BidVault · AI Tender Response Drafter

Onboarding checklist

Collect these from the client before drafting. Anything missing becomes a [CLIENT TO PROVIDE] tag in drafts.

Add documents

Files are converted to clean text and stored in this browser only.

Drop files here or tap to upload .md, .txt, .docx or .pdf

Paste text instead

Document title Document type Content

Vault documents 0

Import client

Tender question

Buyer's question Evaluation criteria / scoring matrix (one sub-point per line)

Word limit Character limit (optional) Buyer name Social value guidance

Extra context for this bid (optional)

Evidence EnforcerEvery claim is backed by a metric from a named contract, or tagged [INSERT METRIC] Evaluator SignpostingBold sub-headings that mirror the buyer's wording, plus scoring triggers Quality loopAn AI evaluator scores every sub-point, weak sections are rewritten, every claim is audited, then a plain-English polish Use win themesNo win themes yet: build them in tab 3

Build the matrix first to see every explicit and implicit ask, then use it as the scoring matrix for the draft.

Win themes & buyer intelligence

Run once per tender, before drafting. The themes are then woven into every answer for this client, so the whole tender pulls the same way.

Buyer name ITT pack and specification (paste, or upload below) Upload ITT files Buyer's published strategy or corporate plan (optional)

Feedback Learner

Under the Procurement Act 2023, buyers send every bidder an assessment summary with scores and reasons. Paste it here to turn it into rules the drafting AI follows on every future bid.

Which answer was this feedback on?

Buyer name Score awarded (0–5)

Buyer's assessment summary / feedback Our submitted answer

House Playbook 0

Import playbook

Shared across all clients. General rules go into every draft; buyer-specific rules only when the buyer name matches.

Add a rule Applies to Buyer

Vault gap analysis

Looks across every matrix and draft for this client and finds the evidence that keeps coming up missing.

Predicted vs awarded scores

Drafts where you've logged the buyer's real score. A small difference means the Red Team scorer can be trusted.

Quality benchmark 0

Save 10–20 past questions as a test set for this client. Re-run it after you change the playbook, and keep the change only if the average score goes up.

Each question runs the full quality loop (about 8–10 AI calls), so a 20-question run takes a while.