As a product studio, we’ve spent a lot of time thinking about one big question: how do you cut through the noise around AI and use it to build a product customers love?

Because the pressure is real. Everyone’s still talking about it. Boards want a strategy. Customers expect smarter experiences. Teams are experimenting. But turning all that into something real, a product that ships, that’s safe, that solves a problem, is still incredibly hard.

So we thought we’d share a conversation from earlier this year, when we sat down with OpenAI, Experian UK&I, Octopus, and Lewis Silkin.

If you're figuring out where to start, what it takes to scale, or how to bring the rest of the business with you, here are their learnings from getting AI into production.

Between agents, copilots, and assistants, it’s tempting to chase whatever’s trending. But the teams successfully shipping AI products are the ones starting from customer problems and focusing on what already makes their experience great.

As Matt Weaver from OpenAI puts it: “The companies doing best are the ones ignoring the hype and focusing on what makes their business different. Then doubling down on that.”

He pointed to Virgin Atlantic as an example: “They used AI to amplify their competitive edge: human experience. They’re building a GenAI-powered travel concierge that shares tips from actual flight crews. It’s a way to make their human touch even stronger."

Alongside doubling down on your differentiator, it also has to solve a real friction point for customers.

Inês Liberato from Lewis Silkin shared how her team approached it: “The ask wasn’t ‘go build me a GenAI product.’ The ask was, ‘how can we make this experience better?’ and the answer just happened to be AI.”

AI has dramatically increased the speed of learning. Teams are moving from idea to working prototype in days, not months.

As Inês put it: “The discovery process has become a lot more tangible. Sometimes you get stuck in planning, mapping risks, debating feasibility, but you never actually test whether something works. AI lets us move faster and get earlier feedback from users.

We built an AI prototype and put it in front of customers. The response was, ‘this is a game changer.’ Showing value that early gave us the buy-in to move it into production.”

AI can be unpredictable, and that’s scary. GPT-3.5’s habit of hallucinating is burned into people’s memories, especially in high-stakes industries. 

That’s why testing isn’t an afterthought anymore. It’s the work that makes sure a product behaves and scales the way you expect and need it to.

Matt explained: “We talk a lot about evals. They’re like unit tests for AI. We advise teams to write them before they write any code, because with AI, tweaking a prompt can make your product much better or much worse.

Without a solid test set, you have no way of knowing whether your changes are actually helping or quietly breaking something else. Evals are what give you the confidence your product will hold up as you continue to scale.”

Building fast is no longer a competitive advantage; it’s the baseline.

Sam Grice from Octopus Legacy put it plainly: “If it takes you 12 months to build something that competitors can do in two weeks, you’re done.”

But speed alone isn’t enough. If the rest of the organisation can’t keep up, even the best ideas get stuck.

Inês explained: “The sequencing is so much faster. With AI, my usual playbook got thrown out. You have to rope in a lot more stakeholders, much earlier. And you need some urgency: a launch deadline, customer traction, something that keeps momentum going. Otherwise, the rest of the org can end up stalling you."

Christine Foster from Experian added: “Compliance and legal, they’re not business prevention. They’re business partners. They have to be there from day one.”

The teams moving fastest have built a culture of experimentation, literacy, and alignment across the company.

Matt explained: “You need to pull two levers at once. One is AI literacy. Everyone in the company should be experimenting, tinkering, trying tools. It’s a foundational requirement now.

The second is leadership vision. It’s easy to end up with a spreadsheet of 100 AI use cases, but what actually moves people is a strategy. You need a clear, ambitious direction. Something like, ‘what would a fully automated AI banker look like three years from now?’”

Inês echoed this: “It’s not just about what tech we can introduce. It’s about asking, what’s the business we want to become? That’s where you get buy-in.”

We’re running free, hands-on workshops showing how product teams are using AI to go from idea to prototype in record time.

You’ll explore the tools, learn the approach, and even try building a prototype yourself. Plus, we’ll cover what to watch out for along the way.

Want to set up a session for your team? Email us at hey@planes.agency