Two years building an AI-native company

Nick Erskine-Shaw|Founder|

Also published on LinkedIn

What we got right, what we got wrong, and why the market is finally ready.

We spent fourteen years building and scaling our last business globally. We hired the teams, opened the markets and learned what growth costs the traditional way: more people, more process, more drag with every step up.

Then AI changed the rules. Not at the edges. At the core. We didn't want to bolt AI onto how we used to work. We wanted to build a company that was AI-native from day one, and help other organisations make the same shift.

So two years ago we started again, with a small question. What does AI actually do to a JD?

It turned out to be the wrong question. Here's what we learned finding the right one, and why the market is finally ready for the answer.

Year one: the JD wasn't the work

Over more than a hundred conversations with brands, one thing kept surfacing. The JD wasn't the work. It was a story about the work, written years ago for a role that had quietly changed shape.

What leaders actually lacked was a baseline. Nobody could say which tasks existed, who did them, how long they took, or which tools and vendors sat underneath.

So most of them skipped it. That was the first swing nearly every organisation took at AI: licences for everyone. An LLM in every employee's hands, a prompt workshop, and hope. Usage climbed, but nobody could say which part of which role AI was changing. We called it learn and burn. People learned as they went, tokens burned in the background, and the organisation kept nothing.

We stopped analysing documents and started mapping the whole organisation. And we let the market steer. Every serious client request, from locking decisions to mapping processes to scaling across thousands of roles, became product.

Going deeper before going wide

Our first version did what we set out to do. It scored roles and showed leaders where AI would land. Clients found it useful.

Useful wasn't enough. A real transformation partner has to go deeper: every task, every process, tool and vendor, every number traceable to its source. And it has to keep measuring after the first report.

That meant more building before we went to market in full flight. It was a hard call. But you can't limp into being the platform an organisation trusts with its workforce decisions. Either the evidence holds up in the boardroom or it doesn't.

Speed is cheap now. Trust isn't.

The Human Layer Graph

What came out of that work is the thing I'm proudest of: the Human Layer Graph, one live model of how an organisation actually runs, task by task, with every fact traced to its source. Where a client's data runs thin, and it always does, our engine (Anima) fills the gaps from industry and market data and then checks them with the people who do the work.

The bigger lesson was what to build on top of it. Consulting gets rebuilt from scratch on every engagement. We turned ours into programmes that run on the graph and get better every time we deliver them. Discovery shows an organisation where it stands. Transform lets leaders see the consequences of a change before they make it, not six months after. Traction proves what actually landed.

The graph never goes stale, so the work never starts from zero again. That's the difference between a project and an operating rhythm.

Running on our own brain

We built Human Layer Lab on the same idea we sell. Every system we use, from email and calls to CRM, code and finance, feeds one connected picture of the business. Every person runs a team of agents.

After fourteen years of scaling the traditional way, the difference is hard to overstate. When everything shares the same context, everything rises at the same speed. A pipeline review across every relationship we hold used to take a week. Now it takes a morning.

We didn't automate everything. Product and engineering, and go-to-market led. The rest followed once the pattern was proven.

And we learned the part the hype leaves out. Connecting everything is not the same as controlling it. You can move fast and miss things just as fast. So we built in governance over what every agent can touch, security from day one, a named owner for every system, a human approving every action that leaves the building, and live visibility over all of it. That control layer is now part of what our customers get.

Year two: into the market

Year two was go-to-market across APAC. By then most organisations were on their second and third swings: agents, internal AI teams, or outsourcing the problem entirely to the big consultancies. Every swing taught the same lesson. The tools work. What's missing is a way to run the hybrid workforce they're creating.

That's when the conversations changed. We won our first listed-company customers, and a competitive deal against one of the world's largest consultancies. Regional consultancies started running their practice on our platform. We lost some too, mostly to firms that chose to build it themselves (although there's still scope to work with those internal teams in 2027...).

The lessons were simple. Big organisations buy proof, not the future. Sell to the two people who actually own the decision, not nine. And price the asset, not the access: an AI-native platform, not AI bolted onto legacy software, at a fraction of the cost.

Owning AI-native consulting

For two years our category had no name. Now the biggest firms are restructuring around it. McKinsey and AWS launched a joint venture in January built on outcome-based pricing. The industry is calling it services-as-software.

For us, AI-native consulting means three things. Agents do the heavy lifting and consultants hold the gate: a human signs every finding and commits every change. The client keeps the platform, a live record of how their work gets done, long after any engagement ends. And our partners get to run a modern practice without building the technology themselves.

Year three

AI didn't make building a company easier. It made it faster. Good decisions compound quicker. Bad ones break things quicker.

Two years ago, few organisations were ready to talk about running a hybrid workforce on a live, AI-native platform. Now they're asking for it. The ones that win the next few years won't have better models. Everyone has the same models. They'll know their own work, run their people and agents on a live picture of it, and prove what changed.

That's what we've spent two years building. The market is ready for it.

Year three starts now.


Sources: California Management Review, From Rate Cards to Outcomes: Consulting's Fourth Transformation, April 2026; HFS Research, Services-as-Software, 2026.

Related reading

Go deeper

We use cookies to improve your experience and analyze traffic.