Layoffs are what prioritization looks like when you do it too late

On July 22, Amazon confirmed layoffs inside its AGI organization — the group chasing frontier AI research. In the same breath, it reaffirmed roughly $200 billion in AI infrastructure spend for the year. The company wasn't retreating from AI. It was, in its own words, refocusing on its "highest-conviction" bets: Nova foundation models, agentic shopping features like Rufus and Alexa+, and internal AI infrastructure. Everything else in the AGI portfolio — including model customization and post-training work — got cut.

That same week, monday.com cut 630 people, about 20% of its staff, to concentrate on its AI Work Platform. Co-founder Eran Zinman was explicit that it wasn't a cost play — it was the company narrowing its bets. Two unrelated companies, same week, same move: fund fewer things, harder.

The tell isn't the layoff. It's the timing.

Neither company is claiming these projects were bad ideas. Amazon didn't say AGI research was worthless — it said the org had too much of it running relative to what it could actually stand behind. That's not a performance problem. It's a portfolio problem, and portfolio problems are supposed to get solved at the front door, before headcount, budget, and eighteen months of roadmap get committed to a project.

Instead, it got solved at the back door, in public, with severance packages.

This is worth separating from the broader "AI layoffs" narrative currently running hot — Sam Altman himself has pointed out that "almost every company that does layoffs is blaming AI, whether or not it really is about AI." Some of that is real washing: cost cuts wearing an AI-strategy costume. But the Amazon and monday.com cases are closer to the opposite problem. These are companies that genuinely over-said-yes. They greenlit parallel bets without a mechanism for ranking them against each other, and now the ranking is happening anyway — just with more collateral damage than it needed to have.

Prioritization has to happen before the commitment, not after

A weighted scoring pass at intake — value, cost, strategic fit, risk of the thing never shipping — doesn't eliminate hard calls. Amazon would still have had to choose Nova and agentic shopping over other bets. But that choice is cheap when it's made before a team is staffed and a roadmap is public. It's expensive — reputationally, financially, and for the people involved — when it's made after.

The pattern shows up below the Amazon-scale headlines too. Any team running project or feature intake without a consistent scoring pass tends to fund things in the order they got pitched, or in the order the loudest stakeholder pushed, rather than in the order that actually matters. The backlog grows past what the team can execute well, and eventually something forces a correction — a budget cycle, a reorg, a bad quarter. The correction is never gentler for having been delayed.

The uncomfortable version of this: if your org needs a layoff round to tell you which projects mattered most, your intake process — not your AI strategy — is the thing that failed first. Scoring criteria that force an honest ranking before work starts don't just make prioritization easier. They make the ugly, public version of it unnecessary.

That's the whole bet behind treating intake as infrastructure rather than a form: the goal isn't to say yes faster, it's to build the muscle that says no early enough for it not to cost anyone their job.

Sources: CNBC, "Amazon lays off some employees in its AGI unit," July 22, 2026; GeekWire, "Amazon cuts jobs in AGI group as it puts more focus on customer-facing AI," July 2026; TechCrunch, "monday.com lays off hundreds to focus on AI," July 22, 2026.