An invitation to tender is not one prompt. It is a pack.
The questionnaire lives in a Word file. The specification is a PDF. Pricing is a spreadsheet. Scoring language is buried in the instructions. Word limits sit next to the question, or three pages earlier, or only in the portal. If you paste a single excerpt into ChatGPT or Claude, you have already lost the buyer's order.
That is why "AI bid writing" disappoints at award stage. The model can write. The team still has to rebuild the ITT by hand.
ITT response software is useful only if it treats that pack as a delivery problem: extract every requirement in the buyer's order, turn it into an answer plan, then draft from evidence your organisation can actually show.
Why award-stage packs break a chat window
A PQQ or selection questionnaire is often one form. An ITT is a set of files that only make sense together.
A writer who starts in a blank response document has to hold all of this in their head:
- the exact question, as the buyer wrote it
- the specification clause that changes what a good answer must contain
- the word, page or character limit
- whether the item is scored, pass/fail, or an attachment
- which pricing or mobilisation assumption the answer must not contradict
ChatGPT and Claude are strong general assistants. They do not keep that structure for you. They cannot see the XLSX schedule you did not paste. They will happily draft a fluent answer to question 3.2 that ignores the limit in 3.0 and the evidence rule in the specification.
The failure is not weak prose. It is an incomplete brief.
Extract in the buyer's order
The first job is not writing. It is making the ITT inspectable.
ProposAI reads DOCX, PDF and XLSX packs and pulls out questions, deadlines, scoring cues, word limits and missing document references. It keeps them in document order, so completeness is something you can check, not something you hope you remembered.
That order matters. Re-filing the ITT into your own headings can help you staff the work. It should not erase the buyer's structure. Reviewers need to walk the pack the way an evaluator will.
On How it works we show this as a simple upload: questionnaire, specification, pricing schedule. The point is not the filenames. It is that the response plan is built from the whole pack, not from the paragraph someone happened to copy.
Then the answer plan, then the draft
Once every item is visible, the team can decide what the work actually is: who owns the answer, what evidence exists, what is missing, and which sections need a technical or commercial review.
Only then is a first draft useful.
ProposAI drafts from your collateral library (policies, case studies, certifications, previous answers) and keeps the source attached. A reviewer should be able to open the passage behind a claim, not take the sentence on trust. Buyer-lens readiness checks then flag unsupported statements and coverage gaps before you export back into the working document.
That sequence is the product: extract → plan → source-linked draft → review. It is not a better essay generator.
Joel Ratnasothy, Managing Director at Interneuron, put the outcome in public terms: "ProposAI gets us to a credible first draft in minutes letting us focus on the content that makes us stand out." The draft is the start of judgement, not the end of it.
What to refuse
Do not paste the ITT into a general model and call it a response process.
Do not treat the specification as optional reading after the questionnaire is "done".
Do not polish answers while certificates, case studies or pricing assumptions are still missing. A fluent paragraph with no source is a risk, not progress.
And do not buy software that only stores a content library if your real pain is reconstructing this pack every time. A library without extraction and review still leaves the ITT on someone's desktop.
Start with one live ITT
If you have a pack in flight, that is the test.
Upload the questionnaire, specification and schedules. See whether the answer plan is in the buyer's order, with limits and scoring attached. Then draft one high-value section from approved material and inspect the sources.