Identifying the AI projects worth launching.
With each business team, we analyse the tasks, volumes, available data and friction points. Every idea leads to a decision: use a tool you already have, run a test, build a solution or stop there. Leadership receives a prioritised roadmap, with the conditions for success and the benefit assumptions to be verified.
Between "train" and "deploy", there is a decision to make.
Training creates momentum. Deploying needs a target. The diagnostic is the step that connects the two: it turns post-training enthusiasm into a roadmap the executive committee can arbitrate.
Train
Teams build skills on their own files and discover what AI can do.
Diagnose
We identify, department by department, where AI truly saves time, and we prioritise.
Implement
We deploy the prioritised cases at the right level: assistant, Copilot Studio agent or custom automation.
Four steps, from field listening to roadmap.
The duration depends on how many departments are covered: a diagnostic limited to one team takes a few weeks, while an engagement across several departments takes longer. The timeline is set during scoping.
Set the perimeter
Which departments, which goals, which tool and governance constraints. We choose together which teams to meet.
and project sponsor
Listen to each department
Short interviews with representative employees from each function: the real day-to-day, the time-consuming tasks, the friction points.
per department
Qualify and prioritise
Each use case is estimated by time saved and implementation effort, then ranked. The quick wins surface on their own.
off-site
Present to the ExCo
Presentation of the mapping, the prioritisation matrix and the roadmap. Each case carries a level recommendation.
and department heads
The Pentagon: five dimensions to frame every project
The Pentagon is a grid of 30 questions we work through with the business team concerned. It covers five dimensions that show whether a case is clear, feasible and useful enough to launch.
- 01DemandWho owns the need, what problem must be solved and what value is expected?
- 02DescriptionHow does the task run, how often, at what volumes and with which exceptions?
- 03DelegationWhat can be entrusted to AI, and which decisions must stay human?
- 04DiligenceDo the data, access rights, source quality and security rules allow it?
- 05DiscernmentHow will the result, the cost and the real benefit be measured?
Setting aside a project that is not framed well enough is also a useful result: time and budget are not committed by mistake.
What does qualifying a case look like?
Example scores: Demand 4/5, Description 4/5, Delegation 3/5, Diligence 2/5, Discernment 3/5.
Customer request handling assistant
- Demand
- 4 / 5
- Description
- 4 / 5
- Delegation
- 3 / 5
- Diligence
- 2 / 5
- Discernment
- 3 / 5
The need and the process are clear: the request has an owner and the task is well described. However, the sources are still scattered, which limits how reliable the answers can be at this stage.
Consolidate the sources, then test on a limited scope.
This chart illustrates the method. It does not reflect a client engagement and is not a scientific measurement.
A usable dossier, ready for the ExCo.
Each priority case sets out the need, a business owner, prerequisites, estimated effort, benefit assumptions, the recommendation and the next decision. Stated gains are estimates to be verified.
The usage mapping
Every use case identified, sorted by department and task type.
The impact / effort matrix
Each case positioned by expected gain and implementation difficulty.
The quick wins identified
What can start right away, with no new tool or budget, to build buy-in.
The roadmap
A deployment sequence over the coming months, milestoned and arbitrable.
The gain estimate
Benefit assumptions per case, to decide on orders of magnitude, to be verified during testing.
The per-case level recommendation
Direct use of Claude or Copilot, a Level 1 assistant, a Level 2 Copilot Studio agent or a more complex automation project.
You only deploy well what you have understood. The diagnostic is there to choose, before you spend.
After the diagnostic, based on your decisions.
The diagnostic is a standalone engagement: the roadmap is yours, and you can implement it with whoever you choose. If you wish, we can take on some of the cases. To first frame the main directions with leadership, see also the Executives & ExCo programme.
Frequently asked questions
The diagnostic begins with scoping, followed by interviews with employees who understand the relevant processes. Tasks, friction points, rules, exceptions and business knowledge are analysed to identify use cases. These are then prioritised according to their value, implementation effort and the conditions needed for success.
Each use case is assessed according to its potential value and the effort required to implement it. Task frequency, time spent, available data, process maturity, risks and adoption capacity can also be considered to distinguish quick gains from more structural projects.
No. The purpose of the diagnostic is precisely to identify opportunities from the teams’ real work. Interviews help make explicit the rules, checks, exceptions and business knowledge that are not always formalised or documented.
The client receives a map of use cases, a prioritisation and a roadmap. Depending on the needs identified, each case may then lead to improving an existing use, implementing an assistant, designing a Microsoft Copilot Studio agent or creating advanced business automation.
Let's define the scope of your diagnostic together.
Book 30 minutes
A scoping call: the departments involved, how mature the teams are, the scope of the diagnostic.
Describe your need
Three lines is enough: reply within 24 h.