We treat artificial intelligence as a cross-cutting theme, applied across all of our priority sectors, rather than as a standalone offering. That methodological choice has a practical consequence: we never begin an AI engagement by choosing a technology.
Three conditions, in this order
An applied-AI project only holds up if three conditions are met at once: a use case of proven value, data that is genuinely available and usable, and an organisation capable of embedding the outcome into its processes.
Order matters. A use case without data is an intention. Data without a use case is a storage cost. And the two combined without organisational capacity produce a prototype nobody uses.
What this changes in the scoping phase
In practice, the scoping phase is as much about identifying use cases of proven value as it is about ruling out those with no usable data behind them. This is often the most useful part of the engagement for the client — and the least dramatic.
Next comes an audit of data availability and quality, then — only then — prototyping on the priority use case, and finally industrialisation, paired with a governance framework and the upskilling of internal teams.
Across our sectors of intervention — agriculture, health, financial inclusion, public administration — this discipline is what separates projects that reach production from those that remain a demonstration.