This is the sort of news that founders I mentor always want to talk about, so I thought it worth a blog post: Microsoft is offering voluntary buyouts to around 60,000 employees. That’s more than a quarter of its global workforce (it had about 223,000 full-time staff at the end of June), and it’s open to people at level 70 and below in participating teams in the US, Canada, the UK and India, with a deadline in early December. The coverage links it pretty directly to the cost of AI: data centre spending across the industry is likely to top $700 billion this year, and somebody has to pay for it.

This isn’t Microsoft’s first go at it either. Earlier this year a voluntary retirement programme was offered to more than 8,700 US employees, and roughly a third of them took it. Then came 4,800 job cuts in July. Reports also say the company expects far fewer people to accept this time around than are eligible.

I’m not going to pretend I know Microsoft’s internal plans (I don’t), and I have a lot of respect for the engineers there. But the question stories like this always raise with founders and CTOs isn’t really about Microsoft. It’s this: “if a company that size is reshaping itself for AI, should we be doing the same?” So let me share how I’d answer it.

Buyouts Are a Blunt Instrument

Here’s the thing about voluntary buyouts: you don’t get to choose who leaves. The people most likely to take the money are the ones with the best options elsewhere, and those are very often the people you’d least like to lose. The ones who stay are, quite likely, the ones who are least confident they’d find something better. You can see the problem here.

Don’t get me wrong, buyouts are a lot more humane than a surprise layoff on a Wednesday morning, and I’ll give Microsoft credit for that. As a cost-cutting tool, they work. As a way of reshaping an organisation for a different way of working, they’re closer to shaking a tree and seeing which apples fall.

Decide the Shape Before the Size

The reorganisations I’ve seen work, as a CTO and since, start from the work rather than from the headcount: what are we going to build, which decisions need a human, which ones don’t, and who owns what. Only then do they get to how many people that takes. The ones that start from a number (“we need to be 15% smaller”) tend to end up with the same org chart minus some boxes, and the same problems with fewer people to solve them.

AI makes this even more important, because the shape of the work is genuinely changing. In the teams I see doing well with AI, the engineers spend less time typing code and more time writing specs, reviewing what agents produce and making calls on trade-offs. That needs fewer layers, not just fewer people. If you take out 20% of your staff and keep every approval step, every steering committee and every “let’s take this offline and reconvene next week”, you haven’t become an AI company. You’ve become a smaller version of the old one.

Speed Is the Real Prize

And this is where I’ll be a bit blunt. I firmly believe that AI’s biggest gift to a company is velocity, and that companies which still make decisions by committee, sitting on a decision for days while it does the rounds, are going to get swallowed by smaller, agentic companies that decide in an afternoon and have agents shipping by the evening. A 20-person company running a few hundred agents doesn’t need a quarter of its staff to leave. It never hired them in the first place.

Sure, a company the size of Microsoft can afford a few slow decisions; it has the cash, the customers and the distribution. Most of the companies I talk to can’t. For them, the useful lesson from this story has very little to do with cutting a quarter of your people, and a lot to do with looking hard at how long it takes to get from “good idea” to “in production”, and removing every step in between that doesn’t actually add anything.

So, if you’re a CTO or a founder looking at this news and wondering whether you should be doing the same: start with the decisions, not the headcount. Map how work actually flows through your team today, find the places where it sits waiting for a human to say yes, and ask whether that human (or that meeting) is really needed. Fix that first. The size of the team will sort itself out.

Bottom line: cutting people is easy, any spreadsheet can do it. Cutting the time it takes your company to make a decision is the hard bit, and it’s the bit that actually matters 🙂