AI augments, never replaces: how we think about Velle's drafting tools

There's a version of "AI for proposals" that generates a glossy draft from a prompt. It demos well and loses proposals, because evaluators can tell when no one actually thought about their project.

We built Velle around a different principle: AI augments, never replaces judgment.

Everything is sourced

Every AI-generated element traces to a specific source — a chunk from a past proposal, a staff CV, a project record, or the RFP itself. Hover any suggestion to see exactly where it came from. Content with no traceable source cannot be saved. Hallucinated claims can't ship by accident.

Everything is flagged

AI content carries a confidence flag based on how much it changed:

• Low — a near-verbatim lift with light edits.
• Medium — synthesized from multiple sources for this RFP.
• High — gap-filled with model knowledge, which must be reviewed and accepted before the section is marked drafted.

You're always the author

Three drafting modes — pull from your library, AI draft, or write from scratch — and you can move between them freely. The AI removes the busywork of finding and recombining; you keep the judgment about what actually belongs in front of this client.

That's the line we hold: the machine does the fetching and the formatting. You do the winning.