AI changed what a product is
AI agents and open protocols broke the assumption that products have edges, with consequences for positioning, pricing, and competitive dynamics.
Products used to have clear, fixed boundaries. Companies defined precisely what the product did, set a price, and sold it that way. All the standard practices around positioning, competition, pricing, and sales were based on the assumption that the product would remain a stable, contained thing.
AI agents, open protocols, and the Model Context Protocol have removed those fixed boundaries. An agent can discover an API and insert it into workflows its makers never expected. The product becomes one element in temporary chains that the agent assembles.
The scope of the product is now determined at the time it is used, not when it is built. This makes normal positioning strategies ineffective. Competition is no longer confined to similar products in the same market. It includes any tool or service the agent can use to accomplish the same task.
Subscription and per-seat pricing assume a human buyer making conscious decisions. Agent usage follows a different pattern: sudden bursts of activity, often running automatically. There are now two decision makers — the person with the budget and the agent selecting tools. Data from usage shows what the agent did, not necessarily what the human valued. This makes it harder to tell whether the product truly fits the market.
Old competitive advantages break down quickly. Features, brand, and switching costs count for less when an agent can swap one tool for another on any call without loyalty. Data quality, reliability, and ease of integration hold up better. Trust is still relevant, but it is harder to establish when the evaluator is frequently another AI system.
The API itself has become the actual product. Consistent reliability matters more than lists of features. Pricing needs to work for usage that is autonomous and irregular. Distribution has moved to the places where agents find and choose tools: registries, default settings in agent platforms, and discovery protocols. These now control visibility and adoption.
Products are no longer self-contained units. They function as interchangeable parts in larger, fluid systems.