A ranked use-case register
Each candidate idea scored on value to the business, feasibility with the data you hold, risk if the output is wrong and effort to deliver, so the debate moves from enthusiasm to trade-offs everyone can see.
Service · AI
Commission a short advisory engagement that decides where AI belongs in your product before money goes into building it. We examine candidate workflows, the data you can lawfully use and the vendors on your shortlist, and leave you with ranked options, pilot criteria and a roadmap your leadership can scrutinise.
Advice only, with no build commitment attached
Pilot stop-and-go criteria written down
Vendor comparison on your own sample tasks
Deliverables
Each candidate idea scored on value to the business, feasibility with the data you hold, risk if the output is wrong and effort to deliver, so the debate moves from enthusiasm to trade-offs everyone can see.
Measured or carefully estimated figures for how long the task takes today, how often it goes wrong and what errors cost, without which no later pilot can claim an improvement.
For the top candidates, a comparison of features already inside tools you licence, specialist SaaS products, a custom build and the option of postponing until data or models mature.
What a pilot must demonstrate, how it will be measured, how many cases it needs and which results would stop further investment, agreed before anyone writes code.
Which data categories may reach which providers, the contractual terms to check, retention and regional constraints, and what exit from a vendor would involve.
Early decisions
The most exciting idea is rarely the best first one. We favour a use case with clear measurement, accessible data and tolerant failure modes, so the organisation learns quickly and safely.
A pilot result means little if nobody has authority to act on it. We name the decision owner and the reviewers whose judgement defines a good output.
Some functions can live with an occasional poor suggestion; others cannot. We record the tolerable error rate and the categories of mistake that would be unacceptable at any rate.
Tying a product to one provider’s proprietary features speeds the first release but complicates switching later. We set out what portability would cost and whether it is worth paying for now.
Depending on sector and market, obligations may arise under data protection law or the EU AI Act. We identify questions for your counsel rather than presenting legal conclusions ourselves.
Engagement
Conversations with product, operations, data and compliance leads, plus a review of existing tools, contracts and datasets, produce the long list of candidate uses.
Promising candidates are scored, and the riskiest assumptions are checked with quick experiments on sample data, such as whether a hosted model can read your documents at all.
For two or three finalists we price build, buy and defer options, run vendor tools against the same sample tasks and note where each one falls short.
A working session with your leadership walks through recommendations, pilot criteria and open risks, followed by written material your team can circulate and revise.
Fit
Leadership wants a plan, several teams have ideas, and nobody has the time or neutrality to compare them on the same terms.
Before signing an enterprise contract or funding a build, you want an independent view of whether the expected benefit holds up.
Previous experiments produced demos but no decision, and you need clearer success criteria before trying again.
Alternatives
If you know what the feature should do and have examples ready, going straight to a scoped build avoids paying for analysis you do not need.
Open registerWhere the real issue is legacy systems, team structure or cloud strategy, a general technology advisory engagement fits better.
Open registerProgrammes touching processes, operating models and many systems at once belong with our digital transformation work.
Open registerMany AI strategy documents describe trends and possibilities but leave nobody able to approve or reject a specific investment. We write ours as a sequence of decisions: which workflow first, what evidence would justify continuing, what it would cost to stop and who decides.
That framing changes the conversation with a board or investor. Instead of defending a vision, the team defends a small bet with known stopping rules.
Claims such as “this will save each agent an hour a day” collapse quickly when nobody knows how long the task takes now. Part of our job is establishing the starting figures: handling time, error rates, rework, escalation volume. Sometimes those figures reveal that a process change or a better form would deliver most of the gain without any model at all, and we say so.
A model may be capable of drafting a credit decision letter; that does not mean your organisation can let it send one. We look at who approves consequential output, how errors would be investigated, what customers are told and how records are kept. These questions often decide feasibility more than model quality does.
Vendor demonstrations run on curated examples. For shortlisted options we run the same small set of your real tasks through each tool, under your data terms, and compare results side by side. Weaknesses that matter to you, such as poor handling of scanned forms or a language you operate in, show up quickly this way.
We do build AI products, so it is fair to ask whether our advice leans towards custom work. The engagement is priced and closed on its own, and our recommendations regularly point to buying an existing tool or deferring altogether. If a build is recommended, you can take the specification to any team.
FAQ
The number of teams and candidate ideas involved, how much data access and vendor testing the shortlist needs, and whether stakeholders sit in one location or many. A single product line with a handful of ideas costs far less than a group-wide review across business units.
Length depends on how fast interviews can be scheduled, how long it takes your organisation to grant access to sample data, and how many finalists need hands-on vendor comparison. We agree the interview list and access steps at kick-off so the critical path is visible.
The use-case register, baseline figures, build-or-buy comparisons, pilot designs with thresholds, the data and provider memo, and the slides from the leadership session, all in editable formats your team can maintain.
Everything we write for the engagement belongs to you after payment and can be shared internally or with other vendors. Our general methods and templates remain ours, but the content specific to your business is yours.
Completion is tied to deliverables, not hours. When the agreed register, comparisons and pilot criteria have been presented and your decision owner confirms they can act on them, the engagement closes.
No. Consulting ends with recommendations. If you want us to build afterwards, that is quoted separately, and you are free to take the roadmap to another team or run it in-house.
Related services
The natural next step when a pilot clears its thresholds and you want a production feature built.
Open registerWhen the baseline cannot be measured because the underlying reporting does not exist yet.
Open registerOften the first build recommended where staff spend hours hunting through internal documents.
Open registerPrimary sources
Dated official documentation, standards, and research that support the factual claims on this page.
The official text of the EU AI Act on EUR-Lex.
The international standard for managing AI within an organisation.
How to read this page
Scope, cost and timeline figures are planning guidance until confirmed in a written proposal. Diagrams and screens are illustrations, not client deliverables, and source-code rights and handover follow the signed agreement. See the delivery process and the full disclaimer.
Next decision
Define outcomes, constraints, evidence, rights and handover before delivery begins.
Commercial rights, repositories, environments, documentation, acceptance and handover remain contract-defined.