AI tender documents: how AI reads a tender pack and what you still check

The tender pack lands: 40 PDFs, 3 addenda and a return date next Friday. Somewhere in there is the one clause that turns a good job into a bad one.

Yes, AI can read tender documents. It turns each file into text and searches it for the details that matter. Then it gives you a short summary, with the page each point came from.

What it can't do is take responsibility for your price. You still check the scope, the risks and anything it marks as unclear.

In short

• AI reads a tender pack by turning every document into searchable text, then pulling out what you ask for.
• It's good at finding dates, scope, exclusions, provisional sums and contract terms across lots of files.
• It can miss notes on drawings, misread bad scans and state things that aren't in the pack.
• Check anything that changes your price or your risk against the document itself.
• Don't upload a pack to a public AI tool unless you know where the files go and the tender allows it.

How can AI read a construction tender pack and find the important details?

AI reads a construction tender pack by turning each document into text it can search. It then pulls out the details you ask for and notes the page each one came from.

Here's what happens, step by step:

1. You give it the whole pack: drawings, specification, prelims, pricing document, contract terms and every addendum.
2. It turns each PDF into text. A good tool also runs scanned pages through text recognition, which reads the letters off the image.
3. It sorts the files by type and lists every one, so you can see what's there.
4. It searches each file for set things: dates, scope, exclusions, provisional sums, contract terms and risks.
5. It writes a summary, with the document and page for every point.
6. You check the points that matter against the page they came from.

The page references are the most important part. A summary you can't trace back is a summary you can't trust.

What is in a tender pack?

A tender pack is the set of documents a client or main contractor sends out so firms can price a job.

Most packs include these:

Prelims (preliminaries) are the costs of running the site, rather than doing the work itself. If the pack has a bill, our guide to how a bill of quantities is laid out explains each section.

Pre-construction information is a legal duty, not a nice extra. CDM 2015 is the set of health and safety rules for managing building work. Under it, the client must give this information as soon as is practicable to every contractor being considered for the job.

It covers the project, how it will be planned and managed, and the health and safety hazards, including how they'll be dealt with. It also includes anything in an existing health and safety file.

Northern Ireland has its own version: the Construction (Design and Management) Regulations (Northern Ireland) 2016.

What can AI pull out of a tender pack?

AI can pull out the facts you need to decide whether to bid and what to price. The most useful ones are below.

A provisional sum is an allowance for work that can't be fully priced yet, because it hasn't been designed. Retention is money the client holds back until the job's finished and any defects are fixed.

The best use of AI here is the boring cross-checking. Does the spec match the drawings? Does an addendum change the hours? Is an item excluded in one place and included in another?

A person reading 40 files at 10pm misses those. That's where AI earns its keep.

What does AI get wrong when it reads tender documents?

AI can miss things, misread them, or state something with confidence that isn't in the pack at all. The UK's National Cyber Security Centre (NCSC) warns that AI language tools can get things wrong and "hallucinate" facts that aren't true.

The usual slip-ups in a tender pack:

• Notes on drawings. A note in the corner of a drawing can change the spec. Some AI tools read the text in a PDF well but struggle with drawings.
• Bad scans. A wonky scanned letter gives garbled text, so a key addendum can get missed or misread.
• Cross-references. "As clause 4.2 of the prelims" only helps if it follows the link to the right document.
• Clashes. When the spec says one thing and the drawing another, it may pick one and not tell you there was a choice.
• What isn't there. If a sheet on the drawing register wasn't sent, it may not notice.
• Made-up detail. Ask it something the pack doesn't answer, and it may fill the gap with a likely-sounding guess.

So treat the summary as a map of where to look, not as the answer.

What do you still have to check yourself?

You still check anything that changes your price or your risk. That means the scope, the quantities, the contract terms and every point the AI marks as unclear.

Run down this list before you price up:

Quantities are their own job. If you're measuring off the drawings, see how to measure PDF drawings for a tender. Before you trust an AI-built price, run through the checks to make before using AI for tender pricing.

Is it safe to put tender documents into AI?

It's safe only if you know where the documents go and the tender lets you share them. Tender packs often hold client details and commercial information that isn't public.

The NCSC's 2023 guidance says what you type into a public AI language tool is visible to the company that runs it. It's stored, and will almost certainly be used to develop the service at some point.

