Is AI sparking a headless renaissance?

ecommerce

Brands selling online have spent the past decade asking themselves: freedom or simplicity? Turns out it was a trick question.

Jenna Holcombe, Co-Founder and Chief Brand Officer

Headless is dead! Long live headless! I designed my first online store for a fashion brand in 2009, before DTC was even a term. A few years later, my co-founders and I started building a headless platform for fashion. And over the years, I've watched headless commerce go from the industry’s hottest idea to something brands were warned away from. Now, with AI agents entering the workforce, I couldn't help but wonder: is headless having its renaissance?

When brands choose a platform, one of the very first questions they face is also one of the biggest: headless for freedom, or all-in-one for simplicity? Headless means the storefront your customers see is built separately from the commerce engine behind it, and the two talk through APIs. All-in-one means the platform gives you both, front end included.

Fashion brands that go the headless route don't do it to follow a buzzy tech trend, and they know full well the trade-off they're signing up for: more work at the start and a front end of their own to maintain, in exchange for total design freedom and a site that loads faster than anything off the shelf. For creative companies that know the power of their brand, giving up control is never even an option. They're hellbent on giving customers a true flagship store experience, whether online or IRL, and a themed storefront with the same product grid as every other store on the internet just can't (never ever will) deliver that.

Headless was never the villain

Somewhere along the way, headless became a dirty word. We all heard some version of the same story, now an urban legend in commerce communities: the replatforming project that ran for a year and never launched😱. All-in-one platforms were more than happy to keep the lore in circulation, and before long, headless was shorthand for complexity, cost and risk.

Still, headless wasn't the true villain of these cautionary tales. The idea itself was sound: creative freedom at the front, and everything accessible through APIs. But trouble was brewing in the back end.

Many headless projects doubled down with a fully composable back end, built from best-of-breed services and connected one by one with bespoke integrations. Every new service meant another connection to maintain and another vendor to call when something broke. Meanwhile, all-in-one platforms were serving up complexity of their own, just better disguised. Their simplicity often depended on a growing stack of third-party apps to cover the gaps.

Either way, none of this showed up in the initial licence fee. But you'd better believe it showed up on the bill in years three, four and five: the hidden tax.

And here's the thing: AI agents are about to make that hidden tax a lot more visible. Whatever you call it — AI agents, agentic operations — it's bringing the headless debate back into focus, fast. In the agentic era, the brands that now have the edge are the ones that embrace headless architecture but keep their commerce core natively connected.

Centra news in your inboxSubscribe

From user experience to agent experience

Nobody joins a fashion brand because they delight in chasing missing attributes or reformatting linesheets. Now AI agents are joining commerce teams and taking the tedium out of the everyday work. More so, they're taking on jobs that cut across many departments: checking a collection before launch, reconciling refunds, working out why a product isn't selling in a market. To do that work, an agent doesn't click through an admin panel. It works through APIs — increasingly through the Model Context Protocol, the open standard that lets AI tools connect to software.

That changes how commerce platforms have to be built. As my co-founder Martin Jensen puts it: "We're no longer designing software only for humans. We're designing software for agents." We've had customer experience, user experience and developer experience. Now there's a fourth: agent experience. And agent experience starts with APIs.

Enter headless. On a headless platform, the storefront is just another application talking to the back end through APIs. To make that work, everything — products, prices, stock, orders — has to be reachable through code. So when agents arrived, they found platforms that already spoke their language. All-in-one platforms, built around their own admin screens, never had to make everything reachable through code. As a result, some features only live in the interface, others in third-party apps with their own access rules. Many also offer their own built-in assistant, but it can only see what lives inside that platform. An API-first platform that speaks an open standard like MCP works the other way round: your team brings the agent it already uses, and it can work across the rest of your stack too.

That gives the platforms that took API-first architecture seriously (and the brands that embraced them) a real advantage. And that advantage isn't just for the big players anymore. AI has made a custom front end something a small team can build, and agents have made a connected back end something a small team can run. Consider the headless playing field leveled.

