The CEOle squirrel holding a single acorn coin next to a sapling growing from planted acorns

Headcount is a one-time return

Once owners stop worrying about being behind on AI, the next question is what it will actually save them, and the first number on the whiteboard is almost always a person. Sometimes it's said out loud and sometimes it's just implied by the spreadsheet, but the business case usually ends with "and then we won't need that role." I get it, because I've been running services businesses for over 20 years and payroll is the biggest line I look at every month too.

Honestly, cutting people is about the weakest return AI can give you, and I'm not being soft, the math just doesn't work.

The math on cutting a role

You can only remove a position once. The savings show up in year one, and then there's nothing new in year two, because you can't cut the same job twice.

It's also capped, because the most you could ever save this way is your payroll, and nobody gets anywhere near that number while still serving customers, which is fine..

And it works against you, because the people you'd cut are the ones you need for the next opportunity, so you get a one-time bump and lose what keeps paying.

The three returns that actually compound

This framing comes from Graeme Crouch's AI readiness work, and he puts it better than I do. The returns that build on themselves fall into three buckets:

  • Capacity: more work through the same people, including the work you're currently skipping
  • Capability: better work from the same people, like juniors producing senior-quality output and new hires getting up to speed faster
  • Continuity: the people who know how things really work stick around, so you carry less key-person risk

Those are the three things a business actually runs out of... hours, the right skill in the right seat, and people who know where the bodies are buried (figuratively, I hope).

What gets me is that cutting headcount doesn't ease any of those three, and it makes all three worse. You end up with less capacity, less capability walking around the office, and the people who stay are covering more ground while wondering if they're next.

Where I've been guilty of this

I've made the shortsighted version of this call before, long before AI was in the picture, with "efficiency" projects that looked great on the spreadsheet (I can be very persuasive when the only one I'm arguing with is a spreadsheet). The spreadsheet never shows the bill that comes later, like slower delivery, tired people and knowledge that walked out the door.

And I've done the real version too. In 2019 we grew from 50-ish people to 77 at our max, we weren't ready for that, and we had to lay off 25 to 30 people at the end of the year just to survive. We hired a CFO who was brutal, she was kind of mean, but she was exactly what we needed, and she's probably the reason we survived. She took away all of our credit cards #FunTimes. We got insanely close, but we never once missed payroll, not in 27 years. As a leader, I think it was Simon Sinek (not Simon Says, though I've played that too) who said you're not in charge of people, they're in your charge. Laying off that many people was insanely stressful, and it's the reason I don't buy a business case that starts with cutting a name off the org chart.

Start with why you want AI

There's an interesting question underneath all of this, which is why you want AI in the first place. AI can make things more efficient and solve real issues, but do you want fewer people or more? And do you have good people?

If you don't have good people, you have a management and business problem, and you should solve that first instead of using AI as an excuse to make those kinds of changes.

Ultimately, AI should be used to make good people better and to increase automation, but use it carefully. It's not just about AI, it's a tool, and you have to make sure it's the right tool for the right job. Do you need something deterministic that gives you the same answer every time, or are you okay with some creativity and that kind of stuff?

What to ask instead

When you build the AI business case, decide what the freed-up capacity is for before you deploy anything. If the answer is more clients without adding people, faster onboarding, or getting month-end out of one person's head, you're building something that pays you again next year.

If the answer is cutting a name off the org chart, that's a one-time win, and next year you're back at the whiteboard.

So what would you actually go after if you kept everyone?

Follow along with the RSS feed, or talk to me about it on LinkedIn.