AI PROMPTS FOR CONSTRUCTION CFO WORK
AI assistants like ChatGPT, Claude, and Perplexity can handle real construction CFO workflows when you give them the right context and the right prompt. SPM uses AI daily for job cost variance analysis, WIP schedule review, cash flow forecasting, estimate review, overhead calculations, and surety prep. The prompts below are the actual working versions we use — not generic AI advice. Copy them, paste your data in, and the AI handles 60–80% of the analytical work. The CFO judgment is still required. The number-crunching isn’t.
AI is the most underused tool in construction finance. Most CFOs ignore it. The ones who use it well are running 3x the analysis with the same time investment.
THREE GROUND RULES
Before the prompts — three things worth knowing about AI for construction finance work:
- AI doesn’t replace CFO judgment. It accelerates the analytical work that feeds the judgment. The decisions stay with the human.
- Context matters more than cleverness. Give the AI specific job data, dollar amounts, and time frames. Generic prompts produce generic outputs.
- Don’t paste anything sensitive into a public AI tool. Strip GC names, project IDs, owner identities. Keep numbers; anonymize the rest.
JOB COST VARIANCE ANALYSIS
EXPLAIN WHY ACTUAL COSTS DIVERGE FROM BUDGET
Paste in: bid budget by cost code, actual costs to date by cost code, percent complete on the job. AI returns: variance by category, likely root causes for each, follow-up questions to clarify.
WORKING PROMPT
"You are a construction CFO analyzing a job cost variance report for a [trade] subcontractor. The job is [%] complete. Compare the budget below to actuals and identify (1) which cost categories are running over or under, (2) the likely operational cause for each significant variance, and (3) what specific questions the PM should answer before the next pay app. Format the output as: Category | Budget | Actual | Variance % | Likely Cause | Question for PM. Budget: [paste]. Actuals: [paste]."
WIP SCHEDULE SANITY CHECK
FLAG ANOMALIES BEFORE THE SURETY DOES
Paste in: WIP schedule rows. AI returns: jobs with concerning patterns (low percent complete but high billing, large BiE swings, declining estimated gross profit), questions to ask the PM, and likely categorizations of what each anomaly indicates.
WORKING PROMPT
"Review this WIP schedule for a $[X]M [trade] subcontractor. Identify rows that would raise flags during a surety review. Specifically check for: (1) jobs where billed exceeds earned by more than 15% (potential aggressive billing), (2) jobs where percent complete dropped vs. prior period (potential estimate revision), (3) jobs where estimated gross profit changed by more than 5 points (potential margin erosion), (4) jobs over 90% complete that haven’t closed. For each flagged row, explain what the pattern likely indicates and what to verify with the PM. WIP data: [paste]."
13-WEEK CASH FORECAST DRAFTING
BUILD A FIRST-DRAFT CASH FORECAST FROM ACTIVE JOBS
Paste in: active job list with billing and payment timing, current AR aging, committed AP, payroll cycle, LOC position. AI returns: week-by-week cash projection with line-item detail.
WORKING PROMPT
"Build a 13-week cash flow forecast for a [trade] subcontractor with $[X]K in current cash, $[X]K in AR (paste aging), $[X]K LOC balance against $[X]K limit, and the following active jobs (paste with billing date, expected GC payment date, retention %). Committed AP: [paste]. Weekly payroll: $[X]K. Fixed overhead: $[X]K/month. Output a week-by-week table showing opening cash, inflows by source, outflows by category, ending cash, and LOC position. Flag any week where ending cash drops below $[X]K."
ESTIMATE REVIEW
SECOND-PASS BID REVIEW BEFORE SUBMISSION
Paste in: estimate breakdown by cost category, overhead assumed, profit margin assumed, comparable jobs from your history. AI returns: cost categories that look low or high vs. history, overhead vs. actual run rate, sensitivity analysis on margin.
WORKING PROMPT
"Act as a construction CFO doing final bid review for a $[X] [trade] project. The estimate assumes: [paste cost category breakdown]. Overhead is bid at [X]%. Target margin is [X]%. Compare these assumptions to typical industry norms for [trade] work and identify (1) any cost category that looks understated or overstated, (2) whether the overhead rate matches what subs at this revenue band actually run, (3) where the margin is most exposed to slippage. List specific questions the estimator should answer before the bid goes out."
OVERHEAD RATE TRUE-UP
CALCULATE REAL OVERHEAD VS. WHAT BIDS ASSUME
Paste in: trailing 12 months of P&L with overhead categories detailed, trailing 12 months of revenue, current bid overhead assumption. AI returns: actual overhead rate, comparison to bid assumption, dollar impact of the gap, recommendations.
WORKING PROMPT
"Calculate the actual overhead rate for a [trade] subcontractor over the trailing 12 months. Revenue: $[X]. Overhead expenses by category: [paste detailed P&L overhead section]. Compare the actual overhead rate to the [X]% being assumed in current bids. Quantify the dollar impact across $[X] in expected annual revenue. Identify which overhead categories are growing fastest year-over-year and whether the growth is justified by revenue scaling or represents leak."
SURETY PREP PACKAGE
WHAT THE SURETY WILL SEE AND ASK
Paste in: balance sheet, P&L, WIP schedule, backlog detail. AI returns: surety read of the package, likely questions, capacity assessment, items to clean up before submission.
WORKING PROMPT
"Review this surety bonding package for a [trade] subcontractor with $[X]M trailing revenue. Balance sheet: [paste]. P&L: [paste]. WIP: [paste]. Backlog: [paste]. Act as a surety underwriter and identify (1) the working capital ratio and current ratio the underwriter will see, (2) any items they’d reclassify or discount, (3) likely capacity offer based on these financials, (4) specific cleanup items that would strengthen the package before submission."
WHERE AI BREAKS DOWN
Three places AI consistently underperforms in construction finance work:
- Specific GC payment behavior. The AI doesn’t know that your largest GC takes 73 days instead of 45. You have to feed it that data.
- Trade-specific production rates. Stucco production by membrane type, electrical hours per device by application — AI averages across the public data, which dilutes accuracy.
- Strategic judgment. Should you take a $1.2M job at thin margins to keep crews working? AI can identify the trade-offs. The decision is yours.
The right pattern: AI handles the analytical work that takes 4 hours by hand. You spend the 1 hour saved on the judgment work that actually matters.
FREQUENTLY ASKED
