Where to apply AI, and in what order

The AI use case ideation session

Ninety minutes with your leadership team, facilitated by us. It is use case discovery done properly: how the business makes money, where the time goes, and which of the ideas people already have would actually pay. Nobody is asked to arrive with a list of use cases, and nobody leaves with a vendor shortlist.

Forty-eight hours later you have a roadmap and a set of recommendations, ranked by the business value each one returns. Some will be AI. Some will be a process fix or a system you already own used properly. Each one says what it is worth and what it would take, so the leadership team can decide what to fund first.

It is one of the most requested pieces of advisory work we run, because it is the fastest way to get from a room full of opinions to an order of work.

• A facilitated 90-minute session
• A roadmap within 48 hours
• Recommendations ranked by business value
• AI and non-AI changes side by side

AI strategy and governance

An AI strategy earns the name when it says what to stop as clearly as what to start. Ours is built from the goals the business already holds, and names the few things worth backing this year, in order, and what can wait.

Without a written rule, the cautious teams ban the tools and the confident ones paste freely. Most organisations have either no rule at all or forty pages from a global framework that exist to be pointed at rather than used. Either way the tools stay banned until the policy is finished, and the policy is never finished.

The governance work gives your organisation one rule, written against the obligations your regulator sets, that people can actually follow. Which plan to run on each tool, with SSO, admin controls and retention set so your security team will pass it. A one-page data-handling rule: what may go into a prompt, what may not, and what to do when it is unclear. And a position your board, your auditors and your CIO can all put their name to.

A draft lands within a week and is tested with the people who have to follow it until the cautious and the confident read it the same way. It is not a compliance certification, and we do not claim one.

• What to back this year, and in what order
• Plans and settings your security team will pass
• A one-page data-handling rule
• A position the board can sign

Enterprise architecture

Most roadmaps are drawn on top of an estate nobody has looked at in years. The capability map is out of date, the application list lives in three spreadsheets, and nobody has added up what the systems cost. So the recommendations assume things that are not true, and the first project finds out the hard way.

We map it as it is: the capabilities the business runs on, rated against how much each one matters, the applications behind them with what they cost and where they sit in their life, and where the data actually lives. That shows where a new tool would help, where an existing one is paid for and unused, and where the real fix is retiring something rather than adding to it.

It stands on its own, and it is also the Diagnose phase of the full review below.

• A business capability map, scored
• The application estate, with costs
• Where to consolidate, retire or invest
• A target state the roadmap can build to

Agentic workflows, designed from the evidence

Once the review has found and analysed a process, we can design and build the agentic workflow that runs it. That order matters. A workflow built from a real process map, with the volumes, the exceptions and the people who handle them written down, does the job. One built from a demo does the demo.

Each workflow is narrow, observable and stoppable, with a person checking anything that commits money or reaches a customer. It is built on what you already run where that will do, and on the platform that suits the work where it will not. You own it when we leave.

• A workflow design from the analysed process
• Human checkpoints where they matter
• The build, tested on your own cases
• Handover to the team that will run it