Think about the first week with a new apprentice. You don’t send them to a job on their own. You let them do the radiator, and you look it over before you pack up. A few months on, the radiators are theirs and you don’t check. The boiler work stays with you for a good while longer. Nobody calls that a policy. It’s just how trust gets built. Software that does your admin should be handled in exactly the same way.
Start with everything coming to you
The safe starting point is simple. The software drafts and you approve. Every quote, every booking outside the obvious, every message to a customer lands on your phone first, and nothing goes out until you say so.
That’s how Datify works out of the box. It asks you before it sends anything to a customer, commits to a price or a booking outside your rules, changes anything to do with money, or moves a gas safety record past draft. Customer replies on WhatsApp start switched off, too. The full list is in what needs your approval.
It’s slower, and that’s fine. Those first couple of weeks are not wasted. They’re how you find out what it gets right. Keep a rough tally as you go: how many drafts you sent as they were, how many you changed, and how many you binned. After a fortnight it might look something like this.
Those numbers are made up. Yours will tell you where to start. The rows you never touch are the ones worth loosening. The rows you keep rewriting stay exactly where they are.
Loosen one narrow rule at a time
“Book whatever comes in” is not a rule you loosen. “Confirm boiler services that fit my diary, in my postcodes, on weekdays” is. The narrower the rule, the easier it is to trust and the easier it is to see when it goes wrong.
A rule is worth loosening when it passes all five of these:
In practice the first rules most owners loosen look like these: confirm a booking when it fits the hours, area and job types you’ve set and the diary is free; send the standard reminder the day before a visit; send a quote for a standard annual service at your set price. Datify only acts on a rule like that when the details are complete, and it checks the connected diary as it is at that moment before booking. If it can’t read the diary or the slot’s gone, it asks you. See setting rules for how they’re written.
Loosen one, then leave it a week or two before the next. If two rules change at once and something odd happens, you won’t know which one did it.
Loosen the rows you never touch. Leave the rows you keep rewriting exactly where they are.
What never goes on the dial
Some things don’t get loosened however good the tally looks. They’re the jobs where being wrong once costs more than all the time you’d save.
The first one isn’t a setting in Datify. It’s fixed. The record is a legal document with your name, registration number and signature on it, and we’ve written about why the engineer signs the paperwork separately.
The money line is worth writing as a number. “Anything over £500 comes to me first” is clearer than “big jobs”, and the same goes for whatever figure suits you. And one more thing Datify holds to: a customer asking for something is never treated as you agreeing to it. If someone on WhatsApp says “the last bloke did it for £60”, that comes to you.
When to turn it back
A dial goes both ways, and turning it back isn’t an admission of anything. You’d do the same with the apprentice after one bad week. Tighten a rule when:
Two things make that easy in Datify. It keeps a record of which rule it used for each thing it did, so when something looks off you can see which rule caused it. And you can pause the receptionist, customer messages, a single task or everything, and it takes effect straight away. Winter is a good time to check your rules anyway. Our piece on winter as a queue covers why.
Start with the software drafting and you approving everything. After a couple of weeks, loosen only the narrow, frequent, low-stakes rules you’ve been approving unchanged, one at a time. Never loosen gas safety paperwork, money beyond the routine, safety calls or anything unusual. Turn a rule back the first time it gets one wrong, or when the business changes.

Writes about what actually happens when you put an AI in charge of real business operations — including the parts that go wrong.