Cleverseed
AI-powered content editing platform. We built a fully custom rich-text editor with contextual AI actions available inline as writers work.
Writers were being asked to leave the page to get help with the page.
Cleverseed's first version wrapped a third-party editor and put AI in a right-hand panel. Every improvement was a three-step manoeuvre: select the text, move to the panel, paste the result back. Session recordings showed writers abandoning the panel entirely and doing the work in a separate tab.
The wrapped editor also fought back. Custom marks could not be added without patching around the vendor's schema, AI suggestions could not be rendered inline, and each new feature took longer to ship than the last. The team asked us to decide whether to keep extending it or replace it, and to do whichever we recommended.
What we built, and why in that order
“Owning the schema was the unglamorous decision that unblocked everything else. We ship features now instead of workarounds.”
The editor is now the product's differentiator rather than its bottleneck.
Cleverseed ships AI features on a weekly cadence, most of them written by the in-house team against the extension API. Inline diffs turned out to matter more than the model quality: writers accept roughly two thirds of suggestions and report trusting the tool because they see exactly what changes before it changes.
The company is now selling into agency teams, where the collaboration and review features built on the same schema are driving the deals. Our involvement has narrowed to architectural review as new node types are added.
Have something that needs to ship?
Scope agreed in writing, architecture first, weekly increments you can review. Start with a 30-minute call or send the brief.