What Should You Automate First in a Growing Business?

Most conversations about automation start with a tool: a workflow builder, a chatbot, an integration someone saw in a demo. The more useful starting point is a question about the business: which recurring task, if it happened reliably and instantly, would make the biggest difference?

Asked that way, the answer is usually unglamorous. Sending the confirmation. Logging the enquiry. Reminding someone that a quote has gone unanswered for three days. Moving a record from one system to another so nobody has to retype it. These tasks are small individually and expensive collectively, because they are exactly the ones that slip when people are busy.

Everything that can be automated reliably can be described in that shape. If you cannot write the rule down — because it depends on reading the situation, weighing a relationship or making an exception — the task needs human judgment, and automating it will produce confident mistakes. That is also the line between rule-based automation and AI, which is covered in a separate article.

For each task you are considering, ask four questions. They are a working framework, not a formula — but they reliably separate good first automations from projects that disappoint.

1. 01How often does it happen?

   A task done forty times a week repays automation far faster than one done twice a month. Frequency multiplies every other benefit.

2. 02How clear is the rule?

   Can you write it as when–if–then without “it depends”? Clear rules automate cleanly. Fuzzy rules need standardising first.

3. 03What does delay cost?

   Some tasks lose value by the hour — a reply to a new enquiry, a reminder before an appointment. Others can wait until Friday without harm.

4. 04What does an error cost?

   A wrongly worded confirmation is embarrassing. A wrongly sent invoice or a message to the wrong patient is serious. High-cost errors need a human checkpoint.

Applied honestly, the four questions point most growing businesses to the same place: the steps between a customer's enquiry and the team's response. They happen constantly, the rules are clear, and delay is costly. In a widely cited Harvard Business Review study of 1.25 million leads, firms that attempted contact within an hour were nearly seven times as likely to qualify the lead as firms that waited even an hour longer ^[1].

• Capture — every form, message or call becomes a record automatically.
• Acknowledge — the customer gets an immediate, honest confirmation, especially out of hours.
• Assign — the enquiry goes to a named person, by rota, territory or product.
• Remind — the owner is prompted if no reply has been sent within the target time.
• Confirm — bookings, payments and next steps are confirmed without anyone retyping details.

None of these replaces a conversation. They make sure the conversation happens, and happens quickly.

• Processes you have not standardised. Automating a messy process produces a faster mess. Agree the steps first.
• Judgment calls. Pricing exceptions, complaints, credit decisions and anything sensitive need a person, even if software prepares the information.
• Rare tasks. Something done monthly is usually cheaper as a checklist than as an automation that must be built and maintained.
• Tasks whose failure you would not notice. Every automation needs someone who would know if it stopped working.

Good automation projects follow the same sequence regardless of the tool. Skipping steps is how businesses end up with impressive workflows that solve the wrong problem.

1. 01Name the problem in business terms

   Not “we need Zapier” but “patients miss appointments and we find out when the slot is already wasted.”

2. 02Map what happens today

   Write down each step, who does it, with what tool, and where it goes wrong. The map often reveals a simpler fix.

3. 03Find the rule

   Identify the part that can be written as when–if–then. That is your automation. The rest stays human.

4. 04Decide the human checkpoint

   Where must a person review, approve or be able to step in? Build that in from the start.

5. 05Build the smallest useful version

   Automate one step end to end, run it for a few weeks, and watch it.

6. 06Measure against the problem

   Did the problem named in step one actually shrink? If not, adjust before automating more.

Vendors will happily estimate the return on an automation for you. A rough estimate of your own is more useful, because it forces you to state the assumptions. Three kinds of value are worth separating, because they behave differently.

1. 01Time returned

   How often the task happens, multiplied by the minutes it takes, multiplied by what that time costs. This is the easiest value to calculate and usually the smallest.

2. 02Errors avoided

   How often the manual version goes wrong — a missed follow-up, a wrong detail retyped — multiplied by what each error costs. Often larger than the time saved, and easy to overlook.

3. 03Speed and consistency

   What it is worth for every customer to get a reply, a reminder or a confirmation on time, every time. Hard to put a number on, and frequently the biggest of the three.

In this illustration, the ten hours returned each month are useful, but the ten enquiries that no longer slip through are the real case for the automation. If your own numbers show neither, choose a different first automation.