What to automate first when you have no time

There is a particular way automation projects fail. Someone maps the most complicated process in the business, builds an elaborate version of it, and it never quite goes live because the edge cases keep multiplying. Meanwhile the simple, boring thing that would have paid for itself in a fortnight never gets built.

A better rule: automate in order of how a task fails, not how complex it is.

Tier one: things that fail silently

Start with work that produces no signal when it does not happen.

Follow-up is the clearest example. If you forget to send a quote chase, nothing breaks. No alert fires. The customer does not complain. The opportunity simply cools and disappears, and you never attribute the loss to the missing email — because there was no missing email, only an absent one.

Silent failures are the highest-value automation targets precisely because you cannot manage them manually. You cannot get better at remembering something whose absence is invisible.

Concretely, that means:

• Quote follow-up after a set number of days
• Re-engagement for leads that went quiet mid-conversation
• Appointment reminders ahead of the day

Automated follow-up across email and SMS covers this. The setup is an afternoon; the return is every deal that would otherwise have gone quiet while you were on site.

Tier two: things that repeat identically

The second tier is work you do the same way every time, where the cost is not risk but repetition.

Every new job that starts with the same eight steps is a workflow template waiting to happen. The value is not only the time saved creating the list. It is that the eighth step stops being optional on a busy week.

This is the difference between your best job and your average job. Both were run by capable people. The best one followed the whole process; the average one skipped two steps under time pressure. Workflow templates make the process the default rather than the aspiration.

A reasonable test for whether something belongs here: if you handed the job to a new team member, would you write them a list? If yes, that list should be a template.

Tier three: things that are merely tedious

Only now is it worth attacking the genuinely annoying, genuinely complicated work — the multi-branch approval chain, the conditional routing, the report nobody reads but everybody asks for.

These are tempting to start with because they are the most irritating. They are the wrong place to start because they are the most likely to have exceptions, and exceptions are what kill automation projects before they ship.

When you do get here, the useful pattern is capturing the decision rather than eliminating it. Dynamic forms with approval flow do not decide whether to approve something. They make sure the approval is requested, recorded and findable — which is the part that actually goes wrong.

Triggers should follow real events

One structural note that saves a lot of grief later.

Automation tied to a schedule (“every Monday, check for…”) drifts out of step with the work. Automation tied to events (“when a booking is confirmed…”) stays in step by construction, because the event is the work.

Prefer triggers that correspond to something that genuinely happened: a lead captured, a booking confirmed, a job created, a task completed. It makes the behaviour predictable, and it makes debugging possible — you can point at the event that should have fired.

A first-fortnight plan

If you have limited time and want the highest return:

• Week one. One follow-up sequence for quotes that have gone quiet. Two messages, different channels, a few days apart.
• Week two. One workflow template for your most common job type. Just the steps, with owners.

That is it. Two things, both boring, both in tier one and two. Resist the temptation to build the clever thing until these have been running long enough that you would notice if they stopped.

You will notice. That is the point.