Construction Prompt Library
Copy-paste prompts for estimators and PMs. Swap the {slots}, keep the rest.
House rules — paste at the top of any prompt that writes outreach
RULES: US register, plain and professional, no hyphens as dashes, no slang, no hype. US terms only — GC (not main contractor), ITB (not ITT), subcontractor appointed / bought out (not vendor onboarded), change order (not variation / compensation event), schedule (not programme), bid (not tender), $ (not £). Use ONLY facts I give you. Never invent wins, numbers, frameworks or projects. If a needed fact is missing, write [GAP: what to add]. The only proof point allowed: on a recent project, ONE senior estimator went from 2 bids to 6 in a month, same headcount, by automating the bid front end from ITB to buyout. State as "on a recent project", never "teams typically", never as US-wide. Always keep the honesty note — name where AI does NOT belong: pricing judgment, relationships and negotiation, final commercial decisions.
1. Research — turn public facts into a brief
1.1 Firm brief from pasted notes
From these public notes about a US contractor, give me a tight brief: sector (commercial building/infrastructure/civil), likely contract form (AIA or ConsensusDocs), size band, recent wins or backlog, project types, and 2 plausible preconstruction pressures given their pipeline. Mark anything you are inferring vs stated. Do not invent facts.
NOTES: {paste what you found}
1.2 Sector + contract classifier
This firm does {paste scope/sectors}. Tell me: primary sector, which contract form dominates (AIA/ConsensusDocs and why), and the 3 terms I must get right when writing to their estimators so I sound native.
2. Asset generation — the 1-pagers
2.1 Champion 1-pager (estimator / PE)
{house rules}
Write a one-page asset titled "Where AI fits in an estimator's week (and where it is noise)".
Audience: hands-on {estimator/PE}, no budget. Frame = makes your week easier.
Open with a line using a real fact about {company}: {fact}.
Cover 3-4 bid tasks where AI helps (bid doc review, sub inquiry drafting, bid leveling, scope-gap/spec checking), one line each in their workflow terms. Add the "where it is not worth it" line. Close by offering a short set of AI prompts they can use — no meeting ask.
Sector/contract: {commercial AIA / infrastructure ConsensusDocs / civil}.
2.2 Decision-maker 1-pager (director)
{house rules}
Write a one-page asset titled "Where AI pays off first in a mid-size GC (and where it is not worth it yet)".
Audience: VP of Preconstruction / Chief Estimator — thinks capacity and margin.
Blocks: 1) one-line thesis 2) why the bid front end pays off first 3) where it pays off (supply-chain solicitation & chasing, bid leveling, scope-gap/spec checking, bid doc review) 4) where it is not worth it yet 5) the 2-to-6 business case, framed as more bids per head not more hires 6) how to start without risk (pilot one package type) 7) a decision filter of 3-4 questions to spot their first candidate workflow.
Open with a real fact about {company}: {fact}. Close with a light offer of a short AI Opportunity Review.
Sector/contract: {...}.
2.3 Sector re-cut
Take this 1-pager and re-cut it for an {infrastructure / civil} contractor: swap AIA language for ConsensusDocs/DOT-spec terms, swap the example tasks for ones that fit {infrastructure/civil} bidding, keep structure and length identical.
ASSET: {paste}
3. The bid-workflow prompts
3.1 Bid doc summarizer
You are helping a US estimator. Read this bid package/ITB and give: scope in 5 bullets, key dates (bid due, start on site, contract period), contract form (AIA/ConsensusDocs + variant), unusual or onerous clauses, and any scope gaps or ambiguities to query. Flag what a subcontractor could exploit later.
BID PACKAGE: {paste}
3.2 Subcontractor RFQ drafter
Draft a subcontractor RFQ for the {trade} package on {project}: scope summary, what to price, drawings/specs to reference, return deadline, and the exact quote format so returns come back levelable. Short, clear, for a busy sub.
3.3 Bid leveling table
Here are {N} subcontractor quotes for the {trade} package: {paste}. Build a like-for-like table: price, inclusions, exclusions, qualifications. Flag where scope differs. Highlight the lowest COMPLIANT bid, not just the lowest number.
3.4 Scope-gap finder
Compare these two quotes for overlapping or missing scope: {A} / {B}. Tell me what falls between them (unpriced by either) and what is double-counted. This is where buyout leaks money.
3.5 Specification checker
Spec section: {paste}. Subcontractor quote: {paste}. Does the quote meet the spec? List every point where they price a lower standard, substitute a product, or stay silent on a requirement.
3.6 Bid query (RFI) drafter
From these bid ambiguities: {list}, draft numbered pre-bid RFIs to the owner/architect. Professional US tone, each specific enough to get a usable answer.
3.7 Risk register starter
From this bid package: {paste}, draft a first-pass bid risk register: commercial, schedule, scope, supply-chain risks. One-line mitigation or contingency flag each.
4. Personalisation — the messages
4.1 Connection note (under 300 chars)
{house rules}
Write a LinkedIn connection note under 300 characters to {first} at {company}. Open with this real hook: {fact}. One line on looking at where AI genuinely helps construction teams and where it does not. No pitch, no CTA. End "Worth connecting. Thanks."
4.2 Step 3 first message (after accept)
{house rules}
Write the first LinkedIn message after {first} accepts. 3 short beats: 1) one line referencing {fact} 2) the 2-to-6 proof, plainly 3) the report link offer + a question about whether bid-to-buyout eats their week. Human, not AI-sounding. No clause-stacked lists.
4.3 Follow-ups (value, not nudge)
{house rules}
Write two LinkedIn follow-ups for {first}, no reply yet. #1: a surprising finding (biggest drain is chasing supply-chain returns, not pricing) + a question on where their cycle slows. #2: a low-pressure close offering a short AI Opportunity Review. Different angles, not "did you see this".
4.4 Re-open a cold thread
{house rules}
{first} said they would read the report then went quiet. Write one short re-open: no chase, give a specific question about their workflow (returns vs leveling) to restart the conversation.
4.5 Register match
Rewrite this message to match the lead's style from their profile: {paste profile read}. If blunt/formal keep it tight and plain; if warm/casual soften slightly. Do not add slang or hype either way.
MESSAGE: {paste}
5. QC — catch AI tells and non-US slips
5.1 De-AI pass
Review this message as a US construction person would. Flag anything that reads AI-generated (clause-stacked lists, over-explaining, "stuck with me", mannered dashes) or British (main contractor, ITT, £, "whilst", "brilliant", "keen to"). Rewrite plain and human. Keep it short.
MESSAGE: {paste}
5.2 Fact audit
Check this asset against the facts I gave: {facts}. Flag any claim not supported by those facts, any invented win/number, and any [GAP] left unfilled. List only the problems.
ASSET: {paste}
How to use this library
- Save the house rules block once. Paste it into any Section 2 or 4 prompt.
- Run Section 1 first to get facts. Use them to fill {fact} everywhere else.
- Build each Section 3 prompt once, reuse it for every estimator.
- Run Section 5 before sending anything out.