The Unspoken AI Rule: Managers Expect You to Use AI. Nobody Said So.

A Rule That Exists Only in Managers' Heads

Most workplace rules are written down somewhere. This one is not.

The Checkr numbers describe a particular kind of rule: an unspoken performance requirement. More than half of managers believe using AI is now part of doing the job well. Fewer than a third of employees believe the same, and over a third cannot tell either way.

That is not a technology gap. It is an expectations gap, and expectations gaps are among the most expensive things in management because they are invisible until someone is rated, passed over or quietly disengages.

Think about what the gap does to each side:

• The employee who holds back looks unambitious, though they may be following the data-safety instincts you want them to have.
• The employee who experiments freely may look like a star, though they may be pasting confidential material into tools nobody approved. We covered that risk in our post on shadow AI.
• The manager believes they have communicated a standard. They have not. They have only formed one.

The Other Half of the Gap: When Managers Hand Over the Thinking

The Omni Calculator findings point at a second, quieter problem. It is not only employees who are unclear. Employees are also watching how their managers use AI, and they are drawing conclusions.

Half say their boss has used AI to answer a question or make a decision that should have been theirs. Two-thirds say AI-generated messages feel less personal. Four in ten say a manager who relies on AI seems less competent.

Read that last number carefully. The same tool that managers see as a sign of competence in their team can read as a sign of incompetence when the manager uses it for feedback, appraisals or tough conversations. A message that is polished but empty is what we called workslop when it travels sideways between peers. When it travels downwards from a boss, it costs something more valuable: credibility.

This matters most in the moments that define a manager. Performance feedback, a difficult conversation, a decision about who gets a stretch role. People can tell when no human thought was involved. The skill of giving and receiving feedback was never about wording. It was about showing the person that you saw them.

Why This Hits Indian Teams Hard

Be careful about borrowing US numbers wholesale, but three features of many Indian workplaces make an unspoken rule more risky, not less.

1. Hierarchy discourages asking. In many teams, saying "I am not sure what is expected of me" feels like admitting weakness. So the uncertainty stays private while the guessing becomes public.
2. Spans of control are widening. As layers shrink, each manager has less time per person, which is exactly when unwritten rules replace conversations. See our post on the great flattening.
3. Client and data sensitivity is high. A large share of work in services, BFSI and IT touches client data. In that context, the safe default for a cautious employee is to do nothing, which looks like low adoption but is in fact low clarity.

Five Things to Say Out Loud This Quarter

You do not need a policy rewrite to close the gap. You need to say the quiet part aloud.

1. State the expectation in one sentence. For example: "For these three kinds of task, I expect you to try AI first. For these two, do not use it." Specific beats aspirational.
2. Name who owns it. One in three employees in the Checkr data did not know who is responsible for pushing AI. Put a name and a channel on it, so questions have somewhere to go.
3. Separate "allowed" from "expected". Many people hear "you may use AI" as "you must" or the reverse. Say which one you mean, and say it separately for each task type.
4. Model it, and show your checking. Managers who use AI openly, and who show how they verified the output, give their team a template. Managers who use it secretly give their team a rumour.
5. Keep the human moments human. Decide as a leadership group that appraisals, feedback and sensitive messages are written by a person, with AI at most as a sounding board.

How Bodhih Helps Teams Close the Expectation Gap

Clear expectations only work if people have the skills to meet them and the confidence to ask.

• AI Literacy Training gives every employee the same practical baseline: what the tools do well, where they fail, what data stays out and how to check results. That turns "I am not sure if I am allowed" into "I know how to do this safely".
• Leading Human + AI Teams Training is built for managers like Rohit. It covers setting expectations for AI use, assigning work between people and agents, defining review standards and keeping feedback human.
• AI for Leaders helps senior teams agree who owns AI adoption, what the principles are and how to talk about them before the rumours start.
• Coaching for Managers builds the conversation skills to ask "what are you unsure about?" and actually hear the answer.
• Critical Thinking Training strengthens the habit of questioning a fluent answer, for managers and employees alike.
• Change Management Training helps leaders carry a team through an expectation shift deliberately, instead of letting it arrive through hints.

To see where your own teams stand before investing, structured assessments can show real skill levels rather than self-ratings, and a short pre and post measure makes the impact of any training visible.

The Bigger Idea

We tend to treat AI adoption as a rollout: tools, licences, training sessions. The surveys suggest the harder part is social. Every team has a set of unwritten rules about what good work looks like, and AI has just added a new, undeclared line to that list.

Rohit is not a bad manager. He is a normal one, running an expectation he never stated. Divya is not resistant. She is careful, and careful people need the rule written down. The gap between them closes with one honest conversation, which costs nothing and can be held on Monday.

The question for this week is simple: if I asked each person on my team what I expect from them on AI, would I get the same answer from all of them, and would it be the answer I would give?

Frequently asked questions