Home · Sales & leads · AI automations
AI for proposals, quotes and estimates: 9 real small-business examples
Updated October 18, 2026 · 9 sourced examples · no paid placement
Short answer: Small firms and contractors use AI to turn what they already have (sales-call recordings, walkthrough notes, past jobs, a client's planning data) into a first-draft proposal, estimate or bid that a person checks before it goes out. In these 9 examples the AI is Claude, ChatGPT, an AI app builder or estimating software, and the reported wins are mostly time: one remodeler says first proposals dropped from 60 to 90 minutes to 10 to 15. One owner tried AI on detailed estimates and went back to writing them himself. The results are what each business reported, not promises.
Below are 9 documented examples from remodelers, a masonry contractor, a landscaper, a collision repair group, a financial planning firm and an insurance agency, told by owners and staff in their own writing, on podcasts and in the trade press: who did it, what the AI drafts, what they reported and where the story came from. After that: when a person should still write the estimate, what to watch out for and how to try it this week.
Other industries too: see real AI automations by task for any small business.
What these 9 businesses actually did
- Remodeling proposals from recorded sales calls: Element Design Build (Ohio) (press, via Qualified Remodeler: 2026 Fred Case Award Winner, Control and Capacity (Patrick O'Toole)). He records first calls and in-home consultations with Fireflies. Claude summarizes each one, identifies action items and helps prepare a JobTread proposal from established templates and historical company data. A Claude project for final proposals pulls together the decisions, selections and cost history stored in JobTread into a final scope of work and budget. He can also ask Claude to build a proposed schedule from a start date using the company's actual task durations. Reported: He says initial proposals that took 60 to 90 minutes now take 10 to 15, and a $150,000 kitchen estimate that took two days can often be done in about three hours. Element grew from $3.74 million in 2024 to $3.92 million in 2025 with the same headcount, which he largely credits to the AI operations layer. "I read every single word," he says, and he tests spreadsheet instructions on one row before applying them. Full case
- Ballpark remodel budgets from past jobs: MCG Homes (Florida) (press, via Qualified Remodeler: 2026 Fred Case Award Finalist — Mathew Meskimen / MCG Homes). MCG chose Anthropic's Claude as its primary business AI and built shared Projects for estimating, sales qualification, HR, SOP search and construction standards so every employee uses the same instructions and company data. Meskimen built a ballpark-estimating tool that lets a salesperson describe scope and property, compare against MCG historical jobs and market data, and draft a budget with MCG cost codes, allowances, labor, subcontractor and material categories plus risk contingencies. JobTread remains the project-management system of record. Reported: Self-reported: work that took about eight hours to compile took the AI tool about 30 minutes in a test, and the result came within about five percent of a staff estimate. Company remodel revenue was $6.03M in 2025 (up from $4.61M in 2024); that is company performance context, not claimed as an AI-only result. Full case
- Custom client proposals as a web page instead of a PDF: Iconoclastic Capital (financial planning, New York) (owner-told, via Kitces Financial Advisor Success podcast ep. 477: Christopher Haigh, Iconoclastic Capital). After the intro and analysis calls, he feeds the meeting transcripts and planning data (for example from RightCapital) into a proposal template he built in v0 (Vercel's AI app builder). It drafts the client's priorities and the scope of work as an interactive web page on a custom link with the client's first name, with short quizzes built in; he tweaks it and sends the link, and corrections show up without re-sending. Before this he sent templated PDFs with a Loom walkthrough. An AI notetaker (Hazel) transcribes meetings and syncs case notes to the CRM (Wealthbox). On the phones, an AI receptionist added to RingCentral is prompted with what callers ask and who they're trying to reach, and routes calls to the right person. Reported: Self-reported: proposals used to take three to four hours each even with templates ('my burnout ratio was as well'); now the template drafts the priorities and he only tweaks them (no new time given). He says the conversion rate was already high. The AI receptionist routes calls that used to end up as missed calls and voicemails 'a lot more effectively' (no figures). Full case
