Marouane Harmach
AI strategy assignments that turn business priorities into governed use cases, experiments and an actionable roadmap.
A limited set of use cases with evidence, dependencies and decision criteria.
Rules for tools, information, validation and incident handling.
Sequenced experiments, capability work and executive arbitration.
Contact Marouane Harmach through Consultor with the organisation, decision, participants and expected timing.
A board does not need to understand model architecture. It needs to know which parts of the organisation's work are exposed, which data would be used, what level of error would be tolerable, and who would carry responsibility for a contested output. Those four questions separate a mature subject from an intuition.
The engagement therefore starts by listening to actual work: repetitive tasks, delays absorbed rather than solved, documents produced in series, customer requests handled manually, decisions waiting on information. It then looks at what already exists, including informal use of generative assistants installed without any framework. Ignoring those practices does not remove them; naming them makes them governable.
Most AI programmes fail on the initial choice rather than on technology. A spectacular but peripheral use case consumes budget and changes nothing. A modest one, repeated daily by fifty people, produces visible effects within weeks.
The selection grid stays deliberately simple: expected value, frequency, feasibility, data availability and sensitivity, integration effort, risk exposure, adoption capacity. Each candidate is described on a single page — user, current situation, expected outcome, the exact role of the system, the human validation point, the success indicator. That sheet becomes the unit of comparison between proposals coming from different departments, which prevents arbitration by intuition or internal influence.
A general charter is not enough. Useful governance distributes precise decisions: who authorises a new tool, who qualifies the data it may process, who checks output quality, who handles an incident, who monitors vendor commitments. It also states what remains prohibited, and why.
The level of control is calibrated on actual risk. An internal drafting aid and a system influencing an individual decision do not call for the same scrutiny. Too light, governance lets everyone improvise; too heavy, it pushes usage outside official channels — the worst of both outcomes.
A strategy approved in committee produces nothing until teams change what they actually do. The engagement therefore plans from the outset how uses will be explained, tested, corrected and supported: workshops on real cases, reference points for checking an output, rules on information that never leaves the organisation, a channel for asking questions without fear.
Cadence matters as much as content. A monthly review of early usage allows rules to be adjusted, unproductive experiments to be stopped and gains to be made visible. That short loop, more than the strategy document itself, is what embeds the practice.
No. Digital transformation addresses processes, systems and services. An AI strategy frames use cases, data, risks and governance specific to artificial intelligence. They are built consistently, under the same executive arbitration.
With an inventory of existing, often informal usage, then one or two frequent, measurable and low-sensitivity use cases. The initial objective is collective learning and the first control points, not maximum gain.
By defining scaling criteria upfront: success indicator, named business owner, required integration, running cost and exit conditions. A pilot without decision criteria continues indefinitely.
CONSULTOR · CASABLANCA
AI, digital transformation, communication or monitoring with Monitor: discuss your context and priorities.
Contact Marouane Harmach →