AI readiness review · 4–6 weeks
We look at five areas: the documents and data you would actually use; the processes and volumes behind them; skills and ownership; controls and legal exposure under the Personal Data Protection Act (PDPA); and the economics of running a model once the novelty fades. You receive a findings note of about fifteen to twenty-five pages, a prioritised shortlist, a risk register, a twelve-month sequence, and a ninety-minute working session. This service does not suit a team that already has a funded pilot with a named owner and a clean data path. In that case a supervised pilot or a vendor assessment is usually the better spend. The review is our flagship and is described in full on its own page.
Use-case shortlisting and business case · 2–4 weeks
Workshops with the people who run the process, plus a small number of interviews with finance and risk. We turn a long list into a short one by asking what volume exists, how messy the input is, who will own the result, and what a wrong draft would cost. You leave with a written shortlist and a business case in ordinary sentences. The case states assumptions so they can be challenged. This work is a poor fit if the organisation wants a catalogue of every possible use of AI. We will decline to pad a list that nobody intends to fund. It is a good fit when two or three functions are competing for the same budget and need a shared ranking.
Tool and vendor assessment, including build versus buy · 3–5 weeks
You may already have a preferred supplier, or you may be choosing between a product, a hosted model, and an internal build. We read security packs, data-processing terms, subprocessors, and the region where data will sit. We sit with your team during a trial that uses your files rather than a vendor demo set. The written comparison covers fitness for the job, operational burden, exit terms, and whether building would actually be cheaper once you count people. We do not resell licences and we do not collect a vendor commission. This service is not a substitute for your own security review on a regulated workload. It is a way to go into that review with questions already framed.
Governance, PDPA alignment and acceptable-use policy · 3–6 weeks
Staff need to know what they may paste into a public chatbot and what must stay inside your tenancy. Managers need a rule for human oversight, which means a person owns the decision where the model only drafts. We help you draft an acceptable-use policy, a simple data-flow picture, and a retention stance you can take to counsel. Information on this site, and in our notes, is general guidance and does not constitute legal advice. We will not stand in for a qualified lawyer, and we will say so in the scoping note. The work suits organisations that already have a privacy contact and want AI uses brought into that conversation rather than parked beside it.
Supervised pilots · advise while your team builds · typically 6–12 weeks
Your people, or a builder you appoint, write and run the system. We join weekly, watch the data path, and help you set measures that match the job. If the job is drafting, we do not pretend the measure is “accuracy of a decision”. If retrieval over documents is involved, we look at whether the corpus is labelled well enough to search. You receive a short running log and a close-out note with a recommendation to continue, reshape, or stop. We do not write production code. If you need a delivery partner, we will say that plainly rather than stretch this service into engineering.
Capability clinics for leadership and operations teams · half-day or full-day
A clinic is a working session, not a product tour. Leadership groups use it to practise reading a vendor claim and to agree what “good enough” would mean for a first pilot. Operations groups use it to walk through a real process and mark where a draft would help and where a human must remain. You receive a two-page recap and a list of decisions the room still owes itself. Clinics are a poor fit as a substitute for a readiness review. They work well as a way to prepare a board paper or to reset a team that has already bought tools without a shared rule.