Your best expert, at every desk.
We turn how your most experienced people work into skills an AI can run, so everyone on the team can use their expertise every day, and it gets better with every run.
A skill is a set of written instructions an AI follows step by step, built from how an expert works.
The data that matters most for AI in your organisation sits in no system.
- It sits with your best people. You know who they are: the person everyone asks.
- They have never written it down, and often can't easily put it into words.
- Your systems record what was decided. Only your expert knows why.
- It doesn't scale past their calendar, and if they left next month, work would stall.
For an owner or a board, that is operational risk. It is also expertise the business depends on and doesn't own.
build
From one expert to the whole team.
It starts with a conversation with the person everyone asks.
your expert
a few hours of interviews
We capture how your expert works.
We interview them on real work: the questions they ask, the trade-offs they weigh, and where they stop to decide. A few hours of their time in total.
the skill
AI drafts, expert signs off
Written instructions an AI follows.
AI drafts the skill from the interviews. Your expert edits it and signs it off before anyone runs it, and it stays their method.
your team
runs it every day
Your team runs it as a skill.
Colleagues run it on their own work every day, and it asks them the questions your expert would ask. The expert reviews the output rather than explaining the same thing again.
It improves every time it's used.
Each run shows a gap. The gap goes back into the instructions, so the next run is better than the last.
The AI does the analysis. It stops where a person has to decide.
A go-to-market skill working out where a team should spend its time. The AI builds the long list, filters it and scores what survives.
Then two options tie on score. The numbers can't separate them, so the skill asks the team which one they have a reason to believe in, and carries on with their answer.
A sample run of a go-to-market prioritisation skill. The AI calibrates, builds a long list of 41 opportunities, filters it to 12 and scores them. Two opportunities tie, so the skill pauses at a human checkpoint and asks the team which one they have a reason to believe in. The team answers, and the skill sequences five opportunities to pursue now and parks seven.
three ways to work with us
Where you start depends on where the expertise sits.
Every route ends with a capability that stays with your team.
Academy
Train
Your team has the expertise and is new to working with AI.
Build & Run teaches your people to stand up an AI-native system on your own stack, guided live. Fixed price, book online.
explore Academy →Applied AI
Build
The expertise is here, but it lives in a few heads.
We turn how your best people work into skills an AI can run for the whole team, from a few hours of their time.
explore Applied AI →Platform
Run
The expertise is already written into skills.
Run each skill on the model that does the job best, test it against examples your experts approve, and feed your team's notes back in. Early access, by invitation.
explore the Platform →Not sure where it sits? A Capability Quick Scan maps it in two weeks and shows which route pays off first.
see the Capability Quick Scan ↓run
How a skill keeps getting better.
Once your team relies on a skill, the Third Vector Platform runs it on several models and shows which does the job best. It tests every new version against examples your experts approve, and sends your team's notes on each output back into the skill.
models
The same skill, run on several models, with a scorecard for each. Pick the one that does the job best.
gold sets
The cases a skill has to get right, approved by your experts. Every new version is tested against them.
feedback
Your team's notes on each output go back to the skill, and shape its next version.
Models come and go. Your skills stay. In early access today, by invitation, with the teams we build skills for.
explore the Platform →Expertise that gets used every day.
process documents
Written once, rarely read, and out of date soon after.
knowledge tools
Store what your experts know, so people can search for it.
skills
Read every time an AI runs them, so your expert's method shapes every piece of work. Anyone on the team can read one, question it and improve it.
Built on Agent Skills, the open standard developed by Anthropic and supported by tools from OpenAI, GitHub, Google and Mistral. Your skills move with you if you change AI provider.
People keep the calls that need judgement.
Each skill has human checkpoints where a decision matters: what the work has to prove, which assumptions to test, which factors decide the case, and when the answer looks right but something feels off.
see where judgement sits →Start with a Capability Quick Scan.
Know where skills will pay off before you commit a budget. In two weeks we stress-test the expertise your business runs on: where it is scarce, where it is critical, and where work waits for it. Then you choose the next step: a skill built with your own expert, training for your team, or both.
what's included
- ✓A short AI fluency survey that shows which teams are ready to use skills, and where training should come first.
- ✓A capability map, ranked with each department head, that shows where your business depends on expertise only a few people hold.
- ✓A business case for the two strongest skills: the work, the expert, the people who would run it, and the hours it frees.
- ✓A read-out with you and the heads involved, where you decide go, change or stop on each skill.
- ✓You keep the survey results, the map and both business cases, whatever you decide.
price€2,500, ex VAT, two weeks
Credited in full if we go on to build a skill with you in a Skill Sprint within 90 days of the read-out.