Product Builder

ABOUT NOUS Nous is an AI agent that takes some of the load of life: making good decisions and acting for people in areas where they're not the experts, or aren't paying enough attention. We started with bills, where we're already saving households thousands, and we're now expanding into new categories. Think Amazon starting with books. It's working. 20x year-on-year growth makes us one of the fastest-scaling startups in the UK, and an NPS of +70 (higher than Apple) says our users love it. On top of bill optimisation, we've just launched two more products: insurance, where we're the first in the UK to get regulatory approval to run the full discovery-and-purchase journey as an agent (not just hand you a link). And subscription management, where we use Open Banking to actively cancel unwanted subscriptions for people, rather than just draw them a pretty graph. We're a certified B Corp, and that's how we hold ourselves to the mission rather than just talking about it. We've stacked the deck: a founding team with multiple exits, and investors including the founders of Monzo, Wise, Booking.com, lastminute.com, Onfido, Funding Circle, Tide and Habito. The problems are hard, and that's the point. Talk is cheap, and that goes for agents too. To have real impact, an agent has to break out of the sandbox and act in the real world, where there are consequences. That means solving all the thorny, frustrating, genuinely interesting problems the real world throws up. HOW WE BUILD (WORTH READING BEFORE THE ROLE) Most companies split the thinkers, the doers and the builders. We don't. Something special happens when they're the same people: a small team that builds and owns the tools for its own problems understands those problems from the inside. So the rule at Nous is simple. The person who feels the problem should be the one who builds the fix. This isn't brand new. A sharp generalist could always wire something together with no-code and low-code tools to solve a problem. The catch was the ceiling: those tools were far more limited than real code, so you hit a wall the moment the problem got interesting. Unless you had superhuman levels of stubbornness and a high tolerance for using tools for things they really weren't designed for. What's changed is that ceiling. With AI tooling and a lot of context, you're now building much closer to real code, and shipping it on the live product. You can change how our WhatsApp agent handles a tricky conversation, dig through Open Banking data to find money we should be saving people, prototype a feature, or automate a process that used to eat someone's week. And you're not doing this around engineering, you're doing it alongside engineers. Engineers build the safe surfaces you ship on (the tools, the infrastructure, the context APIs, the guardrails), and they're who you pull in when you reach the edge of what's safe to touch alone. None of that works without a particular kind of person: not a software engineer, but someone who builds like one. THE ROLE You'll own a product surface and make it better, end to end, without waiting for an engineer to do it for you. The clearest version of this already exists on the team. Ben joined as a PM, not an engineer. Engineering set him up with the right infrastructure and a context API, and now he ships changes to Nora (our customer-facing WhatsApp agent) himself, pulling people in only when he's blocked. Worth being upfront: we've never hired for this profile directly before. The people doing it today grew into it here (Ben from product, Gen from ops) once we set them up with the tooling and context to build. You'd be the first we bring in, so expect the role to shift a bit as we figure out what works. Depending on where you're strongest and where we need you most, the first surface might be: - Nora and our AI agents. Find where she struggles in real conversations, improve how she handles them, build and run the evals that tell us whether she's getting better, and tighten the recovery flows for when things go wrong. - Open Banking and subscriptions. Dig into the data behind our subscription product, spot the patterns, and build the logic that decides what's worth cancelling or flagging. Then work out how we cancel them automatically. Real money back in people's pockets. - Whatever's most broken. New products and surfaces appear constantly. You'll take a rough, messy problem and turn it into something shipped and improving. WHAT YOU'LL ACTUALLY DO - Take a product problem from rough to shipped, with AI tooling (Claude Code, Cursor, whatever gets it done) as your main working environment. - Answer your own questions. When you need to know how many members hit a broken flow last week, write the query and find out. Don't file a ticket and wait. - Ship on the safe surfaces: prompts, agent behaviour, configuration, workflows, data pipelines, internal tools. Know where the edge is, and pull in an engineer before you cr