You don’t need another platform. You need agents trained on your jobs, wired into the tools your team already uses. Built by an operator-engineer, owned by you, audit-ready for diligence.
30-minute Operating Diagnostic. We confirm the math before anyone signs.
The engagement
Built for
Deliverables
Timeline
Architecture
Read the existing stack. Map the data flows. Identify where agents plug in without disrupting human workflows. Document everything in plain operator English.
Ship
First agent live in week one of the build phase. Subsequent agents on a 1–2 week cadence depending on complexity. Each one gets human-in-the-loop until the model is shop-tuned.
Train
Every correction your team makes feeds the model. We retrain on your real files quarterly. Accuracy targets are 70% week 1, 85% week 4, 90%+ by week 12.
The diagnosis
They sell a platform with a restoration sticker. The demo looks great. The migration takes nine months, the team hates it, and the agent still hallucinates on a real Cat 3 file. You’re paying for someone else’s roadmap.
Six PMs producing inconsistent narratives. Eight to ten hours a week of one person rewriting before submission. The work scales linearly with volume — meaning your ops leader can’t take a vacation, and you can’t open a fifth market.
DASH, CompanyCam, Xactimate, QuickBooks, your PSA, Encircle, Albi. Each one has its own truth. Reporting is a part-time analyst pulling CSVs at midnight. The numbers a buyer wants to see are nowhere clean.
Vendor accounts, vendor data, vendor lock-in. The acquirer’s diligence team asks who owns the model, the corpus, the integration code — and the answer is “the SaaS company.” That’s a deduction off the multiple.
Where our cohort lands
Day 0
Day 90
The outcome: A coordinated set of stack-native agents that compound margin, document themselves for diligence, and carry over to your acquirer.
The cost of another quarter
Every quarter you wait, a regional manager spends another 100+ hours on work an agent does in minutes. Every quarter is also a quarter of operating data that doesn’t exist clean for your eventual acquirer. Both compound.
Case study
Anonymized post-mortem from one of the shops we’ve worked with.
Northpoint Restoration · $22M · 4 markets
Recent shop · Regional manager review time 8–10 hrs → ~2 hrs/wk
Multi-location across four markets. Already had DASH, CompanyCam, and Xactimate humming, plus a part-time analyst pulling reports. Pain point was scope-of-loss writing — six PMs producing inconsistent narratives, and the regional manager spent 8–10 hours a week reviewing and rewriting before submission. Built a custom agent trained on their last 18 months of approved estimates that drafts the scope narrative from CompanyCam photos and the Xactimate sketch, flagging missing photos for the PM before submission. First agent shipped in week one as a draft-quality tool, tuned over six weeks against their own approved-estimate corpus. Regional manager review time dropped from 8–10 hours to about 2, PMs stopped resubmitting for missing documentation, and the agent now handles the first draft on roughly 70% of mit jobs.
“I’d been pitched 'AI for restoration’ by four vendors and they all wanted to sell me a platform that replaced tools we already use. Matt asked what we had, learned our stack, and shipped something that lives inside our existing workflow. That’s the only version of this that was ever going to work for us.”
Priya Ramaswamy
VP Operations · Northpoint Restoration
The fit picture
Yes — and here’s why
Everywhere else
At R360
AI platforms: rip-and-replace, lock-in, generic LLM with a sticker
Stack-native agents trained on your 18 months of approved estimates
Generic dev shops: build to spec, walk away, no operator context
Operator-engineer: knows what a Cat 3 file looks like and what a buyer’s diligence team asks for
In-house build: 12-month roadmap, hire an ML team, hope it ships
First agent in week one, accuracy targets at 70/85/90% in 12 weeks, hand-off documented
No — and we’ll say so
We’ll say so on the call. Here’s the short version.
The agents on this engagement
Stack-native. Trained on your jobs. Owned by you when the engagement ends.
Drafts scope-of-loss narratives from your CompanyCam photos and Xactimate sketch.
Trained on your last 18 months of approved estimates. The agent reads the sketch + the photo set, drafts the narrative with the right IICRC reference, and routes it back to your estimator for QA. Your estimator becomes the decision-maker on every file — not the typist.
3–5×
Recent shop · scope output per seat
Reclaims aging dollars you'd already written off.
Tied to QuickBooks (or your AR system). Surfaces every invoice over 45 days, drafts the chase email with the right tone for that customer's payment history, and schedules the follow-up cadence. Recovers the dollars you stopped chasing on Friday afternoons.
$214K
Recent shop · recovered / quarter
Per-job margin, AR aging, supplement capture — live from your stack.
Pulls live from QuickBooks and your field tools. Per-job margin, AR aging, supplement capture, tech profitability, scope cycle time — all defensible numbers a buyer will pay multiples on. Quarter close drops from days to minutes.
15 min
Recent shop · quarterly close
Flags margin risk on a job before it eats the close-out.
Watches every active job for the leading indicators of margin compression — scope creep, supplement-rejection patterns, tech overtime drift, AR slip on the carrier. PMs see the score in their morning queue. The bleed gets named while there's still time to fix it.
90+
Target · job health score (out of 100)
Routes crews to jobs without conflicts.
Reads your active job list, crew skill matrix, and geography. Builds the week. Surfaces conflicts before the lead tech finds out at 7 AM in the truck. Fewer driveway re-routes, more productive hours per crew.
<1
Recent shop · conflict / 30 days
FAQ
No. We integrate with what you have — Xactimate, CompanyCam, Encircle, DASH, Restoration Manager, Albi, whatever your stack is. The whole point of custom agents over off-the-shelf platforms is they live where your team already works. If a vendor is asking you to migrate everything to their platform to get AI, that’s a platform play, not an AI play. Different product.
First agent ships in week one as a usable draft tool — usually scope narrative, photo triage, or supplement drafting depending on your biggest pain. Tuning against your own data takes 4–8 weeks to get to production quality. Full implementation across multiple workflows is typically a 90-day engagement. We don’t sell six-month discovery phases.
You do. The agents run on your accounts (OpenAI, Anthropic, or whatever LLM provider fits), the training corpus is your historical data and stays yours, and the integration code is documented and handed off. We’re not building lock-in. If you want to take it in-house or hand it to another engineer in year two, the seams are clean.
Materially, in the right direction. The audit log per agent — every file, every touch, every correction — is the cleanest operating-data trail a buyer’s diligence team will ever see. Plus the agents and corpus transfer with the business. We’ve designed the engagement specifically for shops looking at a sale in 18–36 months.
30-minute Operating Diagnostic. We see if the math works for both sides before anyone signs.
Independent Restorers
You built a shop. Now it can’t run without you — and that’s exactly why no one will buy it. We pull you out of production without dropping margin, and rebuild the books a buyer will pay for.
TPA Management
TPA programs feel like a lifeline at $1M. They feel like a margin cap at $5M+. We rebuild your scope-defense and supplement workflow so you stop fighting for money you should already be getting paid.
Custom agents on your stack
30-min diagnostic. No pitch.