← All writing

What to Automate First

Every Monday morning, someone in your office opens the shared inbox and starts copying details out of enquiry emails. Name into the spreadsheet. Phone number into the booking system. A note in the CRM so the follow-up does not get missed. It takes an hour, sometimes closer to two. Nobody ever designed this job. It accumulated, one small addition at a time, until it became part of somebody’s morning.

That job is a strong candidate for automation. The problem is that it sits alongside twenty other jobs that also look like candidates, and choosing the wrong one first is the most common way a business ends up deciding that automation was overhyped after all.

Choose the boring task, not the interesting one

Owners usually nominate whatever irritates them personally. That is understandable, but personal irritation is a poor proxy for value. The task that annoys you might happen four times a year. The task nobody mentions might happen forty times a week and cost you a senior person’s afternoon every single time.

There is also a pull towards the impressive option. An AI agent that handles complex client conversations sounds better at a dinner party than a script that moves enquiry data between two systems. The impressive option is also the one with the most ways to go wrong, the hardest thing to test, and the worst place to learn what your business actually needs from automation.

Your first project should be dull, obviously useful, and finished.

What makes a candidate good

Three properties do most of the work.

It happens a lot. Frequency is the multiplier on everything else. A task that takes four minutes and runs sixty times a week is worth more attention than a task that takes two hours and runs once a month. High frequency also means you find out quickly whether the automation works, because it gets exercised constantly rather than sitting idle until quarter end.

It needs little judgment. If the rule can be written as a sentence anyone in the business would agree with, it is a good candidate. “Every enquiry that comes through the website form gets logged in the CRM and gets a reply within five minutes” is a rule. “Decide which enquiries look serious” is not a rule, it is a judgment call, and judgment calls are where automations quietly start making decisions nobody sanctioned.

It is being done by someone too expensive to be doing it. This is the one owners underweight. When a practice manager on a senior salary spends ninety minutes a day on data entry, you are not just paying for data entry. You are paying that rate and also not getting the work only they can do. The real cost of a manual task is the wage plus whatever that person was hired to do instead.

Score the candidates before you build anything

Write down every repetitive task in the business. Not the ones you wish you did, the ones that actually happen. Then score each one from 1 to 5 on four questions.

  1. How often does it run? Once a quarter scores 1. Many times a day scores 5.
  2. How little judgment does it need? Pure judgment scores 1. A rule you could hand to a new starter on day one scores 5.
  3. How expensive are the hands currently doing it? A task already handled by your lowest-cost admin scores 1. A task eating your operations lead or a clinician’s time scores 5.
  4. Can you write the rule down today? If nobody can articulate the current process without a twenty minute conversation, score 1. If you could describe it in a page, score 5.

Add the four numbers. Sixteen or above is a strong first project. Below ten, leave it alone for now, whatever your gut says.

One disqualifier overrides the score. If the process itself is changing month to month, do not automate it yet. Automating an unstable process only makes the instability run faster and gives you something new to unpick every time the process moves. Stabilise it manually first, then automate the stable version.

What this looks like in practice

Suppose a physiotherapy clinic gets around 40 enquiries a week across a web form, a phone line, and a couple of directory listings. The office manager copies each one into the practice management system, sends a reply, and sets a reminder to chase if there is no response. Frequency is high. Judgment is low, since the reply is nearly the same each time. The person doing it is expensive relative to the work. The rule fits on a page. That scores near the top.

Compare it with the quarterly board pack, which takes a full day and involves real interpretation of the numbers. It is painful, but it runs four times a year and needs a human brain throughout. It is not the place to start.

The first one sets the standard

Whatever you automate first establishes how automation gets done in your business. If the first project is specified in writing, tested against real data before it goes live, and handed over with documentation that says what it does and what to check when it misbehaves, that becomes the norm. If the first one is built quickly by whoever was available and lives in that person’s head, that becomes the norm instead, and you will find out the hard way in about six months.

Pick the boring task. Score it honestly. Build it properly.

If you want a second opinion on which task to start with, you can book a free 30 minute call at https://calendly.com/brad-bravoautomate/30min.