- Proposals and coverage comparisons: Evergreen Insurance (Washington) (press, via IndependentAgent.com (Big "I" agency AI survey): How Evergreen Insurance Uses ChatGPT to Improve Efficiency Across the Agency). The leadership team uses ChatGPT for Business to work faster on client communications, proposals and coverage comparisons, to help build out a commercial lines niche, and to write the agency's standards, processes, SOPs and templates. Villafranca also uses it as a thinking partner. Reported: Three people in the office, the leadership team, use it extensively. Villafranca says getting the rest of the team to use it, and learning how to use it best, have been the hard parts. No time or dollar figures were given. Full case
- Bid requests from site walkthrough notes: North Shore Masonry (Illinois) (press, via Qualified Remodeler: The new crew member, how AI agents are reshaping remodeling companies). AI agents built on large language models (ChatGPT, Claude and Gemini for different jobs) are connected through APIs to the company's Gmail, CRM and scheduling. Field staff 'dump a transcript of their walkaround' plus photos, and the agent condenses it into a bid request in the company's own format. Permissions are tight: O'Hara's OpenClaw assistant can only read and draft email, never send it, and sales and admin staff get narrower access than leadership. Drafts are written in his voice after reading thousands of his emails. Reported: O'Hara says last year was a record and Q1 'just got stamped as a record,' while being careful not to credit AI alone; the main win is less admin friction and more consistent estimates. Internal rule: 'Put an eyeball on everything before it goes out.' Full case
- One set of selections feeding every proposal section: KBF Design Gallery (Florida) (press, via Custom Builder: Why This Remodeler Stopped Treating AI Like a Search Engine (Catherine Sweeney)). Otter AI records every client, internal and sales meeting. She aggregated her years of experience into a sales playbook uploaded to Claude Cowork, along with each client's contract, conversations, emails and project data, as a sales intelligence system. A "Single Source of Truth" contract tool built with Claude Cowork takes the material details in one place and parses them into the proposal's sections; when Otter transcripts are uploaded after a meeting, it flags the team on which decisions are still pending and which were made. Reported: She says it has helped her step back from day-to-day sales management into a leadership role. She calls team reception mixed, and says the biggest change was simply recording client meetings: "just having a searchable record of what was discussed is incredibly powerful." No time figures are given. Full case
- Estimates and client messages, written and translated: Gardenboro Landscaping (North Carolina) (owner-told, via Gardenboro Landscaping blog). The owner uses an AI assistant he calls 'Willy' to plan routes and track projects, write and translate client messages and estimates, sketch design ideas, and draft blog and social posts. Reported: Does the work of four extra employees and improved Google visibility (self-reported). Full case
- Photo pre-estimates on collision leads: Body by Cochran (Pennsylvania and Ohio) (press, via Autobody News: $75M Dealership Collision Center MSO Expanding AI Call Center, Estimating Effort (Body by Cochran / David Black)). A 24/7 virtual CSR from BodyShop Booster gathers customer and vehicle details, keeps a record of every call, and starts work within minutes of a shop getting a call whether or not it is open. Tied in is Tractable AI for photo assessments / pre-estimates. A text goes to inquirers with an introductory video; an AI-produced virtual assistant named Amy walks customers through the repair lifecycle. If the first follow-up does not book an appointment, the system flags it to a human. Reported: Per Autobody News / Black: AI contacts 95% of assignments (was below 70%). Capture rate of people contacted nears 90%. Tractable pre-estimates are useful for anything under 25 hours of work and write a slightly better estimate in 90 seconds instead of 30 minutes. BodyShop Booster's Ryan Taylor says 78% of customers go with "first responders." Four of 11 Cochran collision centers were live at the time of the article, with all 11 planned by year-end. No dollar revenue figure is published for the AI itself. Full case
- Tried AI on detailed estimates, went back to writing them himself: Inland Power Wash (California) (owner-told, via Leaders Perception: Why Inland Power Wash trusts humans for estimates and AI for replies (Eric Zavala)). He has handed initial customer communication to AI: responding to common inquiries about services and scheduling free estimates. He tried AI on writing detailed project estimates and surface assessments and went back to doing those himself. The interview doesn't name the tool. Reported: He says AI handles routine questions quickly so he can focus on assessing properties and doing quality work. AI couldn't capture the nuance of judging rust stains, efflorescence and hard water deposits, so estimates and surface assessments stay with him. No figures are given. Full case
These are each company's own results with their own tools and teams, as reported by the source. Vendor case studies are written by the company selling the tool. None of them are promises.
What can AI write for a proposal or estimate?
