AI is not an excuse for layoffs — it's leverage on people

"How much headcount can we save?"

When AI transformation comes up, this is the first question we hear most often. The people asking aren't cold-hearted — mostly they're being honest. Payroll is an SME's biggest cost; AI is the loudest cost-saving hope of the moment; the two meet, and the thought forms itself. The question isn't wrong. But it carries a hidden price: it locks in the direction.

Same tool, two forks in the road

Follow "save headcount" and the path runs roughly like this: process unchanged, output unchanged, AI substituted into some of the seats. The first quarter or two look great on paper. Then the people who remain work something out: teaching the AI means training their own replacement. So they stall where they can and conceal what they can, and the most capable ones start returning recruiters' calls. The money saved on paper leaks back out — through morale, through quality, through attrition.

The other road treats AI as leverage: keep the team roughly intact and multiply what each person produces. The same people handle more clients, afford more experiments, dare to quote for work they used to decline. The hours saved are not profit. They are ammunition.

The question worth asking isn't "how many people can AI save" — it's "how much more can these people, plus AI, get done".

Leverage means moving time, not adding hours

Spread an employee's day on the table. Typing, formatting, collating, hunting for files — the mechanical work usually claims the biggest share. Reviewing, triaging, following up by rule takes another slice. The work that actually needs a brain — pricing, negotiating, telling a customer no — gets squeezed into the gaps.

Today's AI can swallow most of the first category, lend a hand with the second, and hand the person over, whole, to the third. That is all "people leverage" really is: moving human time out of the first two categories into the last. Move it, and one person doing the work of three is no exaggeration. Fail to move it, and every new tool is just a different screen to type into.

An entrepreneur who rebuilt an export business on AI is the extreme case: one person plus AI runs more than eighty websites, with lead-finding, background checks, outreach emails and inquiry analysis all handed to AI. Ask him where he won't let go, though, and the answer is immediate — pricing. AI gets sums wrong and will happily use a two-year-old exchange rate on today's order, so every quote is checked by him before it leaves. Leverage multiplies output; it doesn't multiply responsibility. That small piece he keeps for himself.

Some roles really will disappear

Having said all that — pretending no role will disappear is its own kind of dishonesty. Jobs that are pure entry, pure transcription, pure forwarding will get harder and harder to justify. The responsible move isn't promising "not one person fewer". It's giving those willing to learn a new skill the chance to, giving those who leave a dignified exit, and saying all of it early. If you don't manage the expectations, the rumours will — and rumours always offer worse terms than you would.

Management goes first

One more fact that tends to get stepped around: whether staff use AI is mostly decided by management — not through orders, but through three small things. Has the process been redrawn? If not, AI is an engine bolted onto the old carriage. Have appraisals changed? If performance is still counted in pieces processed, handing work to AI is self-harm. Does the boss use it? Demanding enthusiasm from everyone while never opening the tool yourself is an unconvincing revolution. None of the three is difficult. All three have to be done by management personally.

Closing

Two forks, same tools; the split happens at that first question. At the next kick-off meeting, try swapping it — from "how many people can we save" to "what more could these people get done". The answer takes the company somewhere very different.