The Selling Machine That Never Clocks Out
In April 2026, Mihir Garimella's company Actively AI raised a $45 million Series B, co-led by TCV and First Harmonic, with Bain Capital Ventures, First Round Capital and Alkeon following on. The headline number is the easy part. The harder claim sits underneath it: that the human-led sales model is running out of road.
Actively does not bolt AI features onto a CRM. It assigns a persistent agent to each account. That agent watches activity across a company's systems, reads the signals, forms a hypothesis about what to do next, and moves the deal forward - at 3 a.m. on a Sunday if that is when the signal fires. Garimella has a name for the category he is trying to will into existence: GTM superintelligence. Go-to-market, run by reasoning rather than headcount.
The customers are not pilots-for-show. Ramp, Ironclad and Attentive use it. Samsara deployed it across a go-to-market team of more than a thousand people. When Garimella says execution should not depend on human capacity alone, he is describing what those teams already see on a dashboard.
The framing matters because the market is crowded with tools that promise to make a rep marginally faster - a better draft email, a tidier forecast, a nudge in the right Slack channel. Actively's argument is structurally different. A rep can only work so many accounts in a day; the long tail of accounts that never get a human's full attention is exactly where pipeline quietly leaks. Hand each of those accounts a tireless agent that reasons about its specific situation, and the constraint stops being how many hours the team can work. Garimella describes the company's category as intelligence-led revenue: continuous attention to every account, regardless of who is awake.
We're building AI that helps the most ambitious companies in the world generate more revenue and grow even faster. We call it GTM superintelligence.
Mihir Garimella, launching Actively AIIt is a confident bet, and Garimella has been making confident technical bets since middle school. The pattern is always the same: find a system that should be impossible at its price, and build it anyway.
A Kitchen Full of Fruit Flies
The family came home from a long vacation. Someone had left bananas on the counter. The house, predictably, was a cloud of fruit flies. Most people reach for a swatter and a trash bag. Garimella, then a teenager, kept missing - and started wondering why a creature with a brain the size of a poppy seed and famously bad eyesight was beating him every single time.
The answer turned out to be speed of perception. Fruit flies process visual information roughly ten times faster than humans, which is how they read a swat coming and pivot out of the way before the hand arrives. Garimella took that one fact and turned it into hardware.
The result was Firefly (early on, Flybot): a palm-sized quadrotor that detects and dodges moving obstacles - think falling ceilings in a collapsing building - the way a fruit fly dodges a hand. The pitch was not "cool drone." It was a roughly $250 flying robot that could fly into a burning building or an earthquake-damaged structure, avoid debris on its own, and use sensors to locate the source of a fire or a trapped person, relaying it all back to first responders in real time.
Commercial rescue drones cost orders of magnitude more. That price gap was the whole point.
The work won his age category and the Google Computer Science Award at the 2015 Google Science Fair. He was an Intel ISEF finalist and a multi-year Broadcom MASTERS finalist. He took the idea to the TEDxTeen stage in New York with a talk called "Designing the next generation of first responders," and CNN Business put him in its Tomorrow's Hero series - all before he had a driver's license.
Curiosity is a great skill for inventors. It lets you find inspiration in a lot of places that you may not look.
Mihir GarimellaThe fruit-fly drone was not the first thing he built, either. At ten he made Robo-Mozart, a robot that listens to a violin and turns the tuning pegs itself - because out-of-tune strings annoyed him and the obvious fix was to automate the annoyance away. The fascination with machines reportedly goes back to age two and a robotic dog his parents brought home.
Stanford, and the Quiet Years
Garimella went to Stanford for both a bachelor's and master's in computer science, writing his thesis on machine learning infrastructure. He co-directed TreeHacks, Stanford's largest hackathon, in 2019. But the more telling part of his twenties was the resume he built between the headlines.
He spent a summer at Waymo training the deep learning networks that pick pedestrians, cyclists, motorcyclists and other vehicles out of raw laser data - the perception layer a self-driving car cannot afford to get wrong. He worked at Facebook on connectivity hardware and software aimed at expanding internet access. And he worked at Helia, a startup using deep learning to analyze video, which was later acquired by Scale AI.
It reads like someone deliberately collecting the components of a hard AI company: perception, infrastructure, applied research, and the unglamorous discipline of shipping models that have to work in the real world. Then he assembled them.
There is a thread that runs through all of it. At Stanford he was also a teaching assistant for the introductory and systems computer science courses, the kind of role you take when you actually like explaining how a machine works to someone who has never seen one. Back in Pittsburgh, as a high schooler, he had founded FCHacks, the city's first high school hackathon. The instinct to build the room as well as the thing - to organize the hackathon, direct the bigger one, teach the class - shows up again later when he sets out to define a whole category rather than ship a single feature.
Teaching Machines to Sell
Garimella co-founded Actively AI with fellow Stanford AI researcher Anshul Gupta. First Round Capital led a $5 million seed. Bain Capital Ventures led a $17.5 million Series A. Then came the $45 million Series B in 2026 and a new San Francisco office to go with the New York headquarters at 25 West 39th Street.
The throughline from the drone is hard to miss. Firefly took a biological system - how a fruit fly perceives and reacts faster than a human - and rebuilt it in cheap hardware to do a job humans physically could not. Actively takes a human system - how the best sales rep reads an account and decides what to do next - and rebuilds it in software to run continuously, across every account, without tiring. Same move, different domain. Reverse-engineer the thing that works, then remove the limit that made it scarce.
The investor roster reads like a vote on that thesis. First Round Capital wrote the first check at the seed; Bain Capital Ventures led the Series A; TCV and First Harmonic co-led the Series B, with Alkeon joining. These are firms that buy into infrastructure-scale ideas, not point solutions, and the plan for the new capital - more product, more enterprise customers, top talent, the San Francisco office - is the plan of a company trying to occupy a category before someone else names it.
Whether GTM superintelligence becomes a category or a footnote is still being written. But the person making the bet has a long habit of building the improbable cheaply and shipping it before anyone asks him to. He has been doing it since a swarm of fruit flies embarrassed him in his own kitchen.
On the TEDxTeen Stage
His talk, "Designing the next generation of first responders," lays out the Firefly story in his own words.
YouTube · TEDxTeenDesigning the next generation of first responders →