ALL FIELD NOTES
PLAYBOOK·MAY 18, 2026·14 MIN
THE AUTONOMOUS ROOFING OPERATOR PLAYBOOK
A field-tested playbook for running a roofing company where agents quote, schedule, dispatch, and follow up — while owners decide.
A field-tested playbook for running a roofing company where agents quote, schedule, dispatch, and follow up — while owners decide.
Every roofing company past $3M in revenue hits the same wall. Inbound leads outrun the people who can quote them. Schedules slip because nobody centralizes change orders. Insurance claims sit in a folder for three weeks because the AR coordinator is also doing payroll. The owner ends the quarter with bigger top-line and worse margin, and starts the next quarter by trying to hire their way out of the problem.
Hiring almost never works. It defers the bottleneck by 90 days and re-creates it at a higher cost basis. The companies we work with that have actually scaled past $10M did the opposite: they ran the operation through an agentic system that handles the four functions every growing roofer drowns in — quoting, scheduling, dispatch, and follow-up — and they kept the headcount lean.
This playbook is what that system looks like in 2026. It is not aspirational. It is what fifty operators are running right now on ROOF_OS.
The word gets misused. Autonomous does not mean "there is no human in the loop." It means the system takes the next action by default and escalates only when a hard threshold is crossed.
In practice that looks like this. The quoting agent generates and sends an estimate under a configurable dollar limit (typical: $25K residential, $80K small commercial) without asking. Anything above the limit gets a Slack ping to the estimating lead with the proposed price, the comps it used, and a one-tap approve/edit interface. Same for scheduling: routine reschedules happen automatically; weather pushes affecting more than four crews in one day get escalated.
The threshold is the contract between the owner and the system. Most operators we onboard set it conservatively for the first 30 days and ratchet it up as trust builds. By month three the typical authority limit is 4–6x where it started.
Skip the marketing copy about "AI for every department." In a roofing operation, four agents do the heavy lifting and everything else is decoration.
Takes the lead intake (address, scope hints, photos, insurance status) and produces a written estimate inside 12 minutes. Uses aerial measurement, current local material pricing, and the company's labor productivity averages. References the estimate workflow so the output matches your shop's pricing posture, not a generic SaaS template.
Reads the install schedule, the weather forecast for each job address (from NOAA storm data), and crew capacity. Books, reschedules, and confirms with the homeowner over SMS. Anything that can't be auto-resolved (two-day rain across a region, a crew lead calling out) lands in the dispatch queue with proposed options ranked by P&L impact.
Once a job is scheduled, the dispatch agent orders materials from the supplier portal, pushes the job packet to the crew app, and confirms delivery with the yard. If a delivery slips, the agent re-sequences the day's crews against dispatch automation and updates the homeowner without anyone asking.
Triggers payment reminders on the contract's milestone schedule. Offers a partial-payment option after the second nudge. Files a notice of intent on day 45. After job completion, requests a review via the homeowner's preferred channel. Anything dormant for 90 days gets a final touch and is then archived to keep the funnel honest.
When the four agents are running, the owner's day-to-day collapses. Daily approvals drop from roughly 200 (rough estimate from interviews with operators we onboarded) to fewer than ten. Those ten are the ones the owner actually wants: pricing exceptions, hiring decisions, disputed claims, expansion calls.
What you save is not just time. It's executive function. Most roofing owners who have run their company for ten-plus years describe the cognitive shift as "I stopped waking up at 3 a.m. with a list." The list is gone because the list runs itself.
Phased rollout. Always.
The whole sequence takes 60–90 days if you commit. Operators who try to flip all four agents on day one universally hate the experience and never make it to month two.
Three things stay the same and you should be skeptical of any vendor who promises otherwise.
An operator running the four-agent stack at $10M revenue typically employs: 1 owner, 1 estimating lead (reviews flagged quotes), 1 ops manager (handles flagged scheduling/dispatch), 1 AR/admin (handles flagged AR), 4–6 sales closers, and the production crews. That's a back-office headcount of 4 plus crews. The same revenue without the system usually requires 10–14 back-office bodies.
The math is straightforward. A capable autonomous stack costs roughly the loaded cost of 1–1.5 hires; it does the work of 6–8.
If you've read this far, you already know whether your operation has the bottleneck this playbook addresses. The single highest-leverage move is to book a 30-minute walkthrough and watch the four agents run a sample week of your business from a CSV of your last 60 days of leads.
If you want to read further before talking to anyone, the spokes in this cluster go deep on each piece: replacing four office hires, the agentic OS architecture, from inbound call to signed contract, storm response without a storm team, the quiet death of Excel, permits at scale, and the integration architecture behind it.