The problem
Tender packs are long, and tired builders skim. If you price 3 a month and each takes 3 hours to read properly, that's 9 hours; a missed exclusion or provisional sum can turn a priced job into a loss.
What the AI does
You paste in the text of the specification or tender and ask an AI assistant for a summary: scope, exclusions, programme dates, items that need allowances and questions for the client. You still read the document, but with a checklist in front of you.
A worked example
A four-person building firm in Exeter prices about 3 extension and refurbishment tenders a month. The owner pastes in each specification and asks an AI assistant (ChatGPT, Claude or Gemini) for a risk list before he starts pricing.
Before
About 3 hours reading each pack, with details sometimes missed
After
About 1.5 hours per pack, with a risk list built first
What it's worth
About 4.5 hours a month saved (1.5 hours on 3 tenders), worth around £225 if your time is worth £50 an hour.
How to set it up
- Copy the relevant sections of the tender or specification into ChatGPT, Claude or Gemini as text.
- Ask for scope, exclusions, key dates, items needing allowances and questions to put to the client.
- Read the full document yourself against the summary, especially the numbers.
- Turn its questions into one email to the client before you price.
- Keep the summary with the job file.
What goes wrong
- Long documents may be cut off by usage limits. Split the pack and check you pasted everything.
- It will not catch what is in drawings, so always review those yourself.
- Don't paste confidential client or commercial documents into any AI tool without checking the client allows it.
- Never price from the summary alone.