But API-first isn't the whole answer

Give an agent a stack of separate systems, and it runs into the same walls your team always has: it can only see one fragment at a time. Ask it a simple commercial question — is this product actually sellable in Japan, and if not, what's blocking it? — and the agent has to search several systems (products in one, stock in another, pricing in a third), then reconcile what each one says before it can answer. That's assuming it can reach them all: each system needs its own connection, and some don't offer one yet. The more systems it has to check, the more likely the answer comes back incomplete.

If the past few years of AI hype have taught us anything, it's that AI isn't a silver bullet. The hidden tax of all those integrations and apps doesn't just disappear when agents arrive. It becomes the newest line item on the P&L: the agent tax.

Brands deploy agents expecting them to take on complex work across the business, then discover they can only finish part of the job. So they spend time and tokens solving pieces of problems, while the team fills in the rest. But with the right foundation underneath, the same agents can take a job from start to finish — and that's when they pay for themselves.

The back end is a bigger growth lever than ever

Yes, choosing an ecommerce platform has always been a commercial decision. But agentic operations make the foundation matter even more: once agents join the team, how your systems fit together decides how much they can get done. That makes it a conversation for the whole C-suite, not just the CTO.

Centra is headless, but it's also a suite of commerce systems: DTC, wholesale, product information and order management all share one data layer, and wholesale comes with a hosted showroom that's ready to switch on. Fashion's own logic — markets, size runs, seasons, delivery windows — is built into the data model rather than bolted on. So when an agent needs an answer that spans the webshop, the stockists and the warehouse, it's one question, not an integration project. And brands have noticed: in the weeks since we launched Centra MCP, more than 30 fashion brands have started running AI agents in their back office.

"With Centra and our monitoring tool connected, Claude can follow an issue across our platform, from the order and payment data in Centra to our own logs and code. Problems that used to take several people and several tools to untangle are now solved in one conversation." Andreas JonssonDigital Solutions Manager, Eton

"With Centra and our monitoring tool connected, Claude can follow an issue across our platform, from the order and payment data in Centra to our own logs and code. Problems that used to take several people and several tools to untangle are now solved in one conversation."

Teaching agents the job

Your team knows things no system could tell an agent: that a style with six sizes and one barcode isn't ready for launch, or that switching a wholesale account's price list quietly changes its currency. They know because they've been burned. But an agent doesn’t know this unless it’s trained to know.

So alongside the 300 agent-ready tools in Centra, we've written more than 80 playbooks: step-by-step routines for the recurring work fashion teams do every day, designed by people who have done that work. Think of them as the brief you'd give a new hire, written for an agent.

"Launching our wholesale business required us to replicate much of our D2C structure. With Claude and the Centra MCP, we populated displays, validated delivery windows and completed setup at a pace that simply isn't possible manually."Elin Mengarelli Hofmann Ecommerce and Sales Manager, Maya Delorez

"Launching our wholesale business required us to replicate much of our D2C structure. With Claude and the Centra MCP, we populated displays, validated delivery windows and completed setup at a pace that simply isn't possible manually."

Humans still welcome

If letting an agent near your data freaks you out, good. It should. The teams I talk to who are furthest along have landed on the same approach: they treat agents like a new hire. Limited access at first, more responsibility as trust builds, and a human signing off before anything changes.

Three questions to ask your ecommerce platform about AI agents

If you run commerce for a brand, AI agents are already on your agenda. Here are three questions worth asking:

• Can an agent answer a question that spans DTC, wholesale and stock in one step — or does it need an integration project first?

Can an agent answer a question that spans DTC, wholesale and stock in one step — or does it need an integration project first?

• Does it speak fashion — seasons, size runs, markets — or do your agents have to learn it from scratch?

Does it speak fashion — seasons, size runs, markets — or do your agents have to learn it from scratch?

• Can it act, and can you approve every change before it does?

Can it act, and can you approve every change before it does?

So where does that leave brands?

Headless was in, then out, and now AI is swinging the pendulum back. With freedom easier to build and a connected business simpler to run, that old familiar trade-off is starting to disappear. Freedom or simplicity? For the first time, brands may not have to choose. Headless is dead. Long live headless.

Jenna Holcombe is Co-Founder and Chief Brand Officer at Centra, the commerce platform for fashion brands. She has designed e-commerce experiences for fashion and lifestyle brands since 2009.