What to automate first — a practical order

Automation projects tend to start with whatever the person building them finds most interesting. That is almost never the thing with the shortest payback.

Here is the order we actually recommend, and why.

The test, before anything

Automate a process only if it is repetitive, rule-describable, and currently done inconsistently or late. All three.

Repetitive but done well and on time already? Automating it saves a little labour and adds a system to maintain. That can still be worth it — just be honest that it is a cost-saving, not a revenue play.

Rule-describable is the one people skip. If you cannot write down what should happen in every case, you cannot automate it; you can only automate the happy path and create an exception queue nobody watches.

1. First response to new enquiries

This is first, always, and it is not close.

An enquiry that gets a reply in under five minutes and one that gets a reply tomorrow are different businesses to the person who submitted it. The Harvard Business Review study of 1.25 million leads found firms that made contact within an hour were nearly seven times more likely to qualify the lead than those responding an hour later — and that 23% of the 2,241 companies audited never responded at all.

What to build: instant acknowledgement in the channel the person used, a routing rule that assigns a real owner, and an escalation if nobody has touched the record within a defined window.

That last part is the one everyone omits, and it is the one that catches the leads that would otherwise disappear.

2. Appointment reminders and confirmations

The cheapest revenue in most service businesses is the appointment that already exists and does not happen.

A confirmation at booking, a reminder the day before and one on the morning is a few hours of configuration. It reduces no-shows immediately and it requires no judgement at all, which makes it the safest possible automation.

Add a reply path. A reminder that cannot be answered with “can we move it?” pushes the cancellation into a phone call nobody makes.

3. Follow-up sequences — with exit conditions

Sequences are where automation starts earning properly, and also where it starts embarrassing you.

The rule: every sequence needs a defined exit. Replying, booking, buying or opting out must stop everything, not just the branch that noticed. A customer who has already paid receiving “still thinking it over?” is worse than no follow-up at all.

We wrote about the specific mechanics in workflow exit conditions, because this single issue accounts for a large share of the repair work we are called in for.

4. Review requests

Automated, timed to the moment the work is finished, sent to the channel the customer actually uses. Low effort, compounding return, and almost entirely mechanical.

The one judgement call is timing. Immediately after completion works for transactional work; a few days later works better where the value takes time to become apparent.

5. Reactivation of dormant contacts

You have already paid to acquire these people. Working the list costs almost nothing, which makes the return per hour very high — provided consent is genuinely in place and the send is batched rather than blasted. Details in database reactivation campaigns.

6. Internal reporting and admin

Last, deliberately.

Automating a report saves someone an hour a week. Automating first response can change the number of customers you have. Both are worth doing; only one of them is worth doing first.

What not to automate

Anything requiring a judgement you cannot write down. You will build a rule that is right most of the time and wrong expensively.

A broken process. Automating it makes it break faster and more consistently. Fix the process, then automate the fixed version.

Anything where being wrong is unrecoverable. Invoicing, contract execution, anything compliance-bearing. Automate the preparation and keep a human on the commit.