Its advice is not to include sensitive information in queries to public AI language tools. So before you upload a pack:

• read the tender instructions for any confidentiality rules;
• check the AI tool's privacy terms, including whether it learns from your files;
• use a business tool with clear data terms;
• leave out anything you're not allowed to share.

What does a tender pack summary look like?

A good tender pack summary fits on one page and tells you in 5 minutes whether to bid and what to watch. Use this layout for your own notes, or to check an AI tender documents summary line by line.

Tender pack summary
Job and client:
Who sent it (client or main contractor):
Return date, time and how to submit:
Questions deadline:
Site visit:
Documents in the pack (with revisions):
Scope in one paragraph:
Exclusions and work by others:
Provisional sums:
Working hours and access:
Payment terms and retention:
Late-finish damages:
Clashes between documents:
Questions to send:
Bid or no bid, and why:

Here's a worked example for a sparky pricing the electrical package on a school classroom refurb. The job and figures are made up to show the method.

In this worked example, the AI summary got the exclusions and the provisional sum right. It missed addendum 1, because it was a scanned letter.

That's the line that changes the labour price most. Shorter working days mean more days on site.

The questions to send would be:

• Which light fitting is right?
• Who connects the fire alarm changes to the panel?
• Is there an asbestos survey for the ceiling void?

Which AI tools can read a tender pack?

General AI assistants, tender management software and some PDF tools can all summarise a tender pack. Here's what each kind does, with real examples.

Whichever you use, check every point against the page, and keep confidential packs out of free personal accounts.

Where Nexley fits

Nexley prices your tenders and runs your office, all on WhatsApp. It's AI software for UK tradespeople and small construction firms that prices tenders and measures take-offs. It also writes quotes and runs the back office through WhatsApp, linked to Xero, QuickBooks and Sage.

Nexley reads the tender pack you were sent: the drawings, specification, preliminaries and the pricing document. Quantities come off the PDFs, itemised so you can check any line against the drawing it came from.

Nexley flags provisional sums where the information is incomplete, and prices the bill on your own rates. You check the finished bill before anything is submitted. See how AI tendering software works, or try it free for 7 days.

Common questions

What are tender documents in construction?

Tender documents are the set of files a client or main contractor sends out so firms can price a job. They usually include the invitation to tender, drawings, a specification, prelims, a pricing document and the contract terms.

Any addenda (changes issued during the tender) are part of the pack too.

Can ChatGPT read a tender pack?

Yes, ChatGPT can summarise PDF documents you upload, so it can help with a tender pack. As of October 2026, OpenAI's help pages say paid plans can read the text and images in a PDF you attach to a chat. Not every free account can, and PDFs saved as project files are read as text only.

So on a free account, or in a project's files, a drawing or a scanned letter may not be read at all. Check every point against the page, and don't upload a confidential pack to a free personal account.

Can I trust an AI tender documents summary?

Trust it as a guide to where to look, not as the final word. A good summary gives a page reference for every point, so you can check it fast.

Check every point that changes your price or risk: the scope, exclusions, provisional sums, working hours and contract terms. If a point has no reference, treat it as unconfirmed.

What should I do if two tender documents disagree?

Send a question to whoever issued the tender, before the questions deadline. Say which documents clash and ask which one is right.

If there's no time for an answer, price the one you think is meant and state it as an assumption in your tender. Never quietly pick the cheaper option and hope.

Can AI tell me if a tender is worth bidding for?

AI can lay out the facts that help you decide: the scope, the deadline, the contract terms and the risks. But only you know your workload, your margins and whether you want that client.

A one-page summary makes the call quicker. The call is still yours.

Sources

This page is general guidance, not legal advice. Always read the full tender pack and contract yourself, and get advice on unusual contract terms before you sign.

• The Construction (Design and Management) Regulations 2015, regulation 4: client duties (legislation.gov.uk) (legislation.gov.uk)
• The Construction (Design and Management) Regulations 2015, regulation 2: what pre-construction information means (legislation.gov.uk) (legislation.gov.uk)
• The Construction (Design and Management) Regulations (Northern Ireland) 2016 (legislation.gov.uk) (legislation.gov.uk)
• ChatGPT and large language models: what's the risk? (National Cyber Security Centre) (ncsc.gov.uk)