From a recorded sales call. Vlatko Demrovski records first calls and in-home consultations at Element Design Build with Fireflies; Claude summarizes each one and helps prepare a JobTread proposal from the company's templates and historical data. He says initial proposals that took 60 to 90 minutes now take 10 to 15, and a $150,000 kitchen estimate that took two days can often be done in about three hours. Christopher Haigh of Iconoclastic Capital feeds meeting transcripts and planning data into a proposal template he built in v0, which drafts the client's priorities and scope of work as an interactive web page. Before that, each custom proposal took him three to four hours even with templates.
From past jobs. Mathew Meskimen built a ballpark-estimating tool in a shared Claude project at MCG Homes: a salesperson describes the scope and the property, and it drafts a budget with MCG's cost codes, allowances, labor, subcontractor and material categories, compared against past MCG jobs. In a test, work that took about eight hours to compile took about 30 minutes and came within about five percent of a staff estimate.
From site notes and photos. At North Shore Masonry, field staff dump a transcript of their walkaround plus photos, and an AI agent condenses it into a bid request in the company's own format; the reported win is less admin friction and more consistent estimates. Body by Cochran's collision shops use Tractable for photo pre-estimates, which Autobody News reports write a slightly better estimate in 90 seconds instead of 30 minutes on jobs under 25 hours of work.
Proposals at professional firms. Evergreen Insurance's leadership team uses ChatGPT for Business for proposals and coverage comparisons as well as client emails. At KBF Design Gallery, a contract tool Ashley Sheaffer built with Claude Cowork takes the material details from one place and parses them into the proposal's sections. The problem she was solving: a change as simple as a countertop or backsplash has to be translated in six different places.
Small crews too. Gardenboro Landscaping's owner uses an AI assistant he calls Willy to write and translate estimates and client messages, along with routes and marketing. Vendor tools can also draft quotes: in a Kutlerri.ai case study, its catering sales agent drafts quotes from RASA's menu for the restaurant to approve. That story is written by the company selling the tool.
When should a person still write the estimate?
When judging the job is the hard part. Dennis Vong of Inland Power Wash tried AI on detailed project estimates and surface assessments and went back to doing them himself: it couldn't capture the nuance of judging rust stains, efflorescence and hard water deposits. He kept AI for routine customer questions and for scheduling free estimates.
And a person checks every draft. Demrovski says "I read every single word." North Shore Masonry's COO, Zac O'Hara, has a rule: "Put an eyeball on everything before it goes out," and his AI assistant can read and draft email but never send it. Haigh tweaks each proposal page before he sends the link.
What to watch out for
Your own numbers come first. The biggest time savings here lean on data each business already had: Element's selections and cost history in JobTread, MCG's past jobs and cost codes, Haigh's planning data. If your past jobs and prices aren't written down anywhere, start there.
A test is not a track record. MCG's eight-hours-to-30-minutes figure comes from a test, and Body by Cochran uses pre-estimates only for smaller jobs. Element grew revenue with the same headcount, which Demrovski largely credits to the AI operations layer; O'Hara is careful not to credit AI alone for its record year.
Getting the team on board. At Evergreen Insurance, Patrick Villafranca says getting the rest of the team to use ChatGPT has been the hard part, and Sheaffer calls team reception at KBF mixed.
Tools named in these examples
| Tool | Starting price (base plan) | Checked |
|---|---|---|
| Claude | $17/mo billed annually ($20 monthly) | 2026-10-15 |
| ChatGPT | Free $0; Go $8/mo; Plus $20/mo; Pro $100/mo | 2026-10-17 |
| Fireflies.ai | $10/seat/mo billed annually ($18 monthly); free plan available | 2026-10-15 |
| Otter.ai | Pro $8.33/user/mo billed annually ($16.99 monthly); free Basic plan with 300 min/mo | 2026-10-16 |
Starting prices from each vendor's pricing page on the date shown. AI phone and texting features can be priced separately from the base plan, so check before you buy. v0 by Vercel, Claude Cowork, JobTread, Tractable and Kutlerri.ai are also named in these examples; we haven't price-checked them.
How to start this week
- Pick one document you write over and over: first proposals, ballpark estimates, coverage comparisons or bid requests.
- Collect three or four past ones you were happy with, plus what went into them (call notes, measurements, your price list). That is your template and your price memory.
- After the next sales call or walkthrough, paste your notes and one past example into ChatGPT or Claude and ask for a first draft in the same format. Leave client names and personal details out.
- Check every line, price and quantity yourself before it goes out, and keep the parts that need judgment (site conditions, coverage advice) with a person.
- Once the drafts are reliably close, make it routine: a saved project or prompt, or a template in the software you already quote from.
Turn recorded sales calls into remodeling proposals in 10 to 15 minutes with Fireflies and Claude
He records first calls and in-home consultations with Fireflies.
Claude
Fireflies.aiJJobTreadBuild shared Claude projects for estimating and sales so a ballpark remodel budget takes minutes instead of a full day
MCG chose Anthropic's Claude as its primary business AI and built shared Projects for estimating, sales qualification, HR, SOP search and…
ClaudeJJobTreadTurn each prospect call into a branded, interactive proposal page with AI instead of a static PDF
After the intro and analysis calls, he feeds the meeting transcripts and planning data (for example from RightCapital) into a proposal te…
LoomVBv0 by VercelHHazelUse ChatGPT for Business for client emails, proposals, coverage comparisons and agency templates
The leadership team uses ChatGPT for Business to work faster on client communications, proposals and coverage comparisons, to help build …
ChatGPTTurn walkthrough transcripts into bid requests and draft client emails with an AI agent that can only draft, never send
AI agents built on large language models (ChatGPT, Claude and Gemini for different jobs) are connected through APIs to the company's Gmai…
ChatGPTClaude
Google GeminiRecord every remodel meeting with Otter and have Claude flag which client decisions are made or still pending
Otter AI records every client, internal and sales meeting.
Otter.aiClaudeCCClaude CoworkUse an AI assistant for a landscaping company's estimates, messages and marketing
The owner uses an AI assistant he calls 'Willy' to plan routes and track projects, write and translate client messages and estimates, ske…
ChatGPTPut an AI call center and photo pre-estimate on every collision lead so weekend inquiries don't sit for days
A 24/7 virtual CSR from BodyShop Booster gathers customer and vehicle details, keeps a record of every call, and starts work within minut…
Hand routine customer questions and estimate scheduling to AI, but keep detailed estimates human
He has handed initial customer communication to AI: responding to common inquiries about services and scheduling free estimates.
Quoted: Dennis Vong, owner, Inland Power Wash · via Leaders Perception: Why Inland Power Wash trusts humans for estimates and AI for replies (Eric Zavala) · 2026-09-05Questions
Can AI write a proposal or estimate for my business?
It can draft one. In these examples AI drafts proposals from sales-call recordings (Element Design Build), ballpark remodel budgets from past jobs (MCG Homes), proposal web pages from meeting transcripts (Iconoclastic Capital), bid requests from walkthrough notes (North Shore Masonry), and proposals and coverage comparisons (Evergreen Insurance). A person checks each one before a client sees it.
Can I use ChatGPT to write proposals?
Businesses here do. Evergreen Insurance uses ChatGPT for Business for proposals and coverage comparisons, and North Shore Masonry uses ChatGPT, Claude and Gemini for different jobs in its AI agents. Element Design Build, MCG Homes and KBF Design Gallery use Claude.
How much time does AI save on proposals and estimates?
It depends on the job and on your data. Element Design Build reports first proposals going from 60 to 90 minutes to 10 to 15; MCG Homes reports about eight hours to about 30 minutes in a test, within about five percent of a staff estimate; Body by Cochran's photo pre-estimates take 90 seconds instead of 30 minutes on smaller jobs, per Autobody News. These are what each business reported, not promises.
Should AI price the job on its own?
Not in these examples. Inland Power Wash went back to writing detailed estimates by hand because AI couldn't judge surface conditions, and every business here has a person check the numbers before a client sees them.
Can someone set it up for us?
Yes. The $197 Quick-Win sets up one automation in accounts you own, live within 48 hours of getting access, with a refund if it is not live in 48 hours or is not doing its job in the first 14 days.
Related guides
- AI automations for any small business, by task
- AI client intake and onboarding automation: real examples
- How contractors automate lead and quote follow-up with AI
- AI lead and quote follow-up tools for contractors
- AI automations for small insurance agencies
- AI automations for small accounting firms and bookkeepers
- AI email triage and inbox management: real examples
- Have one automation set up for you ($197)
One short email a day: 3 new AI use cases from real small businesses, each linked to its source, plus a Tool of the Day. Free, no spam, unsubscribe anytime.
Want one set up for you?
The $197 Quick-Win: we set up one AI automation that saves you hours each week, in accounts you own, live within 48 hours of getting access. No call needed. Refund if it isn't live in 48 hours or isn't doing its job in the first 14 days.
See the $197 Quick-Win Bigger setups