Amazon's CEO wrote down what most executives only say privately

An Authority Report is where we read what people with real operating responsibility said in public, in their own words, and work out what it means for a company that isn’t them.

On 17 June 2025 Andy Jassy sent a memo to Amazon employees and published it on the company’s own site. That last detail matters: this is not a leaked internal document or a reporter’s paraphrase, it is a primary source, which makes it the most reliably quotable statement in the entire workforce debate.

What it says

The sentence everyone took is careful in a way the coverage was not:

We will need fewer people doing some of the jobs that are being done today, and more people doing other types of jobs.

And on the aggregate:

It’s hard to know exactly where this nets out over time, but in the next few years, we expect that this will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company.

Two hedges in one paragraph (“hard to know exactly where this nets out”, “some of the jobs”) sitting next to one firm directional claim. That is the shape of a statement written by people who expected it to be quoted for years.

The memo also gives a number that got less attention than it deserved.

More than 1,000 AI services and applications built or in progress, described as a small fraction of what Amazon plans.

Jassy’s memo, published by Amazon on 17 June 2025

The number is the story, not the sentence

A thousand services is the part a mid-sized company should actually think about, because it says what the model of adoption is.

It is not one transformation programme. It is a thousand narrow things, each attached to a specific process, built by the teams that own those processes. The workforce claim in the memo is a consequence of that model, not a plan of its own, nobody at Amazon set out to reduce headcount by X; they set out to build a thousand things and reported what they expect to follow.

That distinction is the difference between a company that gets results and one that runs a pilot. It is also the finding behind the 95 % of enterprise AI pilots that produce no measurable result: the failure is almost never the model, it is that the effort was organised as a programme rather than as a thousand small things owned by the people doing the work.

What Jassy told employees to do

The practical advice in the memo was to be curious about AI, learn it, attend the internal workshops and training. That is unobjectionable and thin, and it is worth noticing how thin it is.

A memo that says the workforce will shrink and then advises curiosity is asking individuals to solve a structural change with personal initiative. It is the standard move, it is not dishonest, and it does not amount to a plan for the person reading it.

The version of this that is actually actionable for a company of any size is more specific: name the processes, name who owns each one, and make the question “should a system do this part” a normal part of owning it. That is what a thousand services looks like from the inside.

What this means at 400 people instead of 350,000

The mechanism transfers. The framing does not. At Amazon’s scale, aggregate workforce effects are measurable and worth stating. At 400 people, the same statement is a morale event with no operational content, and the honest version of the same trend is “we will not need to hire the next three people for this”, which is a different sentence and lands differently.

A thousand narrow things beats one big thing, at every scale. Your version is ten, not a thousand. Same principle: each attached to a process someone owns, each small enough to finish.

Say the specific thing or say nothing. The most damaging version of this memo at a small company is the vague one, an all-hands where AI and headcount appear in the same sentence with no detail. People fill the gap, and they fill it worse than the truth. Either name which work is affected, as Duolingo did and then had to correct, or do not raise it.

Watch the second half of the sentence. “More people doing other types of jobs” is in the memo and disappeared from almost every headline. Whether it turns out to be true is the actual question, and at your scale you will find out much sooner than Amazon will.

The read

Jassy’s memo is the clearest public statement of an expectation many executives hold and few write down, and its value as a citation is exactly that: it is on the company’s own site, hedged carefully, and impossible to dismiss as a misquote.

Its value as a template is close to zero. What is worth copying is not the announcement but the model underneath it, a thousand narrow systems owned by the teams whose work they change. The announcement is the part that generates headlines. The thousand services are the part that generates the result.

Frequently asked questions