Know Where AI Can Create Real Business Value.
Most AI budgets are lost deciding what to build, not building it. Consulting here means measuring your actual processes, scoring the opportunities against each other, and giving you a ranked plan you could hand to any competent engineering team — including one that is not us.
- Vendor-neutral advice
- Scored, not guessed
- You keep every deliverable
Eight questions, one indicative read on where to start
This runs entirely in your browser. Nothing is sent anywhere, nothing is stored, and no contact details are asked for. It will not replace a real assessment — it will tell you whether one is worth booking.
AI Opportunity Assessment
Step 1 of 8
Which industry are you in?
Sector decides what can be automated, what must stay human and where the regulatory lines sit.
Nine pieces of work, delivered as documents you keep
Not all nine appear in every engagement. Which ones you need depends on how far along the decision already is — some clients need the whole path, others only need the build-or-buy call settled.
AI readiness assessment
An honest look at four things that decide whether AI works here: the state of your data, the stability of your processes, the systems it would have to talk to, and whether anyone owns the outcome.
AI strategy
Where AI fits your commercial position — cost, capacity, response speed or a product capability — and what you are deliberately choosing not to do with it this year.
AI roadmap
A sequence rather than a wish list: what to build first because it is contained and provable, what depends on it, and what should wait until the foundations exist.
Process analysis
We follow a real unit of work through your business and time it, including the waiting. The bottleneck is almost never where the org chart suggests it is.
Automation opportunity discovery
Every candidate process scored on volume, variability, error cost, data availability and integration difficulty, so the shortlist is arguable rather than intuitive.
AI use-case identification
Translating business problems into specific system behaviours: what the AI must decide, what it must never decide alone, and what a good answer looks like.
ROI analysis
A cost model built from your own numbers — hours, salary loading, error rework, lost enquiries — set against build cost, running inference cost and realistic adoption.
Technology selection
Which parts should be bought, which built, which left as plain deterministic code, and what the switching cost of each choice looks like in two years.
AI implementation planning
The delivery plan: phases, owners, environments, data access, evaluation criteria, rollout gates and what a rollback would mean for the business.
Three ways to work with us before anything gets built
Pricing is set against scope and discussed directly, because a two-week assessment of one department and a group-wide review are not the same engagement. What is fixed is what you receive.
Strategy Call
Free · 45 minutes
A working conversation, not a pitch. You describe the operation; we tell you where AI plausibly fits, where it does not, and what the honest sequence would be.
You receive
- Two or three specific opportunities named
- A view on whether to start now or fix process first
- Rough order of magnitude on effort
- Written follow-up summary of what was discussed
Best when you are deciding whether this is worth pursuing at all.
Opportunity Sprint
Scoped engagement · 2–3 weeks
A proper assessment. We interview the people doing the work, map the processes end to end, look at the systems and the data, then rank every opportunity we found against what it would cost to build.
You receive
- Process maps for the areas in scope
- Opportunity register with consistent scoring
- Ranked roadmap with dependencies made explicit
- Architecture recommendation and build-or-buy guidance
- Cost model and a first-phase delivery plan
- Executive readout session with your leadership
Best when the decision is real and needs to survive scrutiny from a board or a CFO.
Embedded Advisory
Ongoing · monthly
Senior AI judgement on retainer for teams already building. We sit in your architecture reviews, pressure-test decisions, review evaluation results and keep the roadmap honest as models and costs move.
You receive
- Recurring working sessions with your team
- Architecture and vendor decision reviews
- Evaluation and quality-gate design
- Cost and model-routing reviews as usage grows
- Quarterly roadmap reassessment
Best when you have engineers and need direction, not another delivery vendor.
Every format starts with the same 45-minute call, so nothing is committed before both sides know what the work actually is.
Book the callWhat you get at the end
Six artefacts, all yours, all written to be useful without us in the room. If a document only makes sense while the consultant is presenting it, it was not worth producing.
Process map
Each process in scope documented as it actually runs, including the informal steps, the waiting, the rework loop and the person everything quietly depends on.
Opportunity register
Every candidate scored on the same five axes — volume, variability, error cost, data availability and integration difficulty — so comparisons are like for like.
Ranked roadmap
A phased sequence with dependencies drawn out, showing what must exist before each phase and what value arrives at the end of it.
Architecture recommendation
The target shape: where data lives, what integrates with what, where models sit, what stays deterministic and where humans keep approval rights.
Build-or-buy guidance
Per opportunity, a clear recommendation with the reasoning, the switching cost of each option, and where a subscription genuinely beats a build.
Cost model
A spreadsheet you keep, with your own inputs and adjustable assumptions covering build cost, running cost and the current cost of doing nothing.
When we tell clients not to build AI
It happens often enough that it belongs on the page rather than in a footnote. Saying no early costs us one project and saves you a year, which is a trade we are happy to keep making.
When the answer is no, the assessment still leaves you with something: the process map, the numbers behind the decision, and a written record of the condition that would need to change for the answer to become yes.
The process is broken
Automating a bad process makes bad outcomes arrive faster and in greater number. If approvals loop, if three teams keep separate versions of the truth, or if the rules change depending on who is asking, the process needs fixing before anything is built on top of it. That fix is often free and usually the whole benefit.
The volume is too low
A task done four times a month by one person rarely justifies a system, its integration surface and its maintenance. There is a threshold below which the honest advice is a better template, a shared checklist or twenty lines of script. We will tell you where that threshold sits for your case.
The data does not exist
Models cannot retrieve knowledge that was never written down, and cannot learn patterns from records nobody kept. If the expertise lives in one person’s head and the history is in an inbox, the first project is capture and structure — unglamorous, but it is what makes everything afterwards possible.
A rules-based script is better
When the logic is fully known and the inputs are structured, deterministic code is cheaper to run, faster, fully testable and never wrong in a surprising way. Reach for a model when the input is genuinely messy, ambiguous or written in human language — not because it is the current answer to everything.
What leadership teams ask before engaging
Direct questions, direct answers. Anything commercially specific is better handled on the call.
No, and the deliverables are written so that you do not have to. The roadmap, architecture recommendation and cost model are vendor-neutral enough to be handed to your internal team or put out to tender. Some clients use the sprint output as the specification for an RFP; that is a legitimate outcome and it is priced as consulting, not as a loss leader. We would rather be recommended than retained under duress.
Find out where AI is worth the money
Bring one process that frustrates you. In 45 minutes you will have a clearer view of whether AI belongs anywhere near it, and what the sensible first step would be.
Free 45-minute session · No obligation · NDA on request