Will this replace my team? How to think and talk about it

Before you can have the conversation with your team, you have to finish the one inside your own head.

There is a version of this question that I asked publicly and a version I asked in the dark at 11pm, sitting with my laptop, while the office was empty and no one could hear me work through it honestly.

The public version is clean: "How do I frame AI for my team?" The private version is messier: "Am I about to hollow out jobs I promised to people? And if so, am I lying to them right now by saying nothing?"

This piece is for that 11pm version of you. Not the all-hands conversation (that comes next, and I've written that one separately). This one is earlier. This is the private reckoning that should happen before you open your mouth in any room with more than one person in it.

You cannot give your team a straight answer until you have one yourself.

The honest answer at your scale

The first thing to get clear on is what is actually happening, on the evidence that exists, for a company your size.

At the $1-50M scale, the near-term pattern is task redistribution and augmentation, not wholesale elimination.

Roles shift. Some tasks disappear. Headcount-per-output rises. You will hire differently going forward than you did three years ago. All of that is real and worth saying.

But the constraint at this scale is almost never "we have too many people." It is "we cannot grow fast enough." The freed-up leverage, when you point it at growth, does not produce fewer jobs. It produces the same number of jobs doing bigger things, or more jobs because you can grow faster. That is the honest pattern.

The research supports it directionally, with the caveat that SMB-specific data is thin and most studies are run on larger organizations. The Harvard/BCG jagged-frontier study (which I also referenced in the myths piece) found that AI meaningfully lifted speed and quality on tasks inside its capability frontier, and actually hurt performance when teams over-trusted it outside that frontier. The risk is misapplied confidence, not a machine replacing the person. BCG has framed AI's broader trajectory as reshaping more jobs than it replaces, with their explicit note that this is not an unemployment forecast. And HBR research found that when employees perceive AI as coming to replace them, they disengage and quit at higher rates, before any actual layoff happens. The replacement signal costs you trust and headcount before you ever make a cut.

For the full picture of what AI can and cannot actually do, that explainer is worth your time. The short version: AI handles volume work, drafting, pattern-matching across large amounts of text, structuring, formatting. The things your team does that AI cannot are real judgment calls, novel decisions, reading a room, and anything that requires understanding stakes you haven't explained in advance. The people are not the redundant part.

One more data point worth holding: Sam Altman walked back some of the more apocalyptic job-loss predictions as real-world deployments showed augmentation rather than replacement as the dominant pattern. And firms that cut headcount explicitly to offset AI costs did not outperform firms that kept their staff and used AI to amplify them. The returns come from the amplification, not the cutting.

That is the honest answer. Not "nothing will change." Not "half of you are gone by next year." Something harder and more specific: tasks will shift, some will disappear, you will hire differently, and the play is to point the leverage at growth rather than at reduction.

The posture trap

The reason most CEOs struggle here is not that the honest answer is unclear. It is that the honest answer requires a posture, and most CEOs default to one of the two wrong ones.

False reassurance sounds like this: "AI is just a tool. Nobody's job is at risk. We're augmenting, not replacing." I have heard this said with complete sincerity by CEOs who privately knew it wasn't quite right. The problem is not that they were lying with intent. The problem is that their employees already knew it wasn't quite right too. Every person on your team has seen what AI can do. They have used it themselves, probably more than you have. The moment AI absorbs a task that someone owned, the "nobody's job is at risk" line gets marked as false in their memory, and everything you said after it gets discounted backward.

Fear-mongering sounds like this: "AI will replace half of you eventually." I have seen CEOs say versions of this because they thought honesty required maximum candor, or because they were genuinely afraid themselves and said more than they intended. Setting aside that it's probably wrong at your scale, it triggers the exact disengagement you're trying to avoid. The HBR finding above is worth re-reading here. The replacement signal costs you before the replacement happens.

The posture that holds is the third one, and it is harder to deliver because it requires holding real uncertainty without either collapsing into false comfort or inflating into theatrical warning. It sounds roughly like: "AI changes what we spend time on. Some tasks go away. We will hire differently. The deal is more leverage per person, reinvested in growth, and I will be straight with you as roles evolve. I will not make promises I cannot keep."

That last sentence is the whole thing. Not a promise that nothing changes. Not a warning that everything does. A commitment to honesty as the situation unfolds.

The gap between what you believe and what you say

Here is the integrity problem I see most often in CEOs who are otherwise very straight with their teams.

Privately, they have concluded: some roles will shift. A few may shrink. We will hire more AI-literate people going forward. Some current roles may not exist in the same form in three years. All of that is plausible and probably right.

Publicly, they say something blander. "Augmentation, not replacement." "Everyone's job is safe." The version that dodges the harder conversation.

The gap is not that they are lying outright. It is that they are managing the message rather than telling the truth, and their teams notice. People who work closely with you are reading your register, not just your words. When your private belief and your public statement don't match, the mismatch is legible, even if the content isn't.

The practical advice: you do not need to know exactly which roles will shift or when. But you do need to commit to candor rather than immunity.

"I don't yet know exactly which roles will look different in two years. Nobody does. What I can promise is that I will tell you what I see as I see it, and I will not give you a guarantee I can't keep." That is a real commitment. The false guarantee that later gets disproven is far more expensive than honest uncertainty delivered now. Silence that lets anxiety fill the vacuum is the second-worst option.

You are not obligated to announce what you don't know. You are obligated not to reassure with things you don't believe.

What to do with this

The private work here is getting your own beliefs sorted before you open your mouth in a meeting room. Three things to work through before you say anything.

Get clear on your actual posture on headcount. Is your plan to use AI to grow faster without growing the team? To grow the team differently? To genuinely not know yet? Each of those is a different thing to say. The worst version is to have no view and then improvise one on the spot under question.

Decide what you can honestly commit to. Not "nobody's job will ever change," because you don't know that. Something specific: the direction on growth, the timeline before any significant role decisions, the process by which people will hear about shifts before they become surprises.

Work out the gap between your private belief and your public statement. If those two things are significantly different, that is worth sitting with before you go into any conversation with your team. The gap doesn't close by talking yourself into a rosier private belief. It closes by bringing the public statement closer to what you actually think.

Once you have done that work, the all-hands conversation is a different meeting. Not easy, but at least honest. The piece I wrote on that covers the three fears your team has and isn't saying, the five conversations that handle them, and the specific mistakes that look like kindness and function as cowardice.

If you want to understand the adoption mechanics before you make any commitments about what AI will or won't do to roles, how to get your team to actually use it is the practical next step, and the 30-day rollout framework is where the implementation begins.

The 11pm question deserves a real answer. Get yourself one first.

Andrew

Related reading

• How to talk to your team about AI · the all-hands conversation, done well
• The things non-technical CEOs get wrong about AI · including the "AI will replace my team" myth
• AI capabilities and limits for CEOs · what AI is and isn't actually good at
• How to get your team to actually use AI · the adoption mechanics after you've announced it
• Where to start with AI as a CEO · if you're still figuring out where to begin