Mohammed Cherifi is fractional CPO & Interim Head of Product | Physical AI Product Management at Hyperion Consulting.
Consider a hypothetical industrial inspection service: A mobile robot uses AI to flag unusual thermal patterns around pumps and motors in a manufacturing plant. The customer wants useful maintenance findings before the next shift. Even a capable detector cannot deliver that promise when blocked routes leave equipment uninspected or uncertain alerts overwhelm the technicians reviewing them.
Suppose a customer asks for inspections during production instead of scheduled rounds in cleared aisles. The product team must choose whether to support that arrangement, limit the offer to scheduled access or target plants with more predictable routes. Each option changes coverage, robot permissions, human work and cost-to-serve. That is a product decision with consequences across functions.
Defining The Discipline
I propose treating physical AI product management as a distinct discipline: deciding what value intelligent physical systems should deliver, and for whom, and managing the technical, human, operational and economic commitments needed to deliver that value throughout their lives. Its scope begins with selecting a worthwhile problem and extends through product development, operation, adaptation and retirement.
The unit of management is the complete operated product. Here, that includes the robot, sensors, AI models, factory layout, technician responsibilities and maintenance workflow. An accurate anomaly detector addresses only part of the customer outcome. If the robot cannot reach equipment or technicians cannot review its findings, the service promise must change.
Why A Distinct Body Of Practice?
Established product management already addresses strategy, customer value, economics and coordination across functions. Systems engineering already integrates requirements, architecture and lifecycle concerns. Physical AI product management should build on those foundations. The case for a distinct discipline rests on organizing specialized judgment where learned behavior, physical consequences, operating conditions and human work jointly determine the product promise.
That requires a shared body of practice: recurring decisions, methods for resolving them, identifiable competencies and clear accountability. The same reasoning can connect robotic inspection, machine tending and material handling wherever AI materially shapes physical work. Whether organizations establish a dedicated role or develop existing product leaders, they need an explicit mandate spanning five responsibilities.
Five Responsibilities Across The Product’s Life
First, select the opportunity. Identify the buyer, the people doing the work and the outcome worth paying for. For the inspection service, timely maintenance findings may matter more than the number of images collected. Compare mobile inspection with fixed sensors or manual rounds and estimate installation, supervision and support obligations before committing.
Second, discover in operating conditions. Study equipment access, machine loading, ambient conditions, aisle traffic, connectivity and technician decisions alongside model performance. Identify the assumption most likely to change the investment decision, then design a test around it. A controlled detection trial and a production-shift workflow study answer different questions; both can influence what the product should become.
Third, shape a repeatable product. Decide which robot, sensors, maintenance-system interfaces, technician arrangements and service capabilities belong in the offer. Specify the factory variations the product can accommodate and those requiring additional work. These choices affect the roadmap, partner responsibilities and cost-to-serve, as well as technical architecture.
Fourth, govern deployment and autonomy. Specify what the system may do, under which conditions and with what evidence and human support. I call this decision “authority release.” Inspecting during production requires examining shared routes, worker interaction and recovery arrangements. This is one method within the wider discipline, with engineering and assurance judgments retaining their authority.
Fifth, manage continuing value. Use field information to reassess customer outcomes, support demand and product economics. If blocked routes and false alerts consume technician time, examine inspection schedules, customer segments, pricing or autonomy scope. The appropriate product decision may be to expand, redesign or retire an offer. Operation keeps product strategy open to revision.
Give Product Leadership A Clear Mandate
The product leader needs delegated authority to connect these responsibilities: choosing target customers, shaping commitments, prioritizing investment and revising the offer when its assumptions fail. The role must influence what the organization builds, sells and continues to support, with unresolved tradeoffs made visible.
Engineering, assurance and operations can then retain their specialist decisions and stop rights. Product leadership can incorporate those judgments into the offer and its priorities. The role requires technical literacy, customer and operator discovery, service economics and the ability to resolve disagreements across functions. Organizations can develop these competencies within existing teams.
For the hypothetical inspection service I shared above, one initial offer could promise reviewed findings for agreed equipment during scheduled access windows. Production-time inspection would be a separate commitment requiring its own evidence and service arrangement. A narrower offer can be a stronger product when an organization can deliver it repeatedly and customers value the result.
Start with one physical AI product. Name its complete customer promise, identify who can change it and give that role a mandate from opportunity selection through retirement. Build your methods and competencies around those decisions. That is how I propose putting physical AI product management into practice.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
