Teneo.ai Was Doing Agentic AI Before It Had a Name

In 2001, three engineers in Stockholm decided that software should be able to hold a conversation. This was four years before YouTube, six before the iPhone, and two decades before a chatbot called ChatGPT made the idea feel obvious. They called the company Artificial Solutions. Today it goes by the name of its product, Teneo, and the thing it set out to build - a machine that understands what a person actually means and does something useful about it - is the most contested territory in enterprise software.

Teneo.ai is not a household name. It has never raised a headline-grabbing venture round, never printed a founder on a magazine cover, never had a launch day that trended. What it has instead is a long paper trail: a chatbot that beat human judges in a Turing-style contest in 2008, a proprietary language for modeling dialogue, and a customer list that reads like a directory of European telecoms and Fortune 500 operations. It powers, by its own count, more than 17,000 AI agents in production.

That gap - between what a company does and what people know it for - is the whole story here. Being early is not the same as being safe.

The pioneer's problem

The founders - Johan Ahlund, Johan Gustavsson and Michael Soderstrom - started with interactive web assistants: the little help agents that tried, often badly, to answer questions on a corporate site. The technology was crude by current standards, but the ambition was not. By 2008 the company's Elbot chatbot won the Loebner Prize, an annual competition that asked whether a program could pass for human in conversation. Elbot slipped past a share of the judges, partly on the strength of a dry sense of humor.

Two years later the company reorganized around a platform it called Teneo, built to let humans and software applications talk to each other in natural language. The bet was simple and, for a long time, unfashionable: that real conversational AI needed structure - explicit models of intent, entities and dialogue - not just a large pile of text and a hope that patterns would emerge.

Then large language models arrived and, for a moment, seemed to make all of that structure look quaint. Why hand-model dialogue when a model could generate any sentence you asked for? The answer, which Teneo.ai had been arguing for years, turned out to matter a lot in enterprise settings: a bank cannot ship a bot that confidently invents an account balance, and a hospital line cannot afford a plausible wrong answer. Fluency is not the same as reliability.

How the machine actually talks

The current Teneo platform runs on a hybrid design. It pairs the company's Teneo Linguistic Modeling Language - a proprietary way of describing how a conversation should flow - with large language models from providers such as OpenAI and Anthropic. The linguistic layer handles understanding and control; the LLM handles the flexible, human-sounding parts. Teneo says the combination holds around 99% accuracy on intent and entity detection while validating outputs before they reach a customer, a property it markets as 100% output control.

There is a cost angle too. Routing every request through a large model is expensive at contact-center volume. By using linguistics to decide when the model is actually needed, Teneo reports cutting one enterprise's LLM spend by roughly 98%. The platform runs the same conversation engine across voice, chat and apps, in more than 86 languages.

None of this makes for a flashy demo. It makes for something enterprises quietly value more: a call that gets handled, a refund that clears, a customer who never realizes a machine was on the line.

The customers nobody sees

The clearest way to understand Teneo.ai is to look at where it runs. Telefonica Germany uses it to handle more than 900,000 calls a month. Swisscom routes roughly nine million calls a year through it and reports an 18-point lift in net promoter score as customers get served faster. Medtronic, the medical-device maker, says it saved about $6 million and freed 36,000 agent hours. Other names on the list include AT&T, Circle K, HelloFresh and the Swedish insurer Folksam.

This is the unglamorous middle of the AI economy. It is also where a lot of the real money is. Front-line customer service is expensive, repetitive and, for most companies, a cost center they would happily shrink. A system that removes even a slice of those calls without making customers angrier pays for itself quickly. Medtronic's freed 36,000 hours is not a marketing metric so much as arithmetic.

The money question

Here the story turns. Teneo.ai trades on Nasdaq First North in Stockholm under the ticker TENEO, having gone public in 2019 through a reverse takeover and rebranded from Artificial Solutions in 2024. Being a small public company in a capital-hungry field is difficult, and 2026 has been rough. A legal dispute with a former US reseller partner hurt revenue during the second quarter, and by mid-year the company disclosed that its cash covered only about a month of operations at then-current levels. In late July it reached an agreement with its lender, Capital Four, to restructure roughly 290 million Swedish kronor in senior financing and signaled a rights issue aimed at getting the balance sheet clean.

Contrast that with the other pole of this market. Uniphore, a US-based Business AI company founded in 2008 and now based in Palo Alto, closed a $260 million Series F in late 2025 at a $2.5 billion valuation, with names like Nvidia, AMD, Snowflake and Databricks on the cap table. It has spent the past few years buying its way across the landscape - Jacada, Red Box, ActionIQ, Infoworks, Orby AI - and building what it calls a Business AI Cloud. Two companies, aiming at roughly the same enterprise buyer, arriving from opposite directions: one with 25 years of dialogue engineering and a strained balance sheet, the other with fresh capital and an acquisition budget.

To be clear, there is no publicly confirmed deal between them. Uniphore has not announced an acquisition of Teneo.ai, and Teneo's strategic review is its own process. But the shape of the market makes the question obvious. When a field consolidates, the well-funded consolidators tend to shop among the pioneers - the companies with real technology, real customers and, sometimes, real cash pressure. Whether Teneo raises fresh money, finds a partner or gets absorbed, its engineering keeps running behind millions of calls either way.

What it leaves you with

There is a lesson in Teneo.ai for anyone building in a hot field. Inventing a category buys you technology and credibility. It does not buy you the narrative, the capital or the timing, and those three things often decide who ends up owning the market you opened. The company that spent two decades insisting conversation needed structure was proven right at exactly the moment a wave of better-funded entrants arrived to act on that insight.

Watch what happens next. If you want a clean read on where enterprise AI is actually heading - past the demos and the valuations - the boring back office of the contact center is a better place to look than any launch event. Teneo.ai has been there the whole time. The rest of the industry is only now pulling up a chair.