How a Midwest roofing shop captured 62% more storm-season leads in 90 days.

The shop, before

The customer is a 14-tech roofing shop covering three counties in a hail-belt state. Storm-driven retail and insurance work makes up roughly 70% of revenue; the remainder is steady-state replacement and repair work. The shop runs a 5-person CSR team during business hours and has historically used a third-party answering service after-hours.

Their problem was visible in their own numbers. Pulling 90 days of call data showed three patterns:

• The voicemail cliff. 41% of after-hours calls went to voicemail. Of those, only about 18% were ever called back — and almost never within 24 hours.
• The web-chat ghost. Their website had a chat widget, but it was unstaffed outside business hours. Inbound chats accumulated overnight; most were stale by the time they were picked up.
• The storm-spike. When a hail event hit, inbound volume could spike 8–10x within an hour. Their CSR team had no way to handle that surge, and the answering service was a noticeable downgrade in conversation quality.

What we deployed

The deployment ran the standard 14-day timeline. Three channels went live:

• Missed-call text-back on the main shop number. Any unanswered call gets an SMS within 60 seconds: "Sorry we missed you — this is [Shop Name]'s booking line. What's going on?"
• Web chat on the homepage and the storm-damage landing page, with storm-aware greeting copy.
• Google Business Profile messaging for the GBP-sourced inbound segment.

The AI Employee was tuned against 60 days of the shop's own historical inbound transcripts. The qualifier focused on the four highest-leverage data points for this market: storm date, damage type, insurance carrier, and prior claim history.

The 90-day comparison

We compared the 90 days after full deployment against the 90 days before deployment, with weather data normalised. Both windows had similar hail and wind activity by NWS data, so the comparison isn't seasonal noise.

What the AI Employee actually did

The headline number is the lead capture lift. But the underlying behaviour matters more, because it's what's repeatable.

On a typical after-hours hail-storm night, the AI Employee handled 30–40 inbound conversations between 8pm and 7am. Of those:

• Roughly half were qualified, dispatched, and confirmed for next-day inspections — without any human involvement.
• About 30% were partially qualified and flagged for the morning CSR team with full context.
• The remaining 20% were lower-priority (spam, out-of-area, prior-customer follow-up) and routed appropriately without burning CSR time.

The shop's CSR team kept the same headcount. Their work shifted from initial intake to closing and customer-experience escalation — higher-value work that only humans should be doing.

What this didn't fix

It's worth being honest about what the AI Employee couldn't do for this shop:

• It didn't increase total inbound volume. Lead capture lift came from converting unanswered inbound — not generating new demand.
• It didn't improve close rates on inspections that did get booked. That's a sales-process question, not an intake question.
• It didn't replace any human role. The CSR team's work changed shape; it didn't shrink.

Those are the right boundaries. An AI Employee fixes a specific, measurable problem (lost inbound). It doesn't fix sales execution or generate net-new demand.

What it would cost a similar shop

For a roofing shop of similar size and storm exposure, the typical setup is:

• Setup and 14-day deployment — one-time fee.
• Monthly subscription — based on inbound volume and channels deployed.
• Volume-uncapped pricing for storm-belt customers, since the highest-value moments are exactly when volume spikes.

For specifics tuned to your shop's size, volume, and trade, book a 20-minute call.