AI Office Hours Recap: How First National Roofing Is Automating with the Contractors Cloud MCP Server

Our first AI Office Hours went inside a real roofing company to see what AI automation looks like when it stops being a demo and starts running the business. Our team walked through a live end-to-end job built from two PDFs, and special guest Adam Lynn of First National Roofing, a roofer for 17 years and a longtime Contractors Cloud client, showed the AI agents he has running today.

The full recording is below and on our YouTube channel you can subscribe to. If you missed our AI Assistant webinar this summer, that webinar is also on the blog and is a good starting point before this one.

Contractors Cloud founder Dwight Gunnarson framed the session with a simple claim: learning AI is the most important business decision a contractor can make right now. AI lets you analyze data faster and connect software with little to no developer experience, but it still needs a central place to store and access your data.

Over the past few years Contractors Cloud rebuilt its infrastructure around native Model Context Protocol (MCP) servers, which is what makes everything below possible. Three takeaways stood out.

Takeaway 1: Start with the repetitive tasks eating your staff’s time

Adam’s first words were “I am not a tech guy.” He doesn’t write code. He asks the AI how to do things, and it tells him. When the MCP tools rolled out, he started with one question: what repetitive tasks are consuming my employees’ time? That question produced a list of agents that now run at First National Roofing:

• Supplier invoices. Their supplier sends batch PDFs with many invoices inside. His office associate used to snip and upload each one by hand, about an hour per batch. Now an AI agent has its own email address: it splits the PDF, uploads each invoice to the right project in Contractors Cloud, and notifies her. As of last week, it enters the expenses too.
• EagleViews and adjuster emails. EagleView reports are routed to the agent for automatic upload. Public adjuster emails are read by an agent that creates and updates project Jet Streams.
• Contracts and lead reports. A signed contract triggers project creation and document upload. For paid leads, he gets a morning report of yesterday’s leads and where each one stands.
• Supplements. An agent takes the EagleView, the adjuster’s estimate and Xactimate pricing and produces an estimate or supplement in seconds.
• Material orders. He taught the AI his ordering standards. It reads the EagleView and sales checklist to draft orders, and it compares original orders against returns across hundreds of jobs to tighten the numbers.

His learning curve was a couple of days at most. He treats the AI as an assistant he talks to all day about marketing and process, and he has plans for 21 agent “employees,” including a board of directors that reviews decisions across the company.

Two points he was careful to make. First, automation hasn’t meant firing anyone; a human still reviews the work, and the people freed up move to higher-value tasks. Second, he skipped the small stuff and went straight to the hard problems: how to make staff happier and the business more efficient.

Takeaway 2: The MCP server puts your whole business within reach

Contractors Cloud’s Mike Kampa showed what this looks like inside our app. Starting from just a State Farm scope of loss PDF and an EagleView PDF from a past job, he ran an entire project from Claude without clicking around in the app:

• Project creation. Uploading the scope PDF created the project, filled the estimated project value, selected the roofing and siding custom fields, and wrote an insurance claim details note.
• Scheduling and estimates. It set an inspection appointment on the calendar, parsed the EagleView, and line-item matched the insurance scope, flagging about $1,700 of non-trade items, to build proposal estimates.
• Material and work orders. Applying local Minnesota code rules (ice and water shield to 24 inches inside the exterior wall, for example), it generated material and crew work orders and put them on the whiteboard calendar for Saturday.
• Email and invoicing. It sent the project attachments to the Stratus project email address and generated the project invoice.
• An audit. Asked “What am I missing on this job?”, it caught that the deductible was missing from the State Farm invoice, noted the scope might be nearing expiration, and warned that signed paperwork wasn’t on file.

The point isn’t that each step is impressive on its own. It’s that the AI can read the documents you already receive and do the data entry, scheduling, ordering and checking that normally happens across a dozen screens.

Takeaway 3: Specialized agents, your permissions, and reports that watch for you

A few practical questions from the Q&A are worth repeating.

Your team can use it, with the same permissions. Employees connect through their own accounts, and every business rule, user permission and restriction already configured in Contractors Cloud still applies. If someone asks the AI to do something they can’t do in the app, it tells them they don’t have access.

Build specialized agents, not one that does everything. Adam’s strongest advice was to give each agent one job. Mike echoed this with skills built inside Claude or ChatGPT for ordering standards and formulas: hand one an EagleView and it produces a list tailored to your company’s waste margins and ordering rules, over and above the settings in the app.

Let scheduled reports do the watching. Adam is setting up a report that alerts him when an Atlanta job hits 80 days, leaving 10 days to file a lien before the 90-day deadline. Mike described weekly or monthly Accounts Receivable reports that create historical snapshots so you can see trends, and that can automatically trigger warnings or Jet Streams on accounts more than 30 days delinquent.

What’s next with AI and Contractors Cloud

Dwight closed with what’s coming from Contractors Cloud:

• Continued expansion of the MCP tools and API
• New production features rolling out in Stratus this week
• Integration with Hover, which already has its own MCP server
• Supplier integrations, starting with ABC Supply in the next few weeks

If you’re wondering where to start, borrow Adam’s question: what repetitive task is eating an hour of someone’s day? Pick that one, connect the MCP server, and ask the AI how to automate it. We’ll keep running these office hours periodically, so bring your questions and your own use cases to the next one.