Problem
Is your AI creating value, or dispersion?
- Copilot is available, and everyone uses it their own way
- Prompts circulate with no common frame
- The content produced reads well, but sometimes sits outside the guidelines
- Internal approvals slow adoption down
- AI generates trials without transforming the workflow in any depth
In regulated environments, AI provokes as much caution as curiosity. With no structuring frame, caution creates inaction. With no governance, experimentation creates dispersion.
Plenty of organisations deploy Copilot.
Few genuinely structure how it is used.
Tension
A solid editorial method. An AI with no frame.
SPW ARNE already had:
- demanding editorial guidelines
- a standard model for administrative procedures
- a structured SEO method
- a clear seven-question reference
The system existed. AI was arriving.
The question could have been how to train the teams on Copilot. The real question was how to bring AI in without weakening an editorial system that was already structured, in an organisation where compliance is not optional.
At that point the project changed in kind.
This was not about adopting a tool.
It was about structuring its use.
Role
A structural project, and a transfer of method.
We did not deliver Copilot training. We installed a Prompting Operating System.
- Map the organisational risks of unframed AI use: find the breaking points before they happen.
- Turn the existing method into a prompting architecture: AI does not replace the method, it activates it.
- Build a coherent library of operational prompts: a shared resource, not scattered individual practices.
- Design a steering engine aligned with the seven questions: every prompt follows the questioning logic the teams already know.
- Formalise the validation and control steps: compliance stays intact, speed increases.
- Transfer the method: SPW ARNE steers its own system.
Every deliverable was designed as a brick of governance.
This project was not technological.
It was structural.
Solution
A Prompting Operating System built into SPW ARNE's Microsoft 365 environment.
- Strategic prompt library
50 structured, contextualised prompts, aligned with SPW ARNE’s editorial guidelines and with the real steps of administrative procedures: editorial production that can be steered across the organisation, without depending on individual initiative. - A structured POS Engine
A workflow built in Excel, aligned with the seven questions, with clear sequencing and control points built in: AI stops being a self-service tool. It becomes a component of the editorial system. - Governance secured
Roles clarified, validation processes maintained, trade-offs formalised: in a public environment where every piece of content commits the institution, structure protects compliance as much as it speeds production up. - Internal autonomy consolidated
Method transferred, usage guide built in, the capacity to create new prompts: SPW ARNE no longer depends on an outside expert. It develops its own sovereignty over how it uses AI.
Impact
A reinforced architecture, and time gained.
The strategic framing from this engagement modelled significant gains on two fronts: an estimated 30% reduction in the time spent producing and validating administrative content, and a measurable improvement in editorial compliance from the first weeks of structured adoption.
These projections rest on an analysis of the existing workflows and on the governance recommendations delivered to SPW ARNE.
SPW ARNE moved from a team experimenting with AI to an organisation steering its prompting system.
Copilot did not create disruption.
It reinforced the architecture already there.
Do you have a system able to bring AI in without losing control?
In environments under heavy regulatory constraint, the public sector, environment, agriculture, European compliance, AI without a system is not an opportunity. It is an operational risk.
With a Prompting Operating System:
- You secure usage across the organisation.
- You align AI with the editorial method you already have.
- You develop governance your teams run on their own.
The real question for your leadership is not whether your teams use AI. It is whether you have a system able to steer that use over time, without compromising your validation processes or your guidelines.