BREAKINGKiwi founder lands Silicon Valley money over a poker game GRW AIbuilds Taylor, the AI sales teammate CAMBRIDGEthesis taught machines to talk to themselves XEROshipped product to a million customers $15,000left in the bank, slept in a basement, raised anyway BREAKINGKiwi founder lands Silicon Valley money over a poker game GRW AIbuilds Taylor, the AI sales teammate CAMBRIDGEthesis taught machines to talk to themselves XEROshipped product to a million customers $15,000left in the bank, slept in a basement, raised anyway
Person · Founder · Cambridge ML

Alistair McLeay

He taught language models to argue with themselves for a master's thesis. Now he is teaching them to coach salespeople. Co-Founder and Co-CEO of Grw AI.

Grw AITaylorLLM AgentsWellington → SF
Portrait of Alistair McLeay
ALISTAIR McLEAY, photographed for the record. The background on his own website was drawn by an AI - he just runs the company that builds them.

A salesperson's best coach now boots up in a browser tab.

There is a piece of software in San Francisco named Taylor. It does not carry a quota, does not sandbag the forecast, and does not flinch when a deal goes quiet. Taylor reads the room of an entire sales team, notices the operational work falling through the cracks, and quietly does it. Alistair McLeay is one of the people who built it - and he is in no hurry to make Taylor sound modest.

Grw AI, the company he co-founded and co-leads, has a single stubborn ambition: a superhuman sales leader. Not a chatbot bolted onto a CRM. An agent that coaches, trains, and lifts a whole revenue org the way the best manager you ever had would, except it never sleeps, never plays favourites, and never forgets what you said on last Tuesday's call.

That is the pitch. The proof is in the wiring, and the wiring is where McLeay is most at home. He did not arrive at AI sales coaching through a sales career. He arrived through a Cambridge laboratory, a product job touching a million strangers' accounts, and a card table where the stakes were the whole company.

We're pushing the limits of what's possible with agents - building a superhuman sales leader. — Alistair McLeay, on Grw AI's reason to exist
1M
Xero customers shipped to
2
Companies built before Grw
~$2.2M
Reported seed raised
$15K
In the bank when it counted

Down to fifteen grand, sleeping in a basement, and dealt into the most important poker hand of his life.

The founder mythology of Silicon Valley usually involves a garage. McLeay's involves a basement and a deck of cards. The team flew into San Francisco with roughly fifteen thousand dollars left - the kind of runway measured in weeks, not quarters. They crashed in a basement. And somewhere in that scramble, McLeay ended up across a poker table from the investor who would change everything.

He raised the round. He likes to say it was a bet that sparked the whole thing, which is why the podcast that told the story called him "the all-in founder." The phrase fits a man who left a stable product career to teach neural networks to negotiate.

It is a good story because it is a true one, and because it tells you something about how McLeay operates. He does not hedge. He sizes the pot, reads the table, and pushes the stack forward.

A broke Kiwi founder, $15,000, a basement, and the poker game that funded a superhuman AI. — The story, as told on the Startup Theatre Podcast
▶ Watch / hear "The All-In Founder" on Startup Theatre

A million users, then a deliberate detour into the deep end of machine learning.

Before any of the agent-building, McLeay was a Technical Product Manager at Xero, the accounting software giant, where he owned the Identity Platform. That is the unglamorous plumbing that decides whether a million people can log in and trust that their financial data is theirs alone. Shipping to a million customers teaches a particular humility: software at scale is a promise, and broken promises compound.

Then he did something most product managers only daydream about. He left to go back to school - not for an MBA, but for a master's in Machine Learning & Machine Intelligence at the University of Cambridge. His thesis trained LLM agents through self-play, the technique where a model improves by playing against versions of itself. It is the same family of ideas that taught machines to beat humans at Go. McLeay pointed it at dialogue.

If you want to know why Grw AI talks about "agents" rather than "a chatbot," this is the origin. McLeay did not learn AI from a blog post during the hype cycle. He built the agents by hand, watched them learn from their own conversations, and came away convinced that a sufficiently good agent could do more than answer - it could coach.

Between Cambridge and Grw AI he built two software companies, one in New Zealand and one in the United States. In interviews he has talked through ventures named Gondola and Voyce - the unglamorous, instructive kind of building where you learn what customers will and will not pay for. Two companies is two sets of scar tissue. It is also two reasons investors took the third pitch seriously.

He shipped product to a million Xero customers, then left to train large language models at Cambridge. — The one-line résumé he can't shorten any further

What Taylor actually does when nobody's watching.

Grw AI's thesis is simple and a little uncomfortable: as companies grow fast, sales teams get stretched thin and the critical work quietly slips. Taylor is built to step into the workflow, understand the team's dynamics on its own, and execute - no hand-holding required.

Coaches

Personalised, hyper-contextual guidance for every rep - like the world's best manager, cloned and always on call.

Manages the pipeline

Watches deals, spots the ones going cold, and keeps the operational work from falling through the cracks.

Does the work

Steps into the workflow and executes the necessary tasks, instead of waiting to be asked.

The values Grw AI puts on the wall.

Every startup has a values page. Most read like wallpaper. Grw AI's are unusually pointed - and they map neatly onto how McLeay tells his own story: high effort, total ownership, big bets, and a stubborn insistence that the AI is there to lift people, not replace them.

Exceptionalrelentless
Unusually Responsibletotal ownership
Bold Betspush the stack
People-CenteredAI for humans

Things that don't fit in a pitch deck.

Meta

His personal website was designed and built by Claude Opus 3 in mid-2024. The neural-network background? The model's own idea. The founder of an AI company let an AI do his homepage.

Off-piste

When he is not training agents, he is chasing adrenaline - surfing, mountain climbing, and the kind of travel that does not come with a hotel concierge.

The flex

He raised Silicon Valley money before most New Zealand founders get so much as a first meeting. Geography was not the moat he was told it would be.

Self-play

His Cambridge research had language models improve by talking to themselves. The whole company is, in a sense, that idea grown up: agents that get better at helping people sell.

Where this is all pointed.

The goal is not a better autocomplete for sales emails. It is a superhuman sales leader - an agent that lets a company scale further and faster than was ever possible before, while keeping people at the centre of the work. McLeay is betting his career that the best manager you ever had can be rebuilt, improved, and handed to every team at once.

Building a superhuman AI sales leader. — Grw AI, in five words

Follow the work.

Reporting compiled from public sources: alistairmcleay.com, grw.ai, Crunchbase, and the Startup Theatre Podcast. Funding figures are as reported and may vary across sources. Where the record was thin, we left it thin.