AI Transformation at the Top.
When an AI initiative stalls at partner level, the cause is rarely the technology. Partners who hesitate are following a logic of their own: their economics rest on scarce billable hours, their authority on decades of judgment. A firm-wide push that ignores this produces polite compliance and no change in behavior.
Partners come to AI through a specific moment. The moment is usually private and concrete: an output on a real engagement that shows something the partner had not imagined before. Two hours of research on the competitive landscape in four minutes. A position on a regulatory question they would not have produced this quickly at that depth on their own. A slide structure that captures their logic more precisely than the first draft they had in mind. What works is the feeling:
“With this tool I can do things I could not do before. My capability is getting bigger.”
Weeks lead up to this moment. They are about shaping the conditions so that the moment becomes possible. That is four things:
Three risks block these four conditions before they take hold.
Three steps dissolve these risks:
Going up is work, going down is easy.
A partner’s stance shifts over time, and which step works depends on where they stand right now: the private demo that moves a skeptic up a stage shows a champion, above all, that you don’t know where they stand.
Going down is the invisible part. A champion who loses their stage doesn’t become an opponent; they go quiet, and soon use AI only for themselves.
The right tools.
Which one applies depends on the moment.
Extending implies: what exists is sound, something new is added. Introducing implies: something was missing until now.
The partner associates productivity with the junior conversation and hourly rates - an entirely different conversation.
Historical parallels position the partner as a competent navigator of past transitions.
External observations activate the strategic reflex. Internal comparisons trigger defensiveness.
“What would be the most time-consuming part of your engagements that you would most like to get rid of? That is where we start.”
A concrete anecdote bypasses the abstract resistance.
“True. That is why we start with the steps that work without client data: competitor research, structural drafts, preparation. The sensitive part comes later, once the infrastructure for it is cleanly in place.”
Separates the use cases by data sensitivity before the conversation gets stuck on a single use case.
“That is exactly why. 20 minutes, a concrete engagement of yours, I show you the output. You only watch. If it does nothing for you, we forget it.”
Reduces the cognitive load of getting started to the minimum. Leaves a way out open.
“They can. Only then you become the bottleneck. If the team gets faster and you do not, it waits on you more often, not less - and the pressure concentrates at the top. Most partners only notice this once it is already tangible.”
Shifts from the identity topic to the systems question. Identity-sparing and mechanically clear.
“Let us try it. 20 minutes on one of your engagements. If it is hype, you saw that for yourself. If not, you have the next step.”
Trades the burden of proof for an experiment. The partner becomes the judge.
“That is exactly what you are there for. What gets shorter is the distance between the brief and the first point at which something can be judged. Judgment stays with you.”
Reinforces the role of the partner.
“Let us clarify that in parallel. In practice, the main use cases do not touch the protected data classes, and for the ones that do, there are established paths. We map data classes against tools, then we know where friction sits and where it does not.”
Confirms the concern. Turns it into a work package, not a showstopper.
A training session with mixed seniority treats senior partners like junior associates. They leave the room with the impression that AI is something for juniors.
A demonstration in front of several partners, none of whom has tried it privately. Everyone remembers what goes wrong. What works demands a public reaction.
Going straight from one champion to "all partners next week". Diffusion takes time. Better to win two champions one after the other.
In a partnership, authority is not distributed hierarchically. The obligation replaces the peer signal that actually carries the process.
External speaker, vendor pitch, conference vibes. Partners have spent years learning to filter out such pitches. The substance does not reach them through that form.
Triggers the reflex "I will not be lectured by my juniors". Often factually correct, yet backward in its effect.
Triggers the junior-and-hourly-rate conversation. Efficiency is a good outcome, a poor argument at the start.
Reaches the partner without an anchor. What does reach them: the point in the next engagement where they normally lose two hours.
From a distance they look like a skeptic - don't talk about it, don't praise it, don't speak up. Privately, intensive use of AI, own workflows. Recognizable only through the right question: "Has anything surprised you AI-wise in the last two weeks?"
Don't treat them like a skeptic - that would be an insult they won't say out loud. Make them visible directly: the targeted request to tell a concrete story. Visibility is the gift.
A clear refuser who flips within days after a specific moment. Usually a concrete peer moment: an output seen over a colleague's shoulder, a chance anecdote.
React quickly, before the momentum fades. 48 hours after the moment, a concrete use case - no training, no long preamble. A refuser who has just flipped becomes an active user faster than from any other entry point.
Enthusiastic six weeks ago, a driver of use cases. Now: fewer mentions, briefer responses, no interest in new tools.
A direct question: "Where are you tired right now?" A common answer: "I am carrying this alone." What helps: a real peer community at the same level, and recognition in a credible form - a speaker slot, a role in the program.
Other partners roll their eyes when the ally starts in on AI. The peer card that carries the whole program is worn out.
An immediate pause. Two to four weeks of silence from the ally on the topic. Instead, make another, less obvious partner the visible voice. Change the sender.
"Yes, I use that too." When pressed: no concrete examples, vague phrasing. Performance without substance - often where visible skepticism has become socially unattractive.
Do not expose them. Instead, offer a protected use case: an hour on the next engagement, together. The performance gets a chance to become real experience, without anyone losing face.
Partners come to AI through a chain of moments in which their capability visibly gets bigger - and through an environment that makes those moments likely and prevents the relapse. Whoever knows the staircase, has the language, and avoids the antipatterns recognizes the next right step for the next partner.
- Everett Rogers — Diffusion of Innovations - the adoption stages and the role of opinion leaders
- Deci & Ryan — Self-Determination Theory and Work Organizations - autonomy, competence, relatedness
- Electronic Markets (Springer) — The Rise of AI - Understanding the AI Identity Threat at the Workplace
- Daphna Oyserman — Identity-Based Motivation - when an action feels identity-congruent
- BCG — AI at Work 2025 - Momentum Builds, But Gaps Remain
- McKinsey — Superagency in the Workplace 2025
- Harvard Business Review — Where Senior Leaders Are Struggling with AI Adoption
- Thomson Reuters — The 2,000-Hour Problem - When AI Efficiency Collides with Billable Time