Automation can save significant time, but not every task should be automated. Some tasks benefit enormously from automation: they are repetitive, predictable and rule-based. Others depend on judgement, empathy or personal relationships, and automating them can damage customer experience.
Choosing what to automate wisely is more important than the automation tools themselves.
Good candidates for automation
Tasks that are often suitable for automation include:
- Reminders: appointment, payment and renewal reminders.
- Confirmations: order, booking and payment confirmations.
- Data transfer: moving information between tools.
- Scheduling: booking meetings without back-and-forth emails.
- Reporting: generating regular summaries from existing data.
- Routine follow-ups: welcome sequences and standard check-ins.
These tasks are predictable and follow clear rules.
Tasks to keep human
Some tasks are better handled by people:
- Complex customer problems requiring understanding and judgement.
- Complaints and sensitive situations needing empathy.
- Sales conversations where trust is built.
- Important relationship moments, such as thanking loyal customers.
- Decisions with significant consequences.
Automation can support these tasks, but people should lead them.
Use a simple scoring method
For each candidate task, consider frequency, time required, predictability and risk if something goes wrong. Tasks that are frequent, time-consuming, predictable and low risk are ideal for automation.
Fix the process first
Automating a confusing or inefficient process simply makes it fail faster. Simplify and document the process before automating it.
Start small
Begin with one simple automation that saves time quickly, such as appointment reminders or payment notifications. Early success builds confidence and helps you learn.
Before automating a customer-facing task, consider how it will feel to the customer.
Keep a human backup
Even automated processes need oversight. Ensure someone monitors results and can intervene if something goes wrong.
Consider the customer experience
Before automating a customer-facing task, consider how it will feel to the customer. Automated messages should be clear, helpful and personalised where possible. Customers should always be able to reach a real person when needed.
A practical tool: the automation scorecard
List ten repetitive tasks. Score each from one to five for frequency, time taken, predictability and low risk. Add the scores. Tasks scoring sixteen or more out of twenty are strong candidates; those scoring under ten are usually best left to people.
Questions before automating
Ask: How often does this task occur, and how long does it take? What could go wrong if it runs incorrectly? Will customers notice it is automated, and will they mind? Who will monitor it? How will we know it is working? Clear answers help you automate confidently and safely.
Common mistakes
Businesses often automate tasks that require judgement, automate broken processes, remove easy access to human support or launch automations without monitoring. Others automate rare tasks while frequent repetitive work remains manual. Choosing carefully ensures automation saves time without harming relationships.
An example in practice
A small dental practice automated appointment reminders, booking confirmations and recall notices, saving reception staff several hours each week. It deliberately kept complaints, treatment questions and anxious patient calls with experienced staff. Patients appreciated the timely reminders and continued to value the personal care they received when speaking to the team.
Partial automation
Many tasks do not need full automation. A system might prepare a draft reply, a pre-filled invoice or a summary for a person to check and send. Partial automation keeps human judgement in control while removing most of the repetitive effort, and it is often the best first step for customer-facing work.
Reviewing automations against experience
Every few months, read a sample of automated messages as if you were the customer. Are they still accurate? Do they sound helpful? Do customers reply with questions that suggest confusion? Adjusting wording and timing based on real customer reactions keeps automation from feeling cold or mechanical.