AI & Automation

AI strategy that starts with your process, not a model

An honest assessment of where AI helps your organisation — including where it does not.

Most AI initiatives fail for a reason that has nothing to do with AI: the process being automated was never digital, the data was never clean, or the problem was never the one costing money. Our consulting work is deliberately unglamorous. We map how the work actually flows, find where hours and errors accumulate, and assess honestly which of those AI can address today, which need a system of record first, and which are organisational rather than technical. You get a sequenced plan with costs and expected payback, and a clear list of what we recommend against. That last list is often the most valuable part — the projects not started are the budget preserved.

The problem

What this solves

Boards are asking what the AI strategy is, vendors are answering, and the result is pilots that impress in a demo and never reach production. The underlying issue is sequencing. AI is the third step: digitise, then automate, then add intelligence. Organisations attempting step three from step one produce expensive proofs of concept over data that does not exist yet.

Pilots that never reach production

Impressive demos over data that does not exist operationally.
Sequencing skipped

Attempting intelligence before the process is even digital.
Vendor-led strategy

A roadmap shaped by what a supplier sells rather than what you need.
Unknown running costs

Per-token economics discovered after launch rather than before.
No governance position

Privacy, accuracy and accountability decided ad hoc, under pressure.

Our approach

How we go about it

Start with the operation. We interview the people doing the work and map the flow as it genuinely runs, including the workarounds — which is where the real requirements always live. Then assessment: for each candidate, what data exists, what quality it is in, what the realistic payback looks like, and what the running cost would be. Anything failing that test is written down as not recommended, with the reason. Then a sequenced roadmap, ordered so each step funds and de-risks the next — usually with unglamorous data and process work before any model appears.

Outcomes

What changes for you

Process-first assessment

We map how work actually flows, workarounds included.
Honest not-recommended list

The projects we advise against, with reasons — usually the biggest saving.
Costed, sequenced roadmap

Each step funds and de-risks the next.
Data readiness assessed

Whether you have what a model would need, before anyone builds one.
Governance written down

Privacy, accuracy and accountability decided in the open.

Frequently asked questions

Common questions about ai consulting & strategy.

We do not have much data. Can we still use AI?

Sometimes — assistants grounded in documents need content, not history. But for anything predictive, the honest answer is often no, and the right project is the system of record that starts collecting it. We will tell you which case you are in.

Will you recommend building or buying?

Whichever is correct. We build software for a living and still recommend buying frequently, because a commodity capability built custom is a maintenance cost you fund forever.

How long does an assessment take?

Typically two to four weeks depending on the size of the operation. Most of it is interviews and process mapping; the technical assessment is the shorter half.

Do you implement what you recommend?

We can, and often do — but the assessment stands on its own and is deliberately written so another team could execute it. A recommendation that only works if we build it is not advice.

Related services

Being asked what your AI strategy is?

We will map your operation and tell you where AI pays back — and where it would just add cost.

Start a project Talk to our team