Tashara Fernando, head of data and AI at Man Group, stood up at Code with Claude London 2026 and said this.
There are trading signals running right now in production at Man Group, a regulated investment firm with over $200 billion under management, that were researched, back tested, and proposed by AI. AI came up with the idea. AI got the data. AI ran the back test. AI wrote the strategy proposal. AI productionized the signal. Humans reviewed the output.
She did not tell us what the signal was. She told us the foundation. The foundation is the part that generalizes.
The foundation is skills governance.
Man Group went all in on adoption first. Workshops. Hackathons. Show and tell sessions. Everyone writing skills. Adoption was, in her words, out of this world.
Then the cracks. The skills were written by power users, not by workflow owners. One guy at the firm travelled a lot, hated doing expenses, wrote a skill that ate receipts and produced reports. It worked. He shared it. Then the expense approver came around asking why Claude was generating expense reports against his cost center for people in technology, in the people team, in departments he had nothing to do with.
The cost center code was hardcoded. The skill worked for the guy who wrote it, so it worked for everyone on his team, so it would work for everyone. Nobody had reviewed it. He was not accountable.
That story is funny in expenses. In back testing where signals run real capital, it is a non starter.
Their solve was a common marketplace called My Knowledge. Every skill visible, tagged, with evals. Owned by the actual workflow owner. Tracked. Reviewed. Versioned. Retired when it should be. She compared it to a library. Sections per department. Care for every item. That care is what makes it a library instead of a junk drawer.
The numbers. Around 1,800 people in the firm. 750 use Claude Code. Developers, quants, finance, the people team. Over 100 governed skills and at least as many community skills, all in the library, all visible.
If you do not run systematic trading, the specifics do not apply. The pattern does.
Your organizational context is your moat. The frontier labs are not going to solve it for you. It is not on the internet. They do not know your workflows. You have decades of context. The work is on exposing it, not reinventing it.
Three things to do.
Decide who owns each workflow. Not who uses it most. Who owns it.
Build the platform before you encourage adoption. Visibility, tags, evals, versioning, ownership, lifecycle.
Treat skills like production code. They will run inside AI loops that touch real systems.
Man Group can have AI propose trading signals that run real capital because they built the infrastructure that makes AI output reviewable, comparable, and tied to the right workflow owner. That is the part you build.
Where are you in this cycle? Still in adoption mode? Hit the expense report moment yet? Building the marketplace?