The platform as an intelligent backend

July 5, 2026

Powerful platforms still bottleneck on the UI. Every workflow means training users on screens; the value of a deep feature set is gated behind knowing where to click. For a merchandising platform with years of algorithms inside it, that’s an absurd tax.

So the bet I’m making: treat the platform as an intelligent backend. Expose it as a set of well-described tools — MCP servers plus a skills layer — that an enterprise AI assistant can call. The interface becomes conversational; the platform’s algorithms stay the source of truth.

What changes

A merchandiser asks for what they want in plain language — “rebalance next month’s OTB for the stores that under-sold this range” — and the assistant orchestrates the underlying platform: the right tools, in the right order, with the platform’s own optimization doing the heavy lifting.

Three design principles have mattered most:

  • The tools carry the semantics. An MCP tool description is product surface now. Writing “what this tool is for, when to use it, what it returns” well is the new UX design.
  • Skills encode the workflows. Individual tools are verbs; skills are the sentences — the encoded judgment of how an experienced merchandiser sequences a decision.
  • The platform stays authoritative. The assistant never invents a number. It routes intent to algorithms that were already trusted before AI arrived.

The strategic effect is bigger than convenience: it decouples the platform’s value from its screens. Clients can bring their own assistant, and the deep feature set becomes something you can just talk to.

This one is early. Expect follow-ups as real client workflows land.