How a Phoenix pool service can use AI to juggle 200 customers without dispatch software
A spreadsheet you already keep, plus ChatGPT or Claude, can turn a Monday of scattered notes into an ordered route and ready follow-up messages.
- Works from a spreadsheet or call log you already keep, with columns for customer, address, day and notes
- You paste the day's rows into a chat, ask for a stop order and follow-up drafts, and read the result before anyone drives
- Fits an owner-led service business under about 200 customers with one or two trucks
- Getting a team on the same routine takes training built on your real tasks, not a software purchase
It is 6 a.m. on a Monday in Phoenix. The owner sits in the truck with a call log full of notes, two customers texted about green pools over the weekend, the route lives in someone's memory, and a technician is asking what order to run. An AI chat such as ChatGPT or Claude can turn that same spreadsheet into an ordered route with follow-up notes, in minutes, using the data the owner already keeps. Nothing new gets bought on day one.
The idea is simple. Most pool services already track their customers somewhere: a spreadsheet, a notes app, or a paper call log with the day's stops scratched in. The problem is not the data. The problem is that turning 200 rows of data into a sensible Tuesday takes a human an hour, so it gets skipped, and the route gets run from memory. An AI chat reads the rows, groups them, and drafts the messages, which cuts that hour to a few minutes of review.
Set up the spreadsheet that does the work
Start with one sheet and keep the columns boring and predictable. A row per customer per visit, with these fields, is enough for an AI chat to work with: customer name, service address, ZIP code or cross streets, service day, last visit date, chemicals or equipment checked, anything the technician flagged, and a notes column for anything a customer texted. AI reads clean columns far better than it reads paragraphs, so keep each cell short and consistent.
The notes column is the one that earns its keep. Green pool, pump running loud, customer wants a quote for a new filter, gate code changed. Any of those entries turn a route list into a work list, because the chat can pull them out and attach the right follow-up to the right house. A pool service in Phoenix works across a city of about 500 square miles, so the ZIP code column matters too. It lets the chat group stops by area instead of by the order someone typed them in.
- 01Open a blank sheet
One row per customer, one column per fact.
- 02Name the columns
Customer, address, ZIP, service day, last visit, work done, flags, customer notes.
- 03Fill last week first
Enter the visits you can remember, then add rows as you finish each day.
- 04Keep notes short
One line per cell, the way you would text it.
- 05Save it somewhere reachable
A cloud spreadsheet works, so you can open it from the truck.
Generate tomorrow's route in a chat
Each evening or each morning, select tomorrow's rows, copy them, and paste them into a chat with ChatGPT or Claude. Then give the chat one instruction in plain English. A workable one: You are helping a Phoenix pool service plan tomorrow's route. Here is a list of customers with addresses, ZIP codes, service history and notes. Group them by ZIP code, order the stops so we cross the city the fewest times, and list any customer with a flag or note that needs follow-up. Do not invent work I did not list.
The chat returns a grouped stop order, usually by ZIP code, plus a short list of the houses that need extra attention. The owner's job is the same as with any employee's plan: read it and fix it. The AI does not know that one gated community is a nightmare at 8 a.m. or that one customer prefers Thursdays, so the owner moves those stops by hand and the route is done. That last sentence in the prompt matters. Telling the chat not to invent work keeps it from inventing a filter replacement nobody asked for.
- 01Filter the sheet
Select only tomorrow's service rows, not the whole book.
- 02Copy and paste
Drop the rows into a new chat, keeping the column headers.
- 03Give the instruction
Ask for ZIP grouping, a sensible stop order, and a follow-up list, and say not to invent work.
- 04Read the route
Move any stops the chat could not know about, like gated entries or customer preferences.
- 05Send it to the tech
Text or print the final order for whoever drives it.
Draft the customer follow-up while it is fresh
The second half of the routine is the messages. In the same chat, ask for a short text message to each flagged customer. Something like: Draft a text under 40 words to each customer on the follow-up list. The customer with the green pool gets a plan and a date for a recheck. The customer who asked for a filter quote gets a note that a quote is coming this week. Friendly, plain, no emojis.
The chat returns drafts in seconds. The owner reads each one, edits the ones that sound off, and sends them from the business phone. The value is not that the AI writes better than the owner. It is that the follow-up happens the same evening, while the pump noise is still in the customer's head, instead of three days later or never. Two green-pool texts answered on Monday morning read differently to a customer than the same texts answered Thursday.
One caution belongs here. A chat does not know a customer's service history unless it is in the rows you paste. It also should not be trusted with numbers you did not give it, like chemical readings from a week it never saw. Keep the sheet as the source of truth, paste only the day's rows, and check every draft before it goes out under your business name.
Who this fits, and who it does not
This approach fits an owner-led service business with one or two trucks and somewhere up to a few hundred customers, where one person still knows the routes and the AI saves that person an hour a day. It also fits an owner who wants to learn the routine before deciding whether dispatch software is worth paying for, because running the spreadsheet version for a month teaches exactly what a paid tool would need to do.
It does not fit a business that already runs five trucks, takes same-day emergency calls all day, or needs live technician tracking for customers. A spreadsheet plus a chat is a planning tool, not a dispatch system. It cannot see where a truck is, reschedule in real time, or take a phone call. When those become the daily problem, that is when the paid software question becomes real, and by then the owner knows precisely what to buy.
Getting a team on the same routine
The routine sticks only if the technicians feed it. They need a reason to put the gate code change in the notes column, and a ground rule for what never goes into a chat, like customer payment details. This is where training earns its place. AZ AI Lab offers AI Training for Business: hands-on sessions built on the tasks a team actually does, so a pool crew learns this exact routine on its own sheet rather than on a generic example.
The training covers more than the prompt. People leave with take-home routines they can run the next morning, safe-use ground rules for what staff paste into a chat, and a follow-up check-in. Sessions run 90 minutes and up, for one owner, a small group or a whole team, and happen mostly over Zoom. The first step is the same one AZ AI Lab uses for everything, a free 25-minute fit call that looks at two tasks that repeat in the owner's week and picks a first result worth having. For a pool service, that first task is usually tomorrow's route.
