Retell AI and boost.ai Are Rewiring the Customer Call

Call almost any large company today and something has quietly changed on the other end of the line. The hold music is the same. The menu is the same. But the thing that answers is increasingly not a person, and increasingly, you cannot tell. That shift did not come from one lab or one launch. It came from two companies working the same problem from opposite ends, and neither one looks like the other.

On one side is boost.ai, a Norwegian company that has spent nearly a decade making text-based virtual agents reliable enough for banks and telcos to trust. On the other is Retell AI, a Y Combinator startup that decided conversational AI belonged on the phone, and built voice agents fast enough to hold a real conversation. Put them side by side and you get the clearest map available of where automated customer conversation is actually going.

Two roads out of the chatbot era

For most people, "chatbot" is still a slightly embarrassing word. It calls up the little pop-up in the corner of a website that misunderstood your question three times before offering a phone number. boost.ai's entire history is an argument that the embarrassment was a symptom of early products, not the idea itself.

The company started in 2016 as a small group of Norwegian coders trying to automate support for a single local bank. That constraint mattered. A bank does not tolerate a virtual agent that guesses. So the team built for accuracy first, and the platform that grew out of it is now used by companies like Telenor, DNB and Silvercar by Audi, with virtual agents that handle a large number of separate intents while keeping resolution rates above 90 percent. In 2025 boost.ai was named a Leader in the Gartner Magic Quadrant for Conversational AI Platforms, and more than 500 organizations run on it.

Retell AI took the same underlying technology, large language models, and pointed it somewhere boost.ai mostly did not: the telephone. Founded in 2023 and shaped inside a Y Combinator batch, its founders came out of ByteDance, Google and Meta and made an unusual decision for the moment. Instead of building a general assistant that could do everything a little, they built voice agents that do one thing - answer and place phone calls - and do it well enough to survive contact with real callers.

Why 600 milliseconds changed everything

The reason voice AI felt like a party trick for so long comes down to a single number: latency, the gap between you finishing a sentence and the machine responding. In natural human conversation, that gap is a few hundred milliseconds. Older systems took long enough that you could feel the delay, and once you feel it, the illusion breaks and you start talking to it like a machine.

Retell AI targets roughly 600 milliseconds. That is close enough to human turn-taking that callers stop performing for the robot and just talk. Underneath the speed sits the unglamorous engineering that actually makes a call work: real-time function calling so the agent can look something up mid-sentence, streaming retrieval from a knowledge base, and a framework for designing call flows that handle the fortieth edge case, not just the demo-friendly first one.

The number nobody in venture can stop repeating

Here is the figure that made investors sit up. By reported numbers, Retell AI climbed from around $5M in annual recurring revenue at the start of 2025 to roughly $50M by the end of it, and toward $60M annualized in 2026, up about 650 percent year over year. It did this on roughly $5M raised, with a team of about 30 people.

Work the ratio and you get close to $10 of ARR for every $1 raised. In a market where the default advice is to raise more and hire faster, that is a counterexample sharp enough to hurt. It is also a reminder of a pattern boost.ai proved years earlier in text: the durable advantage in conversational AI is not the model, which everyone can rent. It is the messy last mile - the interruptions, the corrections, the caller who changes their mind halfway through - that most teams never bother to finish.

Same market, opposite instincts

It would be easy to frame this as incumbent versus insurgent, and the funding stories certainly line up that way. boost.ai sold a stake to Nordic Capital in 2021, the patient private-equity route to becoming a category leader. Retell AI took the classic Silicon Valley path through Y Combinator and a lean seed round. Different continents, different money, different tempo.

But the more useful read is that they are attacking the same job from two honest directions. boost.ai owns the case where a problem should be solved before anyone has to speak - a balance check, a password reset, a status lookup handled in text, instantly, at any hour. Retell AI owns the case where the phone is still the channel, where a caller wants to talk, and the win is making that conversation feel natural rather than automated. They even show up in each other's partner ecosystems through integrations, which makes them as much complements as rivals.

What you can take from this

Strip away the funding drama and both companies teach the same lesson to anyone building. Pick a job nobody wants to do the boring parts of, and do those parts obsessively. Retell AI refused to become a general chatbot and only built for the phone. boost.ai refused to ship an agent that guesses and only shipped ones that resolve. Neither tried to do everything, and that restraint is exactly why both worked.

There is a founder detail that fits the pattern. boost.ai's founder and CEO, Lars Selsas, once moved to share a large slice of the company's equity with employees - the kind of long-game decision that reads as strange in a hype cycle and obvious a decade later. It rhymes with Retell AI's choice to stay small and efficient while revenue ran ahead of headcount. Different cultures, same conviction that the fundamentals win.

For everyone else - the people who call companies, wait on hold, and type into little chat windows - the change is arriving whether or not it gets announced. The receptionist changed. The support agent changed. And two companies most people have never heard of, on two different continents, are the reason.

Questions people ask