Automation
What to automate first: how to prioritize by return
When a company decides to automate, the list of ideas appears fast: the report, collections, customer service, registrations. The hard part is choosing where to start. The most common choice is the worst one: starting with what is most interesting or with what someone saw in a video.
How much of the work can be automated
More than most people imagine, and less than the ads promise. The McKinsey Global Institute, in its report "A Future That Works" (2017), estimated that about 60% of occupations have at least 30% of activities that could be automated with technology that already exists. In the same study, fewer than 5% of occupations could be automated entirely.
The practical reading of those two sentences together: almost every job has a slice of repetitive tasks that can be taken off the person's hands, and almost no job disappears. Automating means taking away pieces of work, not replacing people. That is why the right question is "which piece first".
The three questions that decide priority
For each candidate task, answer three questions. They are worth more than any sophisticated matrix.
1. How much time does it take per month?
Multiply the frequency by the time of each run and by the number of people involved. A five-minute task done forty times a day by three people weighs more than a four-hour report done once a month. Volume fools anyone who looks only at the task that bothers them most.
2. Is the rule clear?
Traditional automation needs a rule: if this happens, do that. The more a task depends on judgment, the more expensive and risky the automation gets. Tasks with a clear rule and many repetitions come first. Those that require interpreting text or conversation can use AI, but they call for more care and supervision.
3. What happens when it gets it wrong?
A reminder sent twice is an annoyance. An invoice issued with the wrong amount is a problem. Start where an error is cheap and easy to undo. You learn to operate automations at low risk before touching anything sensitive.
A simple effort-and-return matrix
With the three answers, put each task into one of four groups:
| Group | Characteristic | What to do |
|---|---|---|
| Quick win | A lot of time, clear rule, cheap error | Automate first |
| Project | A lot of time, complex rule or costly error | Plan, map first and test in parallel |
| Convenience | Little time, clear rule | Do it when there is room |
| Trap | Little time, complex rule | Do not automate |
The trap is the most tempting group, because it tends to gather the most "intelligent" tasks. They eat weeks of project time to save minutes.
The candidates that almost always show up at the top
In service and retail operations, these tasks usually fall into the quick-win group:
- Copying data between systems. A sale that has to become a record, a charge and a delivery task.
- Reminders and collections. Due dates, pending documents, tomorrow's meeting. We go into detail in Payment reminder automation.
- First response to contacts. Confirming receipt and collecