Grounded drafts you can defend: every claim cites its source or wears an amber flag. The question workspace puts the buyer's rubric beside the draft beside the evidence retrieved from your library.
Generating a first draft
Generate writes a first draft from your labelled evidence — outcome-labelled answers from past bids, current facts — with a citation on every paragraph. It usually takes under a minute; the page refreshes itself while the draft is written and the job shows on Activity. Spam-clicking Generate queues one draft, not several.
Drafts never retry automatically. A draft that could not be generated says so with the reason and offers Try again — each run uses paid AI capacity, so retrying is always your call. A tender whose company has reached its daily AI budget is refused with an honest message rather than queued to fail later.
A question workspace opened by a WayWiser bid writer shows the bid's Shared with client / Not shared chip in its head; the question inherits the bid's state, and the control lives on the tender page. A draft the client can see may cite a document the writer has not shared — that citation is exactly what the amber flag below marks on the client's side.
Citations and the amber flag
Each paragraph carries chips naming the evidence it drew on. A paragraph WayWiser could not ground in your library carries an amber flag — "Unverified claim — fix with facts bank or edit manually". Fix queues a re-draft of that block from a matching facts-bank entry; the citation attaches and the flag clears. Or edit the text yourself. A flag is a statement of what is not yet evidenced, never a hidden guess.
Only current facts reach drafts: expiring (within 60 days), expired and undetermined facts are held back.
Editing
Blocks are edited in place, moved, tightened (an AI pass that keeps claims, numbers and citations) or deleted; a deleted block can be restored from the undo strip. The word count is shown against the question's limit. Save snapshot keeps a version; saving the same content twice makes one version, not two.
The evaluator review
Run review scores the draft against the buyer's rubric and returns coverage per criterion and the specific gaps to the top band, cited to the guidance. It is a review, never a score forecast: WayWiser does not tell you what mark you will get, and will not until it has enough scored outcomes to calibrate against. A review in flight shows as such; the result is announced once when it lands.
Reviewers and sign-off
Assign a reviewer from your team and request a review; comments and replies thread under the draft and can be resolved. The assigned reviewer, or an admin, signs the question off. Two humans on one bid — no screen-share needed.