A second opinion, from outside the room
The hardest AI decisions are rarely technical. They get made with incomplete information by people who will be held responsible for them afterwards, and the most useful adviser is usually the one with nothing to sell. We read plans, proposals and architectures written by other people - your team, your vendor, your incumbent agency - and say plainly which parts hold up and which do not.
That includes the questions vendors would rather you did not ask. Whether the quoted approach is the simplest one that works. What it will cost to run in year two. Which parts of the proposal are genuinely difficult, and which are a weekend of work priced as a quarter. You end up with an opinion you can take into the negotiation.
Where consultancy earns its keep
Build or buy is the decision most often made backwards. A team picks a tool because it demonstrated well, then spends two years bending its work to fit, or builds something bespoke that a licence would have covered for a tenth of the cost. Getting that call right early is worth more than anything we could do afterwards, and it takes an afternoon rather than a project.
The other half of this work is being on the end of a phone while somebody else builds. Reviewing an architecture before it hardens, checking a model choice against what the data protection team will accept, telling a team that the thing they have spent three weeks on is not going to work. None of that fits a fixed deliverable, which is exactly why it tends not to get bought - and why it is usually the cheapest advice on the table.
- Independent review of an AI plan, proposal or architecture written by someone else.
- Build, buy or wait - including the running costs and lock-in that quotes leave out.
- Model and provider selection against your privacy, residency and budget constraints.
- A written position on AI for a board or leadership team, in language they will read.
- Standing advice alongside an internal team while they build.
Advice without a sales agenda
Strategy and delivery are deliberately kept separate, so our recommendations are not a disguised pitch to build. A significant part of responsible consultancy is telling you when not to build something, when a simpler non-AI solution is the better answer, or when the data simply is not ready. We would rather give you an honest no than a profitable yes.
It is also worth being clear about what this engagement is not: it is advice on AI software, automation and technology adoption, not legal, financial, medical or other regulated professional advice. Where your initiative touches regulated territory, we frame the software and process implications and recommend you take qualified professional advice alongside ours. Our value is clarity and feasibility, not a rubber stamp.
