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.
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 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.