Tools we’ve built.

Nothing about quoting was digital. Each job was priced by hand on a printed sheet, with a pencil and a calculator, and the sheet went into a manila folder. Another person typed it up as a contract and handed it back to go to the customer. Each salesperson kept their own follow-ups in folders in their office.

Now it’s one tool. The salesperson prices the job from the price book, publishes it to the ERP, and generates the contract proposal as a PDF, with no one retyping it. Every open quote has its next follow-up, and notes and to-dos sit on the customer, deal, or quote they’re about. A proposal that took hours takes minutes. Follow-ups went up, and the win rate rose 15% in the first month.

90%less time to put a proposal together

Five weeks of installs at a glance, every crew’s next three weeks on one grid, and a map of the jobs with each installer’s travel range. Drag a job to move it. Each job shows which steps are done, from deposit to drawings, and the schedule warns when a crew is booked on a job whose drawings aren’t approved yet. It replaced a Trello board, spreadsheets, and an Access database, and brought what lived in people’s heads into one place.

It runs 50 to 100 jobs at a time, balancing what the shop has to build, and when, against the crews and installers who put it in. Then we added AI agents: they read the scheduling email, file each message on its job, and check the schedule against set rules to flag problems before they happen. They’re advisory. The company chose to have a person make every scheduling decision.

• Jobs arrive from the ERP when a quote is won
• 50 to 100 jobs at once, with the shop’s schedule and the crews’
• AI agents flag problems; a person makes every decision
• Jobs at risk are flagged three weeks out
• Crews, trucks, shop days, and time off on one grid

Collections had been largely reactive. Now every call, email, meeting, and promise on an account sits in one timeline, so whoever picks up the phone knows the last thing that was said. Follow-ups come due on the dashboard. Reminder letters are drafted as invoices age and wait in an outbox until a person approves them. Nothing goes to a customer without someone’s OK. The finance team runs it every day.

• Invoices and payments synced from the ERP
• Promises tracked until they’re kept or broken
• Statements with or without late interest, in a click

50%+drop in outstanding AR after it went in

The builder is too small for a full-time bookkeeper, and kept hiring and training part-time ones, so the books came out different every time. Now we run them. Every bank and card transaction arrives on its own, and an AI agent codes each one to a house and a cost type, reading checks from a photo. Most come back sure enough to approve in one click. We review the few it isn’t sure of, ask the builder only when a question needs them, and every correction teaches the agent.

It knows the entries a builder’s books need: each check and card charge to its house and cost type, a loan’s interest and draws from the lender’s statement, and a sold house’s cost moved to cost of sale. A person approves each one, and the books stay in QuickBooks.

• The agent codes most of the volume on its own
• Cost to date and loan balance for every house
• A person approves every entry

Reps see their accounts, orders, and commission. Customers drop in a purchase order, the lines fill in, and each line shows when it ships from the stock on hand. The sales team sees the whole business, by agency, by state, and by month.