Our Story.
Getting AI to pay off in a business takes two things: knowing where it gets stuck, and being able to build the fix. We're one of each.
Hamish.
Twenty-five years in media and advertising, ending up as CEO of Dentsu - two thousand people, £1.5 billion of turnover - after a spell as CRO at The Guardian, where he helped turn a £100m cash loss to operational breakeven in three years. Before that, years at Google helping companies through the last big digital shift. He's good at seeing where the growth is hiding in a business and getting people to make the change, which is where most projects die.
The last three years he's been hands-on with AI, building daft useful apps for himself - one that dials in his coffee machine, one that puts his wine-writer wife's expertise in your pocket in the supermarket aisle. He gets the tech well enough to know where to point it at your business.
Stephen.
Twenty-five years an engineer, most of it automation, a lot of that at Cisco. Since we started AgentFlow he's built a platform that turns one consultant's financial analysis into a product they can sell to twenty times the clients. He's built an agent that does a research firm's competitor analysis in a third of the time, and an AI that reads a restaurant's invoices off emails and WhatsApp photos and spots when a supplier's prices creep up.
Find the thing that's slowing a business down and he's the one who turns it into something that just runs.
Between us that's about fifty years of living with business technology and knowing how to make it pay off. We've got the scars and the trophies.
Spend enough time with this stuff and you form a view. Ours comes from a game of Go.
In 2016 Lee Sedol, the best Go player alive, sat down to play Google's AI. Go has more possible positions than there are atoms in the universe, and humans had been the best at it for two and a half thousand years. Then, in the second game, the machine played move 37. No human would have played it. It looked like a beginner's error, a couple of commentators wondered aloud whether the thing had broken, and it turned out to be one of the finest moves anyone had ever seen. It had spotted something twenty-five centuries of human play had missed.
Sedol could have folded after that, and plenty would have. He didn't. Two games later he played move 78, a move so good the machine had nothing for it, and for the only time in the series the AI lost.
We think everyone's got a move 37 coming. Here's what we believe about meeting it.
AI will be better than you at some things. Including things you're proud of. What you do next is the bit that matters.
The real prize is what AI lets you see: the constraints you've designed your whole business around. Taking them out usually means changing how the place runs.
Used well, AI grows a business. Used lazily, it trims a few costs and stops there. We're interested in the first one.
AI gets you to competent quickly, but YOU are what takes your work to excellent.
Never hand it the thinking and walk off. Always drive the conversation.
Make it argue with you, never just agree. And never take its confidence for fact - ask for the sources, then check them.
Try things. Experiment. Fail and learn - we're all figuring out how and when it's right to use AI.
Most of this runs on the AI you're already paying for. Hardly anyone has it set up that way. That's the bit we do.
Automate the grunt. Augment the human. Protect the craft.
Want to see what that looks like for a business like yours? Start on the home page - it'll show you how much of the AI you already pay for is still sitting idle.