Optimizing existing processes with AI is a good start. But at some point, marginal gains plateau. Workflows designed five or ten years ago were never built to incorporate artificial intelligence. Optimizing them is no longer enough — they need to be completely rethought.
This is stage 2 of my methodology: the deep transformation of processes around human-AI co-intelligence. We redesign workflows starting from a fundamental question: if we had to design this workflow today, knowing what AI can do, what would it look like?
Complete redesign: how it works
The redesign is carried out with the teams involved, through structured workshops over 6 to 8 weeks:
- Deconstruction — We break down each workflow into elementary tasks. For each one: who performs it? What judgment is required? What real human added value?
- Redistribution — We assign each task to the most relevant actor: human alone, AI alone, or human-AI pair. This is an exercise in intelligent recomposition.
- Reconception — We redesign the complete workflow with the new roles. Sequences change, checkpoints evolve.
- Prototyping and testing — Implementation on a pilot scope, measurement, adjustment.
Responsibility redistribution
The most sensitive part is not technical. It is the redefinition of roles. When AI takes over the first version of a deliverable, the collaborator's role shifts from producer to supervisor, editor, decision-maker.
- Who validates what the AI produces? Against which criteria?
- Who is accountable for the final result?
- Which skills become critical?
- What new roles emerge?
At the end of the engagement, you have newly documented workflows, a human-AI responsibility matrix per process, an upskilling plan, and before/after comparative metrics.