Stripe just opened a job that says more about the next two years of work than any thinkpiece will.
It's called the Forward Deployed AI Accelerator. The role pays multiple six figures. It sits inside the marketing team. Not IT, not enablement, not some central transformation office. The person hired into it is given a cohort of about twenty marketers and a clear measure of success: how many of those marketers' workflows have been permanently changed, and whether AI is now the first tool they reach for on any task.
That's the entire job. Embed with twenty people. Watch how they actually work. Build the agents, automations, and custom tools that match their specific responsibilities. Coach each one from their first AI win through to being able to build and iterate on their own tools without you. Then generalise what worked so the rest of the cohort can pick it up.
It's a deceptively simple structure, and it's the most honest read on AI adoption I've seen from a major company.
One job posting at one company isn't a trend on its own. Stripe is well known for trying organisational structures that don't catch on elsewhere. But the interesting thing here isn't the role itself. It's that a marketing team, not an engineering team or a central AI office, has formalised the embedded build partner as a permanent function. That's the part worth paying attention to.
Why this matters
Most organisations have been running the same AI playbook for the last two years. Buy licences. Run a training session. Send out a Loom video. Hope.
It hasn't worked. The licences sit unused. The training is forgotten by Friday. A small group of motivated people figure things out on their own and the rest of the business carries on exactly as before. The gap between what the technology can do and what people actually do with it keeps widening.
Stripe's bet is that this gap doesn't close on its own. People will not, in any meaningful number, upskill themselves into a fundamentally new way of working. Not because they're lazy or resistant, but because the tools aren't shaped to their actual job. A marketer running campaign reporting doesn't need a generic chatbot. They need an agent that knows their attribution model, their channels, their reporting cadence, and the specific spreadsheet they update every Monday. You don't get there from a training video. You get there by sitting next to that marketer, watching them work, and building the thing.
The structural insight
The bit most companies will miss is where this role sits. Stripe didn't put it in IT. They didn't put it in a central AI Centre of Excellence. They embedded it directly in the marketing org.
That placement is the whole point. The person doing the work needs to understand marketing as deeply as they understand the technology. They report into the function they're transforming. Their incentives are aligned with the people they're building for, not with an abstract organisational metric.
This is the same pattern that made forward deployed engineers work at Palantir and made solutions architects work in enterprise software. Embed the technical person in the customer's reality. The model is well understood. Stripe is the first major company applying it internally, at scale, for AI.
Where this goes
In twelve months, every marketing org of any size will be hiring some version of this role. Then sales. Then operations. Then finance.
The job title will vary. The structure will be the same. One AI native operator embedded with a cohort of twenty, paid well, with a measurable mandate to transform how that cohort works.
The role probably won't exist five years from now, at least not in this form. By then the expectation will be that everyone in a knowledge work job operates this way natively, the same way we expect everyone to use email. The Forward Deployed AI Accelerator is a transitional role. But it's the right transitional role, and the companies that hire for it now will be eighteen months ahead of the ones still running training sessions.
The obvious objection is that the full Stripe model only works if you can afford it. One embedded operator per twenty people, at multiple six figures each, is not a line item most businesses can absorb. That's fair. But the principle scales down. One person across several functions instead of one per function. An external partner running the same playbook for businesses that can't justify a permanent hire. The structure adapts. The underlying logic, that AI adoption requires someone building alongside the people doing the work, doesn't change with company size.
If you're running a team and you've been waiting for the AI productivity gains everyone keeps promising, this is the answer. Stop asking your people to upskill themselves. Find someone whose only job is to build alongside them.
That's where this is heading. Stripe just said it out loud first.