I'm the President of the HelloFresh Group and sit on our management board.
Over the last 6 months we've quietly rebuilt how we operate with AI. Not a pilot. Not a slide in a board deck. Real P&L impact.
Here's what's working, what broke, and what nobody warns you about 🧵
@amasad Agreed, and it's not abstract for us. Running production workloads across multiple models means no single vendor's pricing or roadmap can hold the business hostage. Openness at the model layer is operational leverage, not a policy stance.
@emollick From inside a company actually shipping on these models: the geopolitics framing is real, but day to day we're picking whichever model clears the bar for the task, not whichever flag it flies. Access and diffusion are the actual fight.
The winners over the next few years won't be the ones who bought the biggest AI program.
They'll be the ones whose people stopped asking permission.
Culture is grown, not installed. We're a tech-food company betting everything on that difference.
Every few weeks someone tries to sell me an expensive "AI transformation."
I say no every time. Our AI adoption keeps accelerating anyway.
Here's what nobody selling "transformation" will tell you 🧵
The hard part was never the tech. The models are ready.
The bottleneck is always the org: its approval chains, its politics, its tolerance for "good enough today" over "perfect next quarter."
No transformation company sells you past that.
@lennysan@netflix This tracks with what we're hiring for now too. Less "expert in one layer," more people who zoom out, see how their piece fits the system, and build scaffolding others can move fast on. That's the real AI-era skill, not prompting.
@CoinDeskPodcast@Mastercard@RajaRajamannar@jpegtrading The "two audiences" framing is sharp — when a customer's AI agent does the choosing, you're marketing to a model as much as a person. Most brand playbooks aren't built for that. Might be the biggest unsolved problem in marketing right now.
@SparkNovus "Raises the production floor, not the idea ceiling" — exactly. AI won't erode brand trust; lazy sameness will. When everyone can generate infinite competent content, the scarce thing is a point of view. That's still a human job.
@Crowdreply_io This is the shift most CMOs are sleeping on. Getting cited in AI answers is becoming the new front page — while everyone still optimizes for a search box users are abandoning. Visibility inside AI is a board-level question now.
@Tech_babby More real than the skeptics think — turning a competitor's ad into on-brand creative in minutes isn't a demo anymore, it's routine. But the tool was never the moat. When anyone can generate the ad, the edge is taste: knowing which one is worth shipping.
@linuxfoundation The payment rail is the easy half. The harder question for anyone running this: when an agent transacts on your behalf, who's accountable for the call? Agentic commerce needs a trust-and-accountability layer as much as a payment one.
@emollick It's more than fun — it's a real budgeting shift. Compute is becoming a line item leaders forecast, like cloud spend a decade ago. The teams treating token spend as capital allocation, not an afterthought, will have an edge.
@garrytan Agreed — and it's not just a builder concern, it's a board one. At scale you can't bet the company on a single vendor's roadmap. The abstraction layer that keeps you portable is real strategy, not plumbing.