Invest in AI and cut costs now
Two instructions land in the same week, and they pull in opposite directions. Spend aggressively on AI, because the market rewards it. Take cost out of the operation this quarter, because the market rewards that too. Everyone at the table nods, and then someone has to go and make both true.
That someone is usually you. So let me answer the question a board actually asks, in the words they use: how much headcount does this save, and when?
The question behind the question
The honest answer for the first year is close to none, and that answer is easier to give with evidence in hand. Layoffs attributed to AI went from roughly 18,000 in 2024 to more than 100,000 in 2025, and past 150,000 in the first six months of 2026. Challenger found AI named as the leading reason four months in a row, reaching about 40 percent of all American layoffs in a single month. Those are real people and real announcements, and they shape what your board reads on a Sunday.
What those announcements describe and what happens inside the operation are two separate things. Keep both in view when you answer.
What Klarna actually did
Klarna is the case everyone quotes, and the full version is more useful than the headline. One assistant did the work of 700 agents and handled two thirds of all conversations within weeks. Headcount dropped from around 5,500 to 3,400. Then the company moved back to a hybrid model with people in the loop for the conversations that needed them.
Read that as a sequence, not a verdict. Volume moved first. Quality caught up second. The operating model settled third, and only then did the numbers mean anything.
The gap between the announcement and the payroll
Gartner asked 321 customer service leaders what AI had done to their workforce, and only about 20 percent had actually reduced headcount. Four in five had not, while nearly all of them were running pilots and buying licences.
NBER went further and asked executives in confidence. About nine in ten said AI had done nothing to employment in their company.
The press release and the payroll are written by different people, and only one of them is checked.
Meanwhile Block attributed some four thousand roles to intelligence tools, and the share price rose 24 percent that day. The same group of companies is heading toward seven hundred billion dollars of AI infrastructure spend in 2026. Both facts are true at once, which is exactly why the paradox feels unsolvable in the room.
The capability you are being asked to buy
You reconcile the two instructions by building something you do not have yet, and you can name it on one slide. It is the ability to put a model in front of live customers, measure what it does to resolution and trust, and change it weekly. That is people, data access, decision rights and a release rhythm.
Give the board a shape they can fund:
- One journey, end to end, with a named owner who can say stop.
- A baseline taken before anything ships: handling time, resolution, repeat contact.
- A weekly review where the numbers move or the design changes.
- A written rule on which decisions the system may take alone.
Cost comes out of the second journey, once the capability from the first one is reusable. Say that out loud, with the Gartner number next to it, and the conversation turns from headcount to sequence.
What you do on Monday
- Write the board's question on a page and answer it in three lines, with the 321-leader finding as your evidence.
- Pick one journey with high volume and low risk, and take the baseline this week.
- Name the owner, the reviewer and the weekly slot before you name the vendor.
- Put the Klarna sequence in your plan: volume, then quality, then operating model.
- Book thirty minutes with finance to agree what counts as savings and when it lands.
The hardest part of this is the business case, because it has to survive a CFO who has heard the promise before. The CX+AI Business Case template walks through the numbers in the order a board wants them: capability first, savings second. The book works through the same paradox with the full evidence behind it. Bekijk de materialen