The sceptic conversation

Every organisation has them: the experienced analyst who says AI output is rubbish, the team lead who's "not convinced it's worth the time", the senior partner who quietly tells their team not to bother. It's tempting to write them off as resistant to change and focus on the enthusiasts. That's a mistake, for two reasons.

First, sceptics are often influential. They tend to be experienced, and other people watch what they do. Second, they're often right about something. The objection might be overstated, but there's usually a real problem underneath: an output that embarrassed them, a worry about their role, a quality bar the tool doesn't meet for their work. If you find it, you either fix something or learn where not to push. Both are worth having.

This playbook is a thirty-minute one-to-one conversation, and what to do afterwards.

Before the conversation

Choose who to talk to

Pick sceptics with influence: people whose opinion others follow, or managers whose team's adoption depends on them.

Ask, don't summon

How you ask matters. "I'd like to discuss your concerns about AI" sounds like a disciplinary. Try:

I'm trying to understand where AI is genuinely useful here and where it isn't. You've got a lot of experience with [their work] and I suspect you've got a clearer view than most. Could I take thirty minutes to hear it?

That's true, it's respectful, and it signals you want their judgement rather than their compliance.

Do your homework

Find out what they do and whether they've used the tools. If someone tells you "they tried it last year and hated it", that's useful context. Don't bring it up as an accusation.

The conversation: thirty minutes

0–5: set it up

Say what you're doing and what you're not:

I'm not here to convince you. I want to understand what you've seen and what worries you, because that's how I'll avoid pushing AI where it doesn't belong.

5–20: find the real objection

Start broad and narrow down. Useful questions:

1. What's your honest view of these tools for the kind of work you do?
2. Have you tried it on something real? What happened?
3. If a colleague told you it had saved them an hour on [their task], what would you worry about?
4. What would it have to do for you to think it was worth using?
5. Is there anything about how it's being rolled out that bothers you, separate from the tool itself?

Listen for the specific thing. The first answer is often general ("it's not reliable"). The second or third is usually specific ("it gave me a citation that didn't exist and I nearly sent it to a client"). The specific thing is what you're looking for.

20–27: respond to that objection, not the general case

Don't deliver a general pitch. Respond to the specific objection they gave you, using the guide below. Sometimes the right response is "you're right, and I'm not going to recommend it for that".

27–30: agree one small next step

Not "give it a go". Something specific and small, on their terms:

• "Would you try it on [one narrow task] once next week, and tell me honestly how it went?"
• "Would you review the guide for your team's workflow and tell me what's wrong with it?"
• "Can I quote your concern, anonymously, in the guidance? It's a better warning than anything I'd write."

The six objections you'll meet most

1. "It gets things wrong"

Often true. Don't argue. Separate tasks where errors are easy to catch from tasks where they're not. "You're right, it's not reliable on facts you can't check quickly. Where it tends to help is restructuring or drafting from material you've given it, where you can see whether it's right. Does any of your work look like that?"

2. "It's quicker to do it myself"

For some tasks, for some people, it is. Offer to measure: "Could we time one both ways? If it's not quicker, I'll stop suggesting it for that." You learn something either way.

3. "It'll replace us" (or the quieter version: "what's this really for?")

This is usually the real objection behind a different stated one. Be honest about what you know. You probably don't make headcount decisions and shouldn't pretend to. "I can't promise what every future decision will be. My remit is helping people get better at their jobs with these tools. I'd rather you were the person in your team who knows how to use it well than not." If there's genuine organisational ambiguity here, take that back to your sponsor. You can't fix it in a one-to-one.

4. "We're not allowed to use it on our work"

Often a policy problem, not a scepticism problem. Check what they believe the rules are. If it's a misunderstanding, clarify it. If it's genuinely unclear, that's a gap in your usage policy and they've just done you a favour.

5. "Our work is too specialised"

Sometimes true for the core judgement, rarely true for the surrounding tasks: summarising notes, drafting the cover email, reformatting for a different audience. Ask: "Which parts of your week aren't specialised?" If the core really is out of reach, say so. That's useful to know.

6. "I don't have time to learn another thing"

A real constraint, often a manager problem. Offer something short and specific to their workflow rather than general training, and check whether their manager has made the time legitimate. If not, that's a sponsor conversation.

After the conversation

• Write it up the same day: their role, the stated objection, the real objection, what you agreed.
• Do what you said you'd do, quickly. If you promised to check the policy, check it and reply within days.
• Feed it into your work. Real objections go into workflow guides as warnings, into your policy as clarifications, into your report as risks.
• Follow up in a fortnight, on the small next step only.

Over time, a sceptic who has been listened to, and whose concern changed something, often becomes your most credible advocate. Not because you persuaded them, but because they can see the programme takes quality seriously.

Mistakes to skip

• Treating scepticism as a training gap. Some sceptics know the tools better than the enthusiasts.
• Arguing with the stated objection. You'll win the argument and lose the person.
• Promising things you can't deliver, particularly about jobs.
• Going round them. If you work with their team while bypassing them, you've made an enemy with influence.
• Ignoring what they told you. The fastest way to confirm their scepticism.

At the end you should have

• Conversations with your most influential sceptics, written up
• A list of real objections, separated from stated ones
• At least one change to a guide, the policy or your plan as a result
• A small, agreed next step with each person, and a follow-up date
• A clearer view of where AI shouldn't be pushed, which is as useful as where it should

The Role Book includes the full sceptic interview template, with follow-up questions for each objection type and how to turn a reviewed sceptic into a team champion.