We shipped six modules and spent a year learning that none of them matter much on their own.

The generation is commodity

Drafting a post, an email or a brief is something every model does competently now, and the gap between the best and the merely adequate keeps narrowing. Building a product on the quality of raw generation means building on ground that shifts underneath you every few months.

What does not commoditise is knowing things about a specific business that are not in any training set.

Grounding is the moat

A draft is usable when it is correct about your pricing, your positioning, your customers and what you shipped last week. That information lives in documents, tickets, threads and people's heads, and getting it into a form a model can use reliably is genuinely hard work.

Once we saw it that way, the Knowledge Center stopped being a supporting feature and became the thing the other five modules are clients of.

What changed in the roadmap

The order of work inverted. Ingest quality, source freshness and citation accuracy moved ahead of new generation surfaces.

  • Every module reads from one knowledge base rather than keeping its own context
  • Citations became mandatory on generated output, not optional
  • Source staleness is surfaced in the product rather than discovered through a bad draft
  • New module work is gated on whether the grounding exists to support it

The uncomfortable implication

It means the value of the product scales with how much work a customer puts into their knowledge base, which is a worse onboarding story than "connect a channel and go". We think it is the true one, and we would rather build against it than around it.

We use cookies

We use strictly necessary cookies to run the site, and optional cookies to understand usage and measure marketing. You can change your choice at any time. Cookie policy