Our approach
Problems before solutions. Levels, not a leap.
Every client walks in with their own mess, but the mess abstracts to the same eighteen problems. We name them first, who bleeds from each and where each comes from, before any technology appears on a slide.
Origins
Where each problem comes from
Every problem carries an origin. Roughly half are amplified or outright introduced by adopting AI. We say so up front, because we govern the problems we bring.
- exists today
was there before any AI
- AI amplifies it
gets worse once AI arrives
- arrives with AI
only exists because you adopt AI
The knowledge problems
Knowledge
what your organisation knows
The bottleneck people
exists todayCritical knowledge is oversubscribed: a few people hold it, and every decision routes through them.
Logic locked in code
exists todayThe business logic only exists inside the code, unreadable to anyone who doesn't program.
Documents that drift
exists todayDocuments drift from reality and compete. Nobody can say which version is the truth.
Confident wrong answers
AI amplifies itThe people who most need answers are the least able to verify them, and a fluent AI makes confident wrongness worse than silence.
No shared place to ask
exists todayThere is no one place where engineering, product and leadership read and ask. Context moves through meetings and chat archaeology.
The workflow problems
Workflow
how the work gets done
Technique in one head
exists todayEach engineer's technique lives in their head. It isn't shared, isn't repeatable, and leaves with them.
A different path every time
exists todaySimilar work follows a different path every time: different gates, different artefacts, different quality.
Faster than review
arrives with AIAI produces work faster than humans can review it. The bottleneck moves to verification, and ungated AI action on a revenue path is the nightmare scenario.
The governance problems
Governance
how safely, and at what cost
The shadow shortcut
arrives with AIIf the approved path is slower than the shadow one, people take the shadow one. No policy fixes that. The approved path has to win the race.
Who accessed what?
AI amplifies itNobody can answer “who accessed what, when”, a standing compliance gap that AI calls make acute.
Data out, poison in
AI amplifies itData crosses the boundary in both directions: customers' personal details reaching models on the way out, poisoned inputs on the way in, and centralised knowledge becomes one queryable door.
Invisible cost, unproven value
AI amplifies itCost is invisible and value is unmeasured. There is no baseline to prove whether AI helped at all.
Built on shifting vendors
arrives with AIThe whole capability rides a fast-moving vendor landscape. Models deprecate, prices shift, and lock-in threatens an investment meant to outlive any single tool.
The adoption problems
Adoption
whether any of it sticks
Tried once, forgotten
arrives with AITools land, habits don't change. If the first experience isn't visibly better than working unassisted, people rationally go back.
Never urgent enough
exists todayKnowledge work always loses to delivery pressure. It's the first thing to slip when a deadline looms, and the system is worthless until someone puts what they know into it.
Leaves when they leave
arrives with AIThe capability stays with the vendor, or with one enthusiast, and dies when they leave. It must outlive any one person or vendor, including us.
Experts holding back
AI amplifies itThe experts whose knowledge the system needs most have the strongest rational reason to withhold it: capture reads as “making me replaceable.” The system has to make them more valuable, not interchangeable.
Too technical a door
arrives with AIThe way in is too technical for most of the organisation. The capability can be perfect. The interface alone caps adoption. For most people, the right door is chat.
Why the order matters
Two dependencies make the four spaces a sequence, not a menu: knowledge feeds everything, because workflows without captured knowledge produce generic output; and adoption carries everything, because a system nobody uses, authors into, or owns is shelfware, however good the rest is.
The ladder
Levels, not a leap
No problem is solved zero to one. Each climbs a maturity ladder, and the one rule is never build a level before its trigger fires. Levels are per problem, not per project: a client can be at L3 on governance while still at L1 on product knowledge.
Status quo
nobody yetKnowledge in heads, workflows personal, AI ungoverned, adoption nil. The baseline of pain, and the baseline we measure against.
Capture
engineersWrite it down where the work happens. The best ways of working become named, repeatable skills, and today's state is measured so improvement can be proven later.
Serve
PMs and non-engineersOne governed door over everything captured. It can read and suggest, but can never change anything itself. Chat becomes the shared place where product and engineering ask the same brain.
Govern
security and the C-suiteThe gateway: credentials kept out of the AI's reach, access control, personal details redacted, a tamper-proof record of every action, cost metered per team. Quality and adoption become measured gates, not hopes.
Scale
the whole organisationThe system starts feeding itself: capture runs automatically, answers draw on everything it has learned, and each workflow earns more independence only against its logged track record.
The tag on each level is who the value lands with, which is the part that decides whether anyone renews.
What we actually build
A handful of things, not eighteen
You don't build eighteen solutions. A handful of built things (the shared memory, the governed door, the gateway, the skills, the checks that catch wrong answers) and a handful of practices (pairing on live work, teaching your team to add skills, naming the experts' fear instead of managing around it) cover the whole grid.
The adoption space is answered almost entirely by things you do, which is why it goes missing from most architecture diagrams, and why most AI initiatives quietly die there.
See what this looks like running in production in our note on what an installed AI brain actually does.
Start here
Which of the eighteen is bleeding the most for you?
That's the first conversation: thirty minutes, your systems, an honest answer about where you are on the ladder.