Where the Decision Went

When a piece of work stops or goes to an agent, the judgement inside it moves somewhere, and the people who used to exercise it find out, at best, in a conversation. Different conversations get filed under “change”, by whoever is running the change. Someone losing a craft needs the loss acknowledged. Someone who has lost a decision right needs a straight answer about where the decision went and who now bears the risk. Give the second person the craft conversation and it will sound condescending to someone who wants to know who to call when the model gets it wrong. Give the first the authority conversation and it will miss.

The second conversation is short when the answer exists: who now decides the cases the agent can’t, with what information and authority, and whom to call when it gets one wrong. If nobody can say, no way of running the conversation will help, because the decision has gone nowhere. The craft conversation is the harder one to run well, and skipping any part of it makes it feel like another corporate exercise.

Acknowledging the Loss

The conversation begins with acknowledgement.

“The work you have built your career on is changing. The specific craft of producing this work, the thing you have spent years becoming excellent at, is the part that AI is reshaping. I am not going to pretend it is not.”

This is the easiest step to skip. The instinct is to be positive and frame the transition as a gain, and the intranet article says the same, with more exclamation marks. But the rest of the conversation depends on not doing that first.

When someone afraid of losing their craft hears their manager jump to the positive framing without first saying what is being lost, they hear denial. So they stop listening to the content and start managing the interaction, which in practice means nodding at the right moments until the meeting is over.

Acknowledgment also has to be specific. “Things are changing” is so generic it could open any all-hands presentation ever given. “The research methodology you spent years mastering, the skill that built your reputation in this practice, is the part being compressed” is specific enough to be heard.

Once the loss is acknowledged, turn to what hasn’t changed.

“The years you spent on the research also decided which questions this practice takes on. Who decides that now is what we should settle.”

This works only if it is grounded in the person’s actual expertise. The general claim that human judgment will always be valuable may well be true, but the person read it in the announcement of the tool.

Move then to co-creation. Build the path together instead of prescribing it.

“What would it look like for you to work differently in a way that still feels connected to what you care about? What would you need to keep deciding what the analysis is for, once you no longer build it?”

Craft identity is partly about control: “I produce this, I do it well, the quality is mine.” Imposing a new role adds to the loss. Designing it together gives back some of that control.

Finally, offer concrete support. The conversation ends with specific commitments.

“Here is what I can do: I will redesign the role description, and I will advocate for compensation that reflects what your work is now worth.”

Abstract promises sound like polite dismissal, because plenty of them are. “I do not have a full picture yet, but here are the specific things I know I can deliver” is more credible than a comprehensive plan that sounds too good to be true.

Before the conversation, work out whether the craft itself is being automated, or whether an economic loss is being presented as an identity concern, perhaps because identity is cheaper to discuss than pay. The approach differs, and getting it wrong is how well-meant conversations go badly. Afterwards, ask yourself whether you acknowledged the loss or softened it. Did the person speak more than you did? If they left with more say over what happens next than they came in with, the conversation worked.

Then let the person say what they are losing, without rushing to fix it.

Nor can the pace be forced from outside. AI capability changes in months, and people change how they see their own work over years, which is awkward for anyone with a quarterly roadmap. Better conditions help, while enthusiasm and deadlines don’t. A deadline gets you compliance, and compliance without commitment quietly routes the hard cases around the tool.

Corporate cheerfulness derails it: “Isn’t this exciting? Think of the possibilities,” said to someone who has just described a loss. They learn that their experience isn’t welcome in its actual form, stop sharing it, and start performing adaptation. The adoption numbers improve, which pleases whoever reports them upwards, and the resistance hardens where nobody is measuring it.

Premature reassurance does the same. At the first sign of distress the manager moves to “your job is safe, your value is clear.” The person may already know the job is safe. That wasn’t the concern, and now they also know the manager didn’t notice.

So does skipping straight to training. A training course is the wrong answer to an identity problem, although it is the easiest answer to budget for and report as done. A person can learn every new skill and still not make the transition, because the harder change is in how they see their own work.

Each mismatched intervention makes the original problem worse. The person concludes, not unreasonably, that the organisation can’t see what is happening to them, and withdraws further. The conversation itself is diagnostic. Someone worried about money steers it towards pay and security. When identity is at stake, it shows in how closely they listen, or how carefully they hide that they’re listening.

Performed Conversations

Of course, once a conversation has been written down in steps, it becomes a template, and a template can be used sincerely or as theatre.

Under pressure to “run the conversation” by a completion date, a manager can borrow the language of acknowledgement, reframe and co-creation after the decision about the person has been made elsewhere, in a meeting the employee didn’t attend. The employee gets every element of a well-crafted conversation, then a restructuring in which the role they were told is “not going away, just shifting” is gone.

So employees have reason to distrust the beautifully worded transition conversation, because the sincere form and the weaponised form use identical vocabulary. Argyris had a name for the gap: espoused theory, what people say they’re doing, against theory-in-use, what their behaviour shows they’re doing, and his finding was that managers almost never see the difference in themselves.¹ So a performed conversation can be performed in good faith. The difference is outside the words, in whether the person running the conversation has authority over outcomes the employee cares about, or whether the conversation is a coat of paint on a decision made elsewhere.

A few patterns separate sincere from performed. A sincere conversation commits to specifics that constrain the manager’s future behaviour: redesign the role description this quarter, or advocate this compensation change to the next band. The manager books a follow-up, with the next specific step on it, and then keeps the appointment. A performed conversation produces aspirational language that binds the manager to nothing, phrases like “we will explore” and “the organisation is committed to supporting,” and the follow-up is an email of thanks, then silence until the next scripted touchpoint.

Once employees have seen the pattern, they read every future conversation through it. The sincere version gets a manager heard. The performed version, which a manager can be cast in without choosing it, costs them credibility for a conversation they didn’t write, while the people who did write it aren’t in the room.

For the team leader, the conversation is something to facilitate for others. A single session with a tick next to it achieves little, because people move over several conversations. The manager’s particular difficulty is that the conversation means admitting uncertainty.

The same conversation applies to the leader’s own role, and is better run inward before running it on anyone else. The craft an executive built their authority on is changing too. If what makes it hard is not knowing which decisions will still be yours, you’re owed the same straight answer as anyone else who has lost one, by whoever is running the change above you.

Checklist

Which conversation is this: a craft that is dissolving, a decision right that moved, or economic displacement dressed as identity threat?
Can you name the specific loss, in this person’s terms?
Is the reframe grounded in this person’s expertise, or is it the general claim about higher-value work?
For each decision that moved, can you say who now makes it and whom to call when the agent gets one wrong?
Do you have authority over the outcomes? If the decision has been made elsewhere, don’t hold the conversation yet.
Will you commit to specifics that constrain what you do later, with dates?
Afterwards: did you acknowledge the loss or soften it, did the reframe feel genuine or performed, and did the person speak more than you did?
Have you planned for more than one conversation?

Notes

Argyris and Schön 1974.↩