A number you don't recognize lights up your phone while you're loading the truck. The voice says you did some work for their neighbor last spring and they want the same thing done. You've got no memory of the job, no name, and no quick way to check, so you fake your way through it and hope.
That's the real price of a contact list spread across six places. Some customers live in your phone, some in old email threads, a few on paper invoices sliding around the floor of the van, and the rest buried in text messages you'll never scroll back far enough to reach. So when somebody asks whether you've worked together before, the honest answer is a shrug.
Here's the part that'll cheer you up: fixing this doesn't mean buying anything. You need one plain spreadsheet, an afternoon, and a little help from AI to handle the boring sorting. Let's pull everyone into a single place you can search and sort in seconds.
What you're actually building
Forget the word CRM for a minute. A CRM (customer relationship manager) is just a fancy name for a list of your customers and what you know about them. You can build a perfectly good one in a free spreadsheet, Google Sheets or Excel, and skip the monthly fee.
The goal is simple. One row per customer, a handful of columns, and a search box that answers "have we worked with this person?" before the call even ends.
Resist the urge to track forty things. Four or five columns you'll actually fill in beat twenty you won't. Start with these:
- Name (and a company name if they have one)
- Phone (the number they actually call from)
- Town (the full address if you've got it, but the town alone is worth a lot)
- Last job (a few words: "replaced water heater", "spring cleanup")
- Date and what they paid (a ballpark is fine)
Add a notes column for the stuff that matters later: the dog that bites, the gate code, the fact that they always pay a month late. That's it. You can always slide in another column next month if you miss one.
Where your customers are hiding
Before AI can tidy anything, you've got to gather the mess into one spot. Most shops find their people scattered across the same handful of places:
- Your phone's contacts
- Sent and received email, where signatures hold names, numbers, and addresses
- Old invoices or estimates (PDFs, or an export from your invoicing app)
- Text threads with customers
- A paper calendar, a notebook, or the glovebox pile of sticky notes
Export what you can into plain text or a file. Phone contacts export to a file, most email and invoicing apps have a "download" or "export" button, and for the paper stuff you can just snap photos. Don't worry about neatness yet. Messy is fine, that's the whole point of the next step.
Let AI do the tedious pasting and sorting
This is where AI earns its keep. You can paste a jumble (a chunk of exported contacts, a few email signatures, the text off an invoice) into an AI chat tool and ask it to pull the pieces apart into a clean table.
Keep the request plain. Something like: "Here's a messy list of my customers from email and invoices. Pull out name, phone, town, and last job into a table with those four columns. If something's missing, leave the cell blank." The tool reads the blob and hands back tidy rows you can drop straight into your spreadsheet.
Work in batches. Feed it twenty or thirty at a time instead of dumping a thousand in at once, since smaller chunks are easier to check and less likely to get garbled. It'll also catch the obvious duplicates if you ask it to (say, "flag any rows that look like the same person twice").
Where AI helps, and where it'll let you down
Be clear-eyed about this, because it'll save you a headache down the road. AI is genuinely good at reshaping messy text into neat columns and spotting likely duplicates. It's fast and it doesn't get bored, which is more than you can say for doing this by hand at the kitchen table.
What it can't do is know things that aren't in front of it. If an address never made it into any of your records, the tool can't conjure it, and sometimes it'll guess instead of leaving the cell blank. It can misread a phone number, merge two different customers both named Dave, or drop the town in the wrong column. None of that is rare.
So treat its work like a sharp helper's work, not gospel. Skim every batch before you trust it. And don't paste anything you wouldn't want floating around (full card numbers, Social Security numbers, that sort of thing) into a tool you don't control. Names, numbers, and job notes are fine.
Keep it clean once it's clean
Your list only stays useful if new customers keep landing in it. The trick is making that one habit, not five.
Pick a single place to add people and a single moment to do it. Maybe every Friday you spend ten minutes adding the week's new jobs off your invoices. Maybe you add them the second you hang up the phone. Whatever you'll actually stick to beats the perfect system you abandon by March.
Once it's all in one sheet, the payoffs show up fast. Sort by town and you can line up three stops in Worcester for one trip instead of three. Search a name and you'll know in two seconds whether they're a repeat customer worth a little extra care. Filter by last job and date, and you've got a ready list of everyone due for a follow-up.
Worth doing this week
You don't need to boil the ocean. Start small and let it grow:
- Open a blank spreadsheet and make five columns: name, phone, town, last job, date and amount.
- Export your phone contacts and one month of invoices into a file or plain text.
- Paste a batch of twenty into an AI tool and ask it to sort them into those columns.
- Check the batch, fix what's off, then paste the clean rows into your sheet.
- Put a recurring ten-minute block on your calendar to add new customers as they come in.
Do that for a couple of weeks and you'll have the one thing most shops never get around to building: a single clean list that answers "have we worked with this person before?" every time. If you'd rather hand the messy first pass off to someone, that's the kind of setup we do with owners at Massachusetts AI Agency, and it's usually a one-afternoon job.