Delays don't cost you when they happen. They cost you when the notice goes out late - or never. Most contracts have strict time bars on extension of time claims. A rough site diary entry won't protect you. A proper notice will. Here's a prompt that turns your delay notes into a draft EOT notice in a few minutes: Draft a formal EXTENSION OF TIME notice for an Australian construction project. Project: [name and number] Principal / Superintendent: [names or placeholders] Date delay event began: [date] Delay event: [e.g. late design information, wet weather, principal-directed change] Facts: - [what happened] - [who was notified and when - include site instruction or email refs] - [which critical-path work was affected] Delay to date: [number of days] Extension claimed: [days - your number, don't let AI guess] Contract clause relied on: [clause number, or write "unsure" and mark it [VERIFY]] Format: Formal letter. 1 page. Factual, no blame language. Sign off: [your name, role, business, date]. Two rules before you send: • Check the notice period in your contract. Time bars are unforgiving and AI doesn't know yours. • Every clause reference gets marked [VERIFY] and checked against the actual contract. Try it on a delay from the last month that never got a proper notice. Tell us in the comments how many days it turned out to be.

One of the most underrated things AI does in construction isn't writing - it's reading. When a Sunshine Coast PM pasted 14 rough notes from a tender walkover into ChatGPT, AI returned not just clean meeting notes - it flagged: "The principal mentioned cleaning up the existing carpark surface but didn't specify whether this is included in scope or a variation. Recommend confirming." He emailed the principal that afternoon. The principal confirmed it was a variation. The carpark cleanup ended up being a $25k extra. Without AI's nudge, the PM would have priced it as included and found out 4 months later. This is what the Day 13 prompt does. Paste your rough meeting notes in. AI structures them and flags the things that could become disputes. That's the one that pays for this course in a single meeting.

Accounts: Here's the one that gives you your Mondays back. AI writes firm, professional follow-up emails that chase payment without burning the relationship - and faster than you can find the invoice number. Draft a professional overdue payment follow-up email for an Australian construction business. Invoice number: [number] Invoice date: [date] Invoice amount: $[amount] Due date: [date] Days overdue: [number] Client name/company: [Client - use a placeholder if preferred] Previous follow-ups: [none / one call on [date] / two emails already sent] Tone: [firm but professional - I want to preserve the relationship / very firm - this is the final notice] Keep it under 150 words. Direct. No apologies for chasing what we're owed. Tip from Day 6: if this is the second or third follow-up, ask AI to add a line referencing the security of payment legislation - it gets a response faster than anything else.

We've covered variations, reports, meeting notes, tenders, emails, debtors, SWMS, and more across the 20 lessons. But there's always something that feels hard to hand off - either because it's too nuanced, too risky, or you just haven't tried it yet. What's yours? Drop it in the comments. I'll answer every one - and if it's something we should build a prompt for, I'll do exactly that.

The 4 AI Rules That Keep You Out of Trouble From Day 4 of the course - the four things to keep front of mind every time you use AI on a real job. Worth re-reading every month. 1. AI drafts. You approve. Never send anything AI wrote without reading every word first. The output goes out with your name on it. 2. Anonymise everything. Never paste real client names, ABNs, contract values, or sensitive project details into a free AI tool. Use [Client Name], [Builder], [Amount] as placeholders. 3. Verify facts and specs. AI's knowledge has a cutoff date. NCC updates, WHS regulations, and material costs change. Always cross-check against current standards before relying on the output. 4. Verify any clause references. AI will confidently cite contract clauses that don't exist, or get the clause number wrong. Mark every AI-generated clause reference as [VERIFY] and check it before the letter goes out. Four rules. Apply them on every job. They're not there to slow you down - they're there to make sure AI actually helps you instead of creating a problem.