Paperwork can eat five to ten hours per week in a trade business. That time is missing on site. For many businesses, 9:30 p.m. at the desk instead of time off is normal.
We tested which AI prompts actually reduce that office work. Every prompt ran multiple times against current models, using realistic business data. These are the three that passed, including the pitfalls most prompt lists leave out.
How do bullet points become a finished proposal?
The classic daily-business problem: the customer has been waiting three days for a number, and the proposal gets written in the evening. With the right prompt, that becomes a 10-minute job. You provide notes about service, materials, and hours; AI provides structure, totals, and the cover note.
We make this prompt completely free. Copy it directly into ChatGPT or Claude:
From the prompt lab · Create a professional proposal from bullet points
Tested ✓Turns loose notes about the job, materials, and hours into a structured proposal with line items, terms, and a professional tone.
The prompt
Take the role of an experienced estimator in a trade business with 20 years of experience. Create a complete, customer-ready proposal from my bullet points. Context: I own a trade business and have little time for office work. The proposal should sound professional but not stiff, like a reliable local contractor. Work step by step: first structure the line items, then calculate the totals, then write the cover note and terms. If quantities or prices are missing or unclear, ask targeted follow-up questions instead of inventing values. # MY DETAILS: Trade and service: [TRADE AND SHORT DESCRIPTION OF THE SERVICE] Materials with quantities: [MATERIAL LIST WITH QUANTITIES] Working time and hourly rate: [ESTIMATED HOURS AND HOURLY RATE] Special notes: [SPECIAL NOTES, E.G. TRAVEL, DISPOSAL] IMPORTANT: Return the proposal with: cover note (max. 4 sentences), line-item table (item, description, quantity, unit price, total price), net total, VAT, gross total, payment terms, and validity period.
Real output (shortened)
Proposal no. 2026-0142 Dear Ms. Example, thank you for your inquiry. We are happy to offer the bathroom renovation as discussed: Item 1: Removal of existing fixtures (8 hrs at $65) ... $520.00 Item 2: Floor/wall tiling, 24 m2 ... $2,880.00 [shortened]
From our test
Reliably produces a clean line-item structure. The model calculates prices correctly as long as quantities are clear. For approximate inputs, the prompt asks targeted follow-up questions instead of guessing.
Pitfalls from the test
- Without your hourly rate in the input, the model invents plausible market rates. Always provide your own.
- Have legal clauses such as warranty and payment terms checked before sending.
Claude Opus 4.8 · tested on Jun 9, 2026 · 4 variants tested
How do you reply confidently to complaints?
Complaints are the moment when customer relationships either tip or become stronger. The difference is the first reply. It has to show understanding without admitting fault. Generic prompts rarely hit that line.
From the prompt lab · Reply confidently to customer complaints
Tested ✓Analyzes the complaint, identifies the emotional core, and drafts a reply that shows understanding, clarifies the facts, and offers a concrete next step.
The prompt
Take the role of an experienced tradesperson who has maintained customer relationships and handled difficult conversations for 20 years. I received a complaint and need a reply that: 1) shows real understanding without admitting fault, 2) calmly explains the situation from my perspective, 3) offers a concrete next step such as an appointment, inspection, or call-back. First briefly analyze what is really bothering the customer emotionally: money, time, respect, or something else? Then write the reply in my name, ready to send, without empty phrases such as "we regret any inconvenience". # MY DETAILS: Customer complaint: [PASTE THE COMPLAINT] What happened from my perspective: [YOUR VIEW IN 2-3 SENTENCES] What I can offer: [YOUR OFFER, E.G. FOLLOW-UP APPOINTMENT]
Real output (shortened)
Dear Mr. Example, I can understand your frustration about the delay. Thank you for raising it directly. [shortened]
From our test
The prompt's strength is separating understanding from admitting fault. That legal and practical distinction is where generic prompts usually fail.
Pitfalls from the test
- For legitimate defects, do not accept promises automatically. Goodwill decisions remain a management decision.
Claude Opus 4.8 · tested on Jun 9, 2026 · 3 variants tested
How do you become visible locally without a marketing budget?
The best job-site photos do nothing if they stay on your phone. This prompt turns a short project description into three ready-to-post variants: factual, personal, and lightly playful, including local hashtags.
From the prompt lab · Turn job-site photos into social media posts
Tested ✓Creates three post variants from a short project description: factual, personal, and playful, including local hashtag suggestions.
The prompt
Take the role of a social media copywriter specialized in trade businesses who knows that authenticity wins more work than polished advertising. Create three Instagram/Facebook post variants from my project description: 1) factual and informative, 2) personal with a behind-the-scenes angle, 3) with a light touch of humor. Each variant: max. 80 words, hook in the first sentence, call to action at the end, 5 hashtag suggestions with local relevance. No customer names, no addresses, no invented details. Use only what I provide. # MY DETAILS: Project in 2-3 sentences: [PROJECT DESCRIPTION] Trade: [YOUR TRADE] Region: [CITY/REGION]
Real output (shortened)
Variant "personal": Before: a bathroom from 1987. After: a small wellness space. Three weeks, a coordinated team, and a customer who now showers longer than they admit. [shortened]
From our test
The humorous variant usually needs a quick edit because humor is local and trade-specific. The factual and personal variants are usable right away.
Pitfalls from the test
- Do not post customer names or recognizable addresses. The prompt reminds you, but it does not replace customer approval.
Claude Opus 4.8 · tested on Jun 9, 2026 · 3 variants tested
Conclusion: tool yes, autopilot no
AI structures and drafts. The decision about what goes out stays with the business.
All three prompts have one thing in common in testing: they work because they keep you in the driver's seat. Prices, goodwill, and approvals stay with you; the model handles the writing work.
If you want more than these three examples: the trades package contains every tested prompt for the niche, from reminders to review replies, each with real output and documented pitfalls.