AI in construction: check the evidence before drafting the tender
Use AI to organise requirements and locate relevant evidence before asking it to draft a tender response. The useful first output is a record of what you can support, what is missing and who must resolve it.
Start with the current requirement
Tender packs often contain several documents, clarifications and versions. Identify which material governs the question you are answering. Preserve the relevant wording and location so a reviewer can return to it.
Ask for the requirement, its supporting source and any uncertainty. Keep the client's request separate from the assistant's suggestion about how to respond. A useful suggestion is not an additional mandatory requirement.
Then draft within the evidence
Once the team has resolved the material gaps, ask for a response using only the approved claims. Mark any remaining questions for the bid owner. Check the draft against the requirement, not just against the previous draft.
Reuse approved company information deliberately. Old project descriptions can contain superseded personnel, certifications or delivery plans. Their presence in a shared folder does not make them current.
Keep internal notes and unsupported suggestions out of the submission copy. Preserve a separate review record so someone can see why a claim was included.
Keep neighbouring workflows separate
Preparing a tender, comparing quotations and chasing subcontractor documents may use similar files, but their decisions and owners differ. Define which job the first pilot covers. Otherwise a useful document search can turn into an unclear promise to automate everything.
For a first trial, choose one response with an experienced reviewer. Count time spent gathering sources and resolving gaps, as well as drafting. Then try the method on a second opportunity before deciding it is ready for wider use.
What a useful result looks like
- A response that answers the actual question.
- Source-backed claims the bid owner can check.
- Missing evidence that remains visible.
- A clear distinction between completed work and proposed approach.
- A final version reviewed by the people responsible for the submission.
Editorial guidance informed by Acuity's work with teams. Examples are fictional and client materials stay private. To record who owns an AI use and when it is reviewed, see AI Register.
