Teneo.ai Wants Your Contact Center to Stop Guessing

Call any big bank, telco or insurer and you already know the drill. You say what you want. The machine hears something else. You repeat it slower, louder, angrier. Somewhere in that gap - between what a caller means and what the software decides they said - is a very old, very expensive problem. Teneo.ai has spent 25 years standing in exactly that gap, and its pitch right now is refreshingly narrow: make the guessing stop.

Teneo is the platform, Teneo AI AB is the company, and until August 2024 the whole thing was called Artificial Solutions. It was founded in Stockholm in 2001 - which means it was building conversational software when "conversational AI" was still a research paper and a novelty demo. That longevity is either a liability or the entire point, depending on who you ask. The company would like you to think it is the point.

The reframed story is agentic AI: software that does not just answer a question but reasons through it, takes an action, and closes the loop. Teneo says it runs more than 17,000 of these AI agents in production, claims 99% intent accuracy, and - the line that gets a compliance officer to lean in - promises 99% output control, which it describes as zero hallucinations. Whether those numbers hold in every deployment is a fair question. That they are the numbers the market now argues about is the more interesting fact.

01The layer, not the replacement

Here is the part that makes Teneo unusual in a crowded field. Most AI vendors want to be the thing you buy instead of what you have. Teneo's whole commercial argument is the opposite: keep your contact center, keep your routing, keep the phone tree your customers grudgingly tolerate - and let Teneo sit on top and do the understanding.

Its Voice AI Accelerator for Genesys Cloud is the clearest example. Instead of asking an enterprise to migrate away from Genesys, it turns the existing IVR trees into conversational experiences through a native connector. The company reports the same approach integrates with Amazon, Microsoft, Google, Five9 and NICE. The names in this story - Genesys, Observe.AI, Amazon, Amelia, Infobip - are usually framed as rivals. Teneo would rather frame most of them as sockets it plugs into.

The strategy is quietly clever. Enterprise buyers have been burned by AI pilots that died in procurement or never survived contact with a real switchboard. "Add a smart layer to what you own" is a much easier internal memo to write than "replace the system 4,000 agents were trained on last year."

02Why accuracy became the whole argument

A few years ago the contact center AI conversation was about demos. Could the bot sound human? Could it crack a joke? That era is closing. The argument now is about a number the industry calls containment - the share of calls fully resolved without a human ever picking up - and about the cost of being wrong.

The math is unforgiving at scale. Teneo cites one enterprise customer saving $32.4 million a month, and points to work like Telefonica Germany handling more than 900,000 monthly calls and Swisscom running millions of interactions a year with a rising satisfaction score. When you are moving that volume, one percentage point of accuracy is not a rounding error. It is thousands of frustrated humans, transferred agents and abandoned calls - every single day.

Output control is the sleeper feature here. Every large model can hallucinate, and in a regulated call - a mortgage, a medical query, an insurance claim - a confidently wrong answer is worse than no answer. Teneo's bet is that enterprises will pay for a system that can be constrained and audited over one that is merely impressive. In the hallucination era, "boring and correct" is turning into a selling point.

03A 25-year head start, or a 25-year problem

Teneo's back story is genuinely odd for an AI company in 2026. Its lineage traces to Elbot, a chatbot that won the 2008 Loebner Prize - an annual Turing-test-style contest for the most convincingly human program. So the company was already chasing the machine-that-talks-like-us goal a decade and a half before ChatGPT made it dinner-table conversation.

That history cuts both ways. The skeptic's version: an old NLU company repainting itself in agentic colors. The believer's version: while newer entrants are discovering that language is hard, that phone lines are messy, and that enterprises do not tolerate a 90% solution, Teneo has been failing and fixing in production for two decades. Under CEO Per Ottosson, who took the job in November 2020, the company rebranded to Teneo.ai in 2024 and now trades publicly in Stockholm. The rebrand was cosmetic. The claim underneath it - that patience in this specific problem compounds - is the real wager.

04What you can actually do with it

Strip away the positioning and Teneo is a toolkit for a specific job: automating high-volume customer conversations without losing control of them. In practice that looks like a handful of concrete moves.

Turn an old IVR into a real conversation. Point Teneo at an existing Genesys or Amazon setup and callers can speak naturally instead of pressing 1, 2, or 3. Automate the repetitive 60%. Balance checks, appointment changes, order status - the calls that do not need a human get resolved by an agent that can actually take the action. Go multilingual without multiplying teams. One instance covers 86+ languages, which matters for any company operating across borders. Keep a paper trail. The output-control layer is aimed squarely at the compliance and risk teams who have veto power over these projects.

The recognition is starting to line up with the story. In DMG Consulting's 2025 Conversational AI Solutions for the Enterprise report, Teneo posted perfect 5.0 scores across categories including implementation, support, pricing and overall vendor satisfaction. Analyst scorecards are not gospel, but for a buyer building a shortlist, they are the difference between "on the list" and "never heard of them."

05The honest caveat

Every figure in this piece comes from Teneo or from analysts writing about it, and the field it plays in is loud with identical-sounding claims. Genesys, Observe.AI, Amazon, Amelia and Infobip all promise some version of smarter, cheaper, more automated service. The useful signal is not the accuracy number in isolation - everyone quotes a good one - but whether a vendor will name the customer, the volume and the containment rate. Teneo tends to. That willingness to publish specifics, more than any single percentage, is what a serious buyer should weigh.

The larger point outlasts Teneo itself. The contact center is quietly becoming one of the first places where agentic AI has to work every day, at scale, with real money and real regulators watching. The winners will not be whoever demos best. They will be whoever can be wrong the least - and prove it.