have enthusiastic AI early adopters in your company who also showcase the impact of AI tooling on their work, but does this scale sufficiently to the whole team and across departments (marketing, finance, customer support, engineering)?

What each side has to do

Leaders — change the structure, then act on what it returns:

  • Change the reward system. Reviews, OKRs and sprint credit measure the what. Until they also count the how — a demonstrated, reshaped workflow — exploring AI stays the irrational choice for the person doing the work. Reward leveraged impact, shown, as visibly as a shipped ticket.
  • Make exploration a standing cadence, not a one-off. A hack-day bolted onto a roadmap that still punishes every non-roadmap hour is theatre — run it once for show and you teach the opposite of what you intended. Install a recurring explore → try → share what works → scale loop as a first-class, company-wide initiative built into the planning cadence itself, not an exception to it. Each cycle is protected (not the first thing cut when the sprint runs hot), expected (a demonstrated outcome owed back), and owned by the person; the "share what works" step is what feeds the next cycle's scaling.
  • Adopt the wins visibly — feed them into how the company plans and measures — and run it across every role, not just engineering. The functions with the least existing automation often have the most slack to recover; "show how your role changes" is what makes this the company's default rather than one team's.

The people doing the work — make the window real:

  • Demonstrate it: a concrete before/after on your own work, shown rather than described — not a tour of tools.
  • Share honestly: what worked, what didn't, where the tooling should trend next. The honest "this didn't work" is as valuable as the wins — it stops everyone else re-paying the same cost. The share-back is not the wrap-up; it is the point.

When the boost is assumed but not measured

There is a failure that looks like the opposite of non-adoption and is just as costly. Once AI is in the building, plans quietly start assuming the boost — scopes grow, timelines compress, more is committed because "with AI this is faster." But adoption is uneven and no one is measuring it: a few people have genuinely changed how they work, most have not, and you cannot yet predict who or by how much. Planning capacity against an average uplift that does not exist for most of the team does not produce more output. It produces more unfinished projects, spiked cross-project dependencies, and the context-switching tax of too many things in flight — measurably worse than before the tools arrived.

So do not price the AI gain into commitments until you can observe it. Treat adoptability as something to measure per person and per role — who has actually changed their workflow, on what, with what effect — and let demonstrated uplift raise expectations, never the assumption that the tools imply it.

What the data says

You do not have to take this as one consultant's impression; the large workplace surveys converge on it:

  • About 71% of organizations regularly use generative AI in at least one function, but only ~21% have fundamentally redesigned any workflow around it, and only ~5% see material bottom-line impact. (McKinsey, State of AI, 2025)
  • 78% of AI-using employees bring their own tools to work, while 60% of leaders worry their company lacks a plan and vision to put AI to use — the two-sided gap, measured. (Microsoft–LinkedIn Work Trend Index, 2024)
  • Leader weekly AI use jumped from 37% to 82% between 2023 and 2025 — people are no longer failing to try the tools. Set against the figures above (used widely, workflows rarely redesigned), that is the argument in one line: trying is now common, changing how the work is done is not. (Wharton, 2025; redesign figures: McKinsey, above)

Building the AI systems is one half of the work; making an organisation actually adopt them is the other — and it is work I take on: guiding the AI transition, reconstructing reward systems and ways of working, and making sure teams use the tools the right way (explore → try → share → scale). If that is where you are, get in touch.

Sources