How boost.ai Turned Norwegian Banking Chat Into a Global AI Play

Most conversational AI stories start with a demo that dazzles and a product that never ships. boost.ai went the other way. It started with a single, unglamorous job: answer the questions a Norwegian bank's customers kept asking, over and over, in a chat window. A decade later that narrow beginning has turned into a company running more than 600 live AI agents for 450-plus organizations, and sitting in the Leader quadrant of Gartner's 2025 Magic Quadrant for Conversational AI Platforms.

The company was founded in 2016 in Stavanger, Norway - a city better known for North Sea oil than for enterprise software. A small group of coders, led by Lars Ropeid Selsås, wanted to automate the repetitive back-and-forth of customer service. Not the flashy edge cases, the boring middle: password resets, account questions, "where is my payment." The bet was that if you could resolve the questions people actually repeat, you would earn the right to handle the hard ones.

01 / The productWhat you can actually do with it

boost.ai sells a conversational AI platform, which is a crowded phrase. In practice it means a place where a company can build, launch and manage virtual agents across both chat and voice, without needing an engineering team for every change. The builder is no-code, so the people who run support - not just developers - can shape how the agent answers. On top of that sits generative AI, wrapped in what the company calls enterprise guardrails: controls that keep an AI answer inside the lines a bank or insurer can defend.

That last part is the whole game for regulated industries. A retailer can tolerate a chatbot that occasionally improvises. A bank cannot. So boost.ai leans into the unsexy checklist - ISO 27001 and ISO 27701 certifications, auditable answers, testing before deployment - because that checklist is what unlocks the biggest budgets.

In 2025 the company shipped Test Studio, a feature that lets enterprises validate how an AI agent performs before customers ever meet it. It sounds mundane. It is also exactly the kind of thing a compliance officer asks for on the first call. The pattern repeats across boost.ai's roadmap: pick the boring, trust-building feature over the demo-friendly one.

02 / The numbersResolution is the only slide that matters

Vendors in this space love vanity metrics. boost.ai's headline number is a useful one: resolution rate above 90% on deployed agents. A resolution is a conversation the AI finished without punting to a human. That is the figure that maps directly to a cost line, which is why enterprises care about it more than "messages sent" or "users reached."

There is a quieter number worth dwelling on: 4,500-plus certified AI trainers. Those trainers do not work for boost.ai. They work inside customer organizations - the support leads and product owners who tune the agents day to day. By certifying users instead of hoarding the tuning in-house, boost.ai turns its customers into its distribution and its retention. Once a bank has trained 40 of its own people on your platform, switching costs get very real.

03 / The fieldUniphore, Intercom, Five9 - and a company from Stavanger

boost.ai does not compete in a vacuum. The enterprise conversational AI market is loud and well-funded, and the names it runs into differ depending on the deal. Uniphore comes at the problem from analytics and CX intelligence. Intercom pushes its Fin AI agent bundled inside a broader customer-service suite. Five9 owns two decades of contact-center voice and is layering AI on top of telephony. Each has real distribution boost.ai lacks.

boost.ai's answer to bigger rivals is focus. It is not trying to be a full contact-center suite or a horizontal support tool. It goes narrow and deep on regulated sectors - banking, insurance, telecom, public sector - where trust, certification and predictable behavior outweigh a flashy feature list. That is how a company most people cannot place on a map ends up in the same Gartner quadrant as far larger vendors.

04 / The money and the peopleSlow capital, compounding product

In 2021, Nordic Capital invested in boost.ai - notably the first deal for its mid-market-focused Evolution fund. Private equity backing shaped the company's temperament: build durable revenue in defensible markets rather than chase the fastest-growing logo. It is a patient kind of capital, and it fits a product that compounds slowly through trust.

A year later, in December 2022, Jerry Haywood took over as CEO, succeeding founder Lars Ropeid Selsås. Haywood is British and arrived from LivePerson, where he ran the EMEA business, with earlier stops at IBM, Cisco and Salesforce. Bringing in an enterprise-sales operator to succeed a technical founder is a familiar move - and a signal about where boost.ai thought its next chapter would be won: in the room with Fortune 1000 buyers, not just in the codebase.

The customer list backs that up. boost.ai's virtual agents serve customers of Nordea, DNB, Telenor and Credit Union of Colorado, with more recent wins including Jack Henry and Sage. These are not experimental pilots. They are organizations that answer to regulators and cannot afford an AI agent that goes off-script.

05 / The takeawayWhy the boring bet is aging well

The current AI cycle rewards spectacle. Every founder promises magic, every launch is a demo. boost.ai's story argues for the opposite discipline. Start with one narrow customer. Automate the questions people genuinely repeat. Make the product auditable enough that the most cautious buyer in the building can say yes. Then let trust compound.

None of that trends on launch day. But it is holding up. A handful of coders in Stavanger built a platform now handling 150 million conversations a year, and did it far from any traditional tech hub by staying closer to their customers than their competitors did. In a market obsessed with the flashiest answer, boost.ai bet that enterprises would eventually pay for the calm one. So far, that bet is winning.

FAQQuestions people ask about boost.ai