AI permit research, job-site daily logs and change-order scans at an Alabama builder-remodeler
Old Number One Customs (Jake Bartol, Auburn, AL)
What they did with AI
Jake Bartol, owner of Old Number One Customs, a licensed residential builder and remodeler in Auburn, Alabama, describes the AI running on every active project. An AI notetaker on site turns conversations into daily logs that sync to project management. A daily scan of client communications flags possible scope creep, and within 24 hours the client gets a written, priced change order to approve or decline. AI permit research pulls county and city code requirements and overlays such as historic districts, Lake Martin shoreline rules and FEMA flood zones, which he then confirms with the permit office. AI also gives tier-based starting estimates from the site's project configurator and drafts invoices, change orders and emails. He says AI handles the information work that used to take 40% of his time.
Related use cases
AI agents for a roofing back office: invoices, supplements and material orders at First National Roofing
First National Roofing
Adam Lynn of First National Roofing, a roofer for 17 years who says he is "not a tech guy," described in a Contractors Cloud webinar recap how he runs AI agents connected to his roofing CRM through its MCP server. Per the vendor's write-up, an agent with its own email address splits batch supplier-invoice PDFs (about an hour of manual work per batch), files each invoice to the right project, notifies the office associate and now also enters the expenses. EagleView reports are routed to an agent for automatic upload, public adjuster emails are read and logged to the project, a signed contract triggers project creation and document upload, and paid leads produce a morning status report. Another agent drafts supplements from the EagleView report, the adjuster's estimate and Xactimate pricing, and drafts material orders using his ordering standards. Lynn says his learning curve was a couple of days, a human still reviews the work, and no one was laid off; he is also setting up an alert for jobs nearing the 90-day lien-filing deadline in Atlanta. The account is a vendor-published webinar recap (Contractors Cloud's own demo used Claude over MCP); no hard before-and-after figures beyond the invoice-batch time are given.
AI agents for a large remodeler's back office: AP processing, vendor reconciliation and job-cost syncs
Neil Kelly Company
Neil Kelly Company, an 80-year-old design-build remodeler in Portland, Oregon, now runs about 30 background AI agents, according to Jeremy Wood, its senior director of IT, on JobTread's Builder Stories podcast. They check estimates against scopes of work, process accounts payable, reconcile vendor invoices (including pulling invoices from a supplier portal), and sync new CRM opportunities and Sage Intacct job costs into JobTread twice a day. Wood says the AI writes and tests ordinary scheduled data pipelines rather than guessing each time, so numbers like job costs are deterministic, and an AP clerk reviews what the agents surface. He also had AI agents grade construction software against a requirements rubric, then set up a JobTread test in an afternoon with a handful of jobs whose job costing he says reconciled to the penny against Intacct. Employees started with ChatGPT before the company built an agent layer; he advises small shops to begin read-only (for example a morning email summary) and to use company accounts that don't train on their data. These are the IT director's own statements in a podcast hosted by JobTread's CEO, with no audited results.
AI project management for general contractors: an in-house assistant for schedules, specs, submittals and RFIs
Rogers-O'Brien Construction
Rogers-O'Brien Construction, a third-generation family-owned Texas general contractor, built its own AI assistant, Compass, connected to Procore, its scheduling and estimating platforms, its data warehouse, Microsoft 365 and its internal playbook, so employees can ask about a project, spec, schedule, trade partner or policy in one place. In its first year RO says Compass answered more than 350,000 questions, about 705 people use it in a typical month and nearly a third of the company uses it daily, with more than 25,000 spec searches this year. A field director used it to turn conceptual information into a planning-level CPM schedule and compare two options (the analysis indicated reusing an existing building could save five to six months). A quality manager uses it to find which spec section applies to a submittal and build a task list from submittals and RFIs out of about 80 emails a day. A preconstruction team reconciled a healthcare project's budget against the estimate in minutes instead of an afternoon, and RO says a schedule narrative that took a Friday afternoon can now take 15 minutes. Superintendents can now ask questions aloud through Compass Voice. Figures are company-reported.