Decide what to build, buy and stop
Which parts of the business AI improves, what the return looks like once someone checks the arithmetic, and what to do first. Then we help you do it.
Pilots are cheap. Stopping things is expensive.
Two years into this, most organisations hold a drawer of pilots, a licence bill that keeps climbing, and no account of what any of it returned. Getting a model to do something impressive in a demo was never the hard part. Choosing which workflows are worth changing is, and so is telling the sponsor of a favourite pilot that it ends here.
What the work involves
- 01
Workflow mapping
We sit with the teams doing the work and map where the hours go, which is rarely where the process document says they go. Automation candidates fall out of that map rather than out of a vendor's feature list.
- 02
Sizing the opportunity
For each candidate: hours addressable, a realistic automation rate, the quality risk and the cost of being wrong. Numbers you would put in front of a CFO, including the ones that argue against going ahead.
- 03
Build, buy or wait
A recommendation per candidate with the reasoning written down. Wait is a real answer and we use it where a category is moving faster than your procurement cycle.
- 04
Operating model
Who owns AI decisions, who approves a new tool, who reviews the output and where the budget sits. Programmes stall here more often than they stall on technology.
- 05
Sequenced roadmap
Twelve months, ordered by return and by dependency, with the compliance implications marked on each item. Finding a high-risk classification after the build is the costly way to learn it.
- 06
Measurement
Baselines taken before anything changes, and the two or three numbers each initiative will be judged on. Without a before, nobody can argue the after.
How this usually runs
What moves the price is how many workflows are in scope and how many people have to agree on the answer. The diagnostic is fixed price, so the step that decides whether the rest is worth buying never carries open-ended risk.
Opportunity diagnostic, 2 to 3 weeks
Fixed price. Workflow map, sized opportunities, a build-buy-wait call on each, and a sequenced roadmap.
Leadership session, one day
A working day with your executive team to pressure-test the roadmap and commit to what happens next quarter.
Transformation support, 3 to 12 months
We stay through delivery: chairing the steering group, holding the sequence honest, and doing the parts your team has no capacity for.
What you get
- Workflow map with time and cost attributed to each step
- Ranked opportunity register with the assumptions stated on every line
- Build, buy or wait recommendation per initiative
- Operating model covering decision rights, approval path and budget ownership
- Twelve-month roadmap sequenced by return and dependency
- Baseline measurements and the success measures for each initiative
A good fit if
- Leadership teams with pilots that never converted
- Companies whose licence spend grew without a matching account of return
- Founders deciding whether to build an AI capability in-house
- Businesses whose competitors have started making claims nobody can verify
Before you ask
Will you tell us not to use AI somewhere?
Often. The two failure modes we see most are automating a process that should have been deleted, and putting a model where a database query would do. A diagnostic that never returns a negative recommendation is not a diagnostic.
Do you resell tools or take commissions?
No. We hold no reseller agreements and take no referral fees, so the build-versus-buy call has nothing riding on it either way.
We already have a strategy deck from a large firm.
Send it over. Sometimes the analysis is sound and what is missing is sequencing and someone to run it. We will say so rather than starting again and charging you for the privilege.
Start with a scoping call
Thirty minutes. Tell us what you run and where it is going wrong, and we will tell you what we would look at first.