Computer Vision — Orange Business
Product strategy and partner-led commercialization · Industrial pilot
The starting point
An internal team had spent roughly two years on face-recognition technology without making sales. The company was still paying for people and infrastructure while deciding whether to close the direction.
I inherited it after those two years. The question was whether the technology could serve a valuable customer workflow, or whether the product approach had to change.
Start with an industrial decision, not a recognition algorithm.
What I did
- Reassess what could sell I researched use cases, segments, competitors and unit economics. The analysis indicated that an offer built only on the in-house face-recognition core would not be competitive enough. I evaluated partner platforms on licensing, deployment, detector integration and support.
- Build around complementary strengths I built a partner-platform strategy combining mature video analytics with Orange’s cloud, connectivity and managed services, plus specialist computer-vision expertise. The portfolio grew beyond recognition into industrial and operational use cases.
- Design the workflow after detection The work covered mining safety, PPE checks, manufacturing tracking and vessel operations. In one maritime scenario, helmet and life-jacket detection were considered alongside door status and access logic. The product question was what decision a detection should trigger, beyond whether a model could spot an object.
- Use pilots to test value I linked the partners, investment work, presales and pilot development. Not every opportunity deserved more funding; one manufacturing concept stopped because the customer didn’t accept the business case.
The outcome
In approximately six months, the direction went from closure risk to investment, a partner-based cloud platform, and initial industrial pilots and opportunities.
This turnaround reached productization and pilots, not proven commercial scale. No signed revenue or safety improvement is claimed. External disruption in February 2022 halted the commercial progress. Specialist teams owned model development.