Is the time of blind AI-tool-pushing over? There seems to be a shift in how companies approach AI. Less piecemeal initiatives, more strategic thinking and adoption, which is increasingly affecting how design, engineering and product work together.

My summer of 2025 wasn’t all holidays and chill. I spent it deep in my master’s thesis, crossing interviews with designers and developers of all backgrounds and seniorities, building graphs and pivot tables from my own survey, and reading Figma’s AI report along with a stack of articles to get the full picture. Intense, but genuinely one of the most exciting subjects I’ve dug into, AI’s impact on design-dev collaboration.

My conclusion, in short, was that companies knew AI mattered strategically, and knew employees were already using it, but almost none of them were creating real strategies for its adoption, only piecemeal ones. In some cases, AI was prohibited outright, over confidentiality or security concerns, or even, sometimes, just “cause.”

Fast forward to this May, and three days at UXDX EMEA in Berlin gave me the itch to revisit that finding. Has anything actually changed at company level, or are we still mostly testing in the dark?

AI is becoming more than a tool to push

Let’s start with the timeline. If I had to summarise the shift between my 2025 thesis and today, I’d borrow Gartner’s own terminology: we seem to have moved past the “Peak of Inflated Expectations” and into the “Trough of Disillusionment.”

In 2025, companies handed out piecemeal tools largely because of the hype, a mix of FOMO and a lack of perspective. In 2026, confronted with ROI realities, security risks and workflow friction, many are being forced to grow up.

We are moving from chaotic, individual adoption toward more structured, collective strategies. For design, product and tech teams, this generally means less shadow AI and isolated testing, and more collective thinking and shared governance, even if it also means real upheaval in day to day practices.

But no average tells the whole story. At UXDX, I heard both halves of it: teams for whom nothing has really changed since my thesis, and teams for whom almost everything has.

No AI strategy doesn’t mean no results

At N26, Marcus Knight’s story shows what individual initiative can achieve even without a company-wide framework. He told me plainly: “the only strategy we had from a company level was ‘everyone needs to use AI’. I don’t care how you use it, but you need to use it.”

This approach led him to a testing mindset and an experiment he presented to the Berlin public: a Claude Code skill nicknamed Sheldon, modeled on the questions his senior engineer used to raise at every handoff, that checks a design file’s readiness before an engineer even opens it. It is, in his words, an attempt to “clone” the colleague whose pushback he once found so frustrating (most designers will relate), and it has already saved time and moved blockers for both sides, proof that even unstrategized experimentation can smooth out a relationship, not just speed up a task.

Looking back, though, he wishes the company had gone further than the push itself: “It’s not enough to just say, “just make sure you’re using AI.” But what does that actually mean from a strategic point of view?”

He’s not alone in that gap. The AI in Design 2026 report, by Designer Fund and Foundation Capital, puts a number on it: 73% of designers feel rising pressure around output, quality and speed, but only 28% of the leaders they surveyed say their company has made any formal update to match it.

Global company shift, or strategy taking shape

Not every company fits that mold, though. Figma’s 2026 report sorts companies into four adoption profiles based on intent, and one stands in sharp contrast to field companies: “Directive Companies” (27%), where AI use is dictated top-down. Their results appear to support my thesis’s conclusion: structured adoption seems to create better conditions for AI to become a collaboration opportunity rather than a burden, rather than proving that outright.

At these companies, AI’s impact on productivity has more than doubled, its impact on tools has nearly doubled, and its impact on collaboration has nearly tripled, the three largest gains observed in any adoption group. The trade off is that it took these companies longer to see AI’s impact, since teams were left to implement top-down strategies on their own. But employees are quickly finding their footing, even as their companies keep pushing them to go faster and further.

The numbers back this up beyond Figma alone. Designer Fund’s AI report found that 87% of designers now report at least moderate organisational support for AI adoption, and more tellingly, the gap is closing: in 2025, early-stage startups were twice as likely as larger companies to have adopted AI tools, but today 60% of both early and growth-stage companies report strong organisational support, evidence that structured support is no longer the privilege of the fastest movers. Perhaps the clearest signal is this: the share of companies making no AI investment at all dropped from 15% to just 4% between 2024 and 2026.

One of the clearest examples of that shift is Fin (formerly Intercom, later acquired by Salesforce).

Back in August 2022, Intercom was substantially underperforming the market average. After AI made it possible to clone almost any SaaS product in a day or two, the company chose a complete pivot: strategy, organization, even its name, rebuilt around AI.

At UXDX, co-founder Des Traynor put it bluntly: “the only thing that changed is absolutely everything.”

Today, the company outperforms it by far: $400 million in revenue, $100 million of which comes from Fin, their AI agent, alone. The shift took about two and a half years, and Traynor is candid about the fact that it was neither easy nor straightforward (some customers, some team members, even some shareholders resisted along the way). His summary of the process: “you have to go too far to know that you’ve gone far enough.”

Another example of an AI adoption mindset shift that I encountered online in the French media Le Ticket, even if way less radical: the analytics company Amplitude recently organized an AI Week. They closed one week for all 700 employees to build with AI, test it, and understand what they could and shouldn’t expect or do with it.

The results speak for themselves: today, 97.5% of employees use Claude in their daily work, and more than 550 out of roughly 700 employees use Lovable. Beyond the tools, something cultural shifted too. One employee summed it up well: “we’re not trying to hide that we’re working with AI anymore, it’s a lot more openly assumed now” (translated from French).

It’s also a case study in something Designer Fund’s research quantifies: companies with strong organisational support for AI build a culture of tinkering, where everyone is expected to experiment, twice as often as companies with weak support, and designers inside that culture are twice as likely to feel more creative and capable. Amplitude’s AI Week reads like that mechanism in practice.

Role boundaries and collaboration are changing: good or bad news?

Another detail that caught my attention during Des Traynor’s presentation is that at Fin, every designer has to ship something to production. A live example of a team leaning into Design engineering, a subject I explored in my previous article.

And it’s not limited to design and engineering skills, it also extends to Product. Fifty-six percent of non-designers are already doing some form of design work, according to Figma, and the AI in Design 2026 report found that 65% of designers say they’re taking on more product or engineering responsibilities themselves.

It echoes something Rory Madden, UXDX co-creator, also predicts: as the lines between roles continue to blur, team sizes themselves may shrink, from the usual 8 to 10 specialists down to 2 or 3 multi-skilled people, “any extra person on a team adds more overhead and adds more coordination effort”, he told me.

So, is it good news or bad news? Both, depending on what’s around it. A third of the AI in Design report respondents describe roles and ownership as messier than before, and the share reporting decreased collaboration has quadrupled in a year, from 5% to 20%. It could be explained by a shift toward more solo work, more time spent prompting or in terminals, less live back-and-forth with teammates. Some of the connective tissue of collaboration, in their words, hasn’t kept pace.

But for the rest, the same blur is producing the opposite: the number of product builders reporting a significant, positive impact of AI on how their team works together has grown 6x in two years, according to Figma.

Pamela Mead, SVP of Global Design at SumUp, was speaking on a UXDX panel about who should feel most nervous about AI between Product Managers, Designers and Engineers. The panel’s answer, in the end, was everyone. But Mead underlined that design is taking the hardest hit right now, and needs to get back to its strategic roots to bring that back to the table. One of her phrases summed up what that takes:

“You need to have the knowledge, you need to have the experience, and you need to have the trust of each other.”

Those three pillars don’t show up on their own, they’re deliberately and relentlessly built, and are the cornerstones of what separates the two outcomes in that split.

If, according to Figma, designers who lean into AI are 25% more likely to say they’re happier at work, a wider 2026 survey by Lenny’s Newsletter paints a rougher picture for the profession as a whole. There, designers are reporting the most AI-related anxiety and the worst-rated managers of any tech role.

These findings on the impact on collaboration are a sharp contrast with where I left things a year ago: back then, 77% of the designers I personally surveyed reported AI made them individually faster, but almost none of them reported any change in how they actually worked with developers. The increase in productivity was real, but the impact on collaboration was merely a whisper. One year later, it is clearly felt, for better or worse depending on the team.

From awareness to construction, a “work in progress”

So, one year later, where are we? Well, it depends who you ask.

Some companies, like Fin, moved fast toward a fully defined strategy. Others are still earlier in that journey, working through what a company-wide approach should even look like. Neither pace is wrong, they reflect different constraints, different industries, different starting points. What’s changed since my thesis is that the conversation has shifted from awareness to construction, from “everyone should use it” toward what a real strategy, and real governance, actually requires.

Whether that holds in another year, I genuinely don’t know, and neither does anyone I spoke to, which I found more reassuring than any confident prediction would have been. What I do know is that the phase we’re in now is exactly the phase where the real decisions get made, quietly, company by company, team by team. And that we’ll need to trust each other, more than ever.

Thanks to UXDX for providing access to the event, and to Marcus Knight and Rory Madden for sharing their perspectives.

AI in Product teams: In 2026, the growing impact on collaboration was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.