What Changes for Product Teams in 2027

The backlog version of the 2027 predictions

Paul van Oijen7 min readDesign Insights
What Changes for Product Teams in 2027

On December 2, fifteen days after this goes out, the EU's grace period ends for products already on the market that generate images, audio or text without machine-readable labels. It's a small deadline, and the 2027 trend reports won't mention it. A surprising amount of next year's product work looks like that.

We spend our weeks inside B2B and AI product teams, and their backlogs rarely look like the trend reports. Here are five things we expect those teams to deal with in 2027, with the dates and sources behind each, and what we'd do about them this quarter. The first is written into law. The rest are our bets.

Compliance shows up in the interface

Start with the part that has dates attached, which I suspect is the part teams outside Europe know least about. Several of these requirements become things a user sees on screen.

DateWhat appliesWhat it means for your product
Aug 2, 2026 (already applies)EU AI Act, Article 50 transparencyTell people when they're talking to an AI, and label AI-generated content
Dec 2, 2026Article 50 grace period ends for AI systems already on the marketMachine-readable marking of generated media for existing products too
Jan 1, 2027Colorado's replacement AI law takes effectNotice, explanation and human review for automated consequential decisions
Jan 12, 2027EU Data Act ban on cloud switching chargesSwitching duties already apply; from this date you can't charge for them
Dec 2, 2027EU AI Act high-risk obligations (hiring, credit, education and more)Human oversight, logging and explanations, visible in the product
Dec 11, 2027Cyber Resilience Act applies in fullSecurity by design, update flows and vulnerability handling

Sources: the European Commission on the AI omnibus, which moved the high-risk date from August 2026 to December 2027, plus law firm summaries from Jones Walker, Gibson Dunn, Goodwin and Jones Day. We're designers, so treat this as a product view and check the legal detail with counsel.

Accessibility belongs on this list too. The European Accessibility Act has applied since June 2025. Purely business-facing software is mostly outside its scope, but anything a consumer touches is in, and that includes the widget your customer embeds on their own site. WCAG 2.2 AA is still the standard to build to, since WCAG 3 is a working draft and unlikely to be final before 2028.

We're a Finnish studio, so we see these from close up. The teams that handle them well treat each one as a design brief: someone decides what the disclosure says and where it sits, weeks before the deadline.

Your next user might not be a person

In January, the Model Context Protocol shipped MCP Apps, which let a tool return a working interface inside a chat with Claude, ChatGPT or VS Code. The same month, Google previewed an "auto browse" mode for Gemini in Chrome that clicks through websites on a user's behalf. Agents now fill in forms and check out in products that were designed for people with a mouse.

The numbers are still small. DataDome measured AI agents and crawlers at 1.7% of web requests by mid-2026, up from 1.2% a year earlier. For a B2B product, the more useful question is where your customers do their work. If they spend their day in ChatGPT or Claude, some of your product's value will be delivered there, through an interface you only partly control.

What we'd do now: make sure your main actions are reachable without your own screens, through an API or an MCP server. And treat accessibility as agent readiness, because an agent reads your page much like a screen reader does. Clean semantic markup and clear labels help both.

AI features have to survive real accounts

The 2026 data on AI agents is a strange mix of adoption and disappointment. Zapier's survey of enterprise leaders found 71% have agents in production, yet only 30% say those agents complete more than 60% of their tasks without a human stepping in. Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027.

Forrester's 2027 predictions for software development go further and expect at least one major failure from AI-written code that shipped without proper review. Their phrase is "at least one big kitchen will go up in flames."

The AI features that keep their users in 2027 will be the ones that survive a Tuesday afternoon in a real customer account. That's a design problem as much as a model problem: how people check the output, approve it, fix it and undo it. The Bruukki case study shows how we approached it for one AI product.

Screens get cheap

Gartner's strategic predictions for 2027 and beyond include this one: "By 2029, 80% of new applications will be intentionally disposable, used for less than one year." McKinsey's 2026 State of AI found that 32% of organizations had already skipped buying at least one software product because they could build it in-house with agentic coding tools.

Inside product teams the lines are blurring too. In Figma's 2026 AI report, the share of designers taking part in development doubled to 41%, and developers doing design work rose from 44% to 60%.

I think this is mostly good. A designer who can put a working prototype in front of a customer on Monday learns more than one who presents a deck on Friday. But when anyone can generate an interface in an afternoon, a lot of products start to look alike. The same component library, down to the gradient on the empty state. Your customers have seen it already, often that same week.

What gets scarcer is judgment about what to leave out, and what should feel specific to your customers.

Senior designers get more valuable, and the path in gets harder

Figma's hiring study from February is the clearest picture we've seen. 82% of design leaders said their need for designers was steady or rising. 56% saw rising demand for senior designers, against 25% for junior ones. And 79% wanted designers who know how to design AI products specifically.

This one worries me. If AI tools do the work juniors used to learn on, and companies only hire people who already know how, the profession has a pipeline problem in five years. I don't have a neat answer. The teams we admire pair juniors with seniors on real work and let them make mistakes in production, which is slower and costs more, and it's still the only way we know to grow good designers. (We wrote a letter to those juniors earlier this year, called Dear designer, don't code.)

What we'd do this quarter

If you lead a product team, five things are worth an hour each before January:

  • List every AI feature in your product and check whether people can tell when they're dealing with AI. Since August, in the EU, they have to be able to.
  • Try exporting all of a customer's data the way a customer would. EU switching duties already apply, and from January 12 you can't charge for them.
  • Pick your most-used workflow and ask whether an agent could complete it, and whether you'd want it to without a confirmation step.
  • Find the AI feature with the biggest gap between how it demos and how often people use it, and watch three customers try it.
  • Look at your last three new screens next to a competitor's. If you couldn't tell whose is whose, start there.

What we're less sure about

Agent traffic is growing fast from a small base, and nobody has reliable data yet on how much buying agents actually do. Agentic browsers have had a bumpy year. Some of what's announced in 2027 will quietly disappear by 2028, the way a fair share of 2025's agent launches already have.

And regulation moves. The EU has already pushed the high-risk deadline back once. Check the dates again in the spring.

Frequently asked questions

What will change in product design in 2027?

Expect AI agents to use products on people's behalf, interfaces to get cheaper to build and harder to tell apart, and more regulation to show up in the UI, such as AI disclosures, data export and human oversight for high-risk AI features.

When do the EU AI Act transparency rules apply?

The Article 50 transparency obligations have applied since August 2, 2026. The grace period for marking AI-generated content in systems already on the market ends on December 2, 2026. High-risk obligations were moved to December 2, 2027.

Will AI replace product designers by 2027?

Current hiring data points the other way for experienced designers. In Figma's 2026 hiring study, 82% of design leaders said their need for designers was steady or rising, with much stronger demand for senior designers than junior ones.


If any of this is already on your roadmap and you'd like senior help with it, book a 30-minute intro call. We'll tell you what we'd tackle first, and whether you need us for it.

Keep reading