The mini-CRM you'll actually keep: eight fields, no more

A customer messages you on WhatsApp: "hey, that quote you sent me — where are we with it?" And off you go. It's not in your phone, because that thread was email. In email there are two versions of it. In the spreadsheet the row is half-filled, because you were in a rush that day. Five minutes later you reply with less context than you had when you spoke to them. What's missing isn't a CRM. It's one place — a single one — holding what you know about each customer.

What breaks when a customer lives in five places

Scattered information doesn't hurt every day. It hurts in three specific moments, and all three cost money.

The first is when you reply. Every "let me check and get back to you" turns a thirty-second answer into a conversation that drags on for a day. The customer doesn't think you're disorganised; they think you don't have them front of mind, which is worse.

The second is follow-up. Quotes are lost to silence far more often than to price. Nobody decides to stop chasing a €2,000 job: the week goes by, something else lands, and the quote stops existing because it wasn't written anywhere that forced you to look at it.

The third matters most over time: you can't hand any of it over. Not to the person starting in September, not to an AI agent. Both need the same thing — who this customer is and where you left off — and right now that lives in your head.

Why the "proper" CRM dies after three weeks

Most small businesses have tried this once. Someone sets up a serious CRM, with a pipeline, stages and forty fields. It gets filled in for two weeks; by the third only the person who set it up bothers; a month later it's dead and nobody says so out loud.

That isn't a discipline problem. Entering data costs you now and pays off later, often for someone else. If filling in the record takes ninety seconds and gives you nothing back in the moment, your brain quietly decides that today is a no.

Which leaves the only two rules that matter: filling it in has to take under thirty seconds, and reading it has to help you today. Everything else — integrations, pretty dashboards, lead scoring — is up for debate. Those two aren't.

The eight fields you'll actually fill in

A customer record that works for a small business fits in eight columns:

1. Who they are. Name, and whether it's a company or a private customer. One field, not three.
2. Where they actually talk to you. Not the email they typed into the form: the mobile number they message from at nine at night.
3. How they found you. Referral from Marta, Google, Instagram, the sign in the window. One word will do.
4. What they want, in one line. "Bathroom refit, 6 m², wants large-format tile." In your own words, not in categories.
5. Status. Four values at most: new, quoted, customer, closed. Twelve stages is a pipeline from a textbook, not your business.
6. Last time you spoke. A date. This is the field that will surprise you most.
7. Next step, and when. "Call Tuesday the 12th to confirm measurements." If this box is empty, that customer has been forgotten, whether you know it or not.
8. What isn't written anywhere else. Pays on 60-day terms. Prefers a call, hates voice notes. Their partner is the one who decides. These notes are 80% of the value of the whole thing.

And what stays out: invoices and VAT (that lives in your accounting tool, and it's better off there), the full conversation history (WhatsApp already has it), and any field you fill in "just in case". If you haven't looked at a column in two months, delete it.

Where the record should live

In the most boring thing you already have open. A Google Sheet — one tab, one row per customer — covers almost any business under ten people. If you want views and filters, Airtable or Notion. And if you already invoice through Holded, Odoo or Xero, the customer record that exists there is a serious candidate: one less place to maintain.

I'd only insist on two things. Two people must be able to write at once without clobbering each other, so no spreadsheet living on the office computer. And you must be able to export to CSV in one click: the day you want to plug in an agent or switch tools, that export is your insurance.

The rule that makes it maintain itself

Records don't fail because of the record, they fail because of the moment. "I'll update it when something happens" isn't a moment, it's an intention. It has to hook onto something you already do.

What works is picking one single writing moment and sticking to it:

• Project businesses (builders, agencies, professional firms): the record gets touched right after a quote goes out or the call ends, while it's fresh. Thirty seconds.
• Volume businesses (shop, clinic, workshop): ten minutes on Friday, with the week's list in front of you.

Two warnings. One: one record, one owner. If you and Marta both update the same customer whenever you remember, within a month you have two versions of the truth and neither is trustworthy. Two: the field holding the whole thing up is number seven, the next step. Sort the sheet by that date and it stops being a file — it's your week's to-do list, already prioritised. Without it you own a very tidy graveyard of contacts.

When the agent reads and writes the record

This is where it stops depending on your willpower. An AI agent handling your messages can do two things with the record, in this order.

First, read. Before replying, it checks that customer's row and answers with context: "the quote we sent on the 24th", "same as last time, we'll leave it at reception". The gap between that and a bot starting from zero every time is huge, and all it needs is a tidy, readable record.

Then, write. Every closed conversation updates the date, the status and a one-line summary on its own. That's the point where the mini-CRM stops depending on you remembering on a Friday afternoon.

With two limits I'd put in writing on day one. The agent writes to specific fields (date, status, summary, proposed next step) and leaves the rest alone — no deleting rows, no changing payment terms. And field eight you write yourself, in a tone you wouldn't be embarrassed by if the customer ever read it: it's personal data, even when it's four words in a spreadsheet. You approve, they execute.

When not to bother with any of this

Not every business needs this, and building it at the wrong time just adds a chore:

• If you live off five big clients on annual contracts, you don't have a data problem: you have five relationships that fit in your head and a calendar. A folder each and you're done.
• If you sell online at volume and most buyers never come back, a record per person tells you very little. Your store and your support tool already own that.
• If your real problem is that customers aren't coming in, tidying up the few you have won't fix it. Obvious, and people do it anyway, because tidying feels like progress.

A mini-CRM doesn't create opportunities. It stops you dropping the ones you already have. Less exciting, considerably more profitable.

Start empty, don't migrate

The temptation is to spend a Saturday dumping in three years of history. Don't: you'll end up with two hundred dead rows and zero desire to open the file again.

Start empty. This week, everyone who talks to you goes into the sheet with their eight fields, and that's it. Within a month you'll have your active customers in there, which are the ones that matter; the old ones fill themselves in the day they get back in touch. In week four, sort by "next step" and count how many rows haven't been touched in over a month. That number usually stings, and it's usually the first time you've seen your actual book of business.

Once the record is alive, handing off the upkeep is the natural next step — it's what Yaqbot's operations agent does: reads the record before replying, leaves it updated afterwards. But the sheet comes first; the tool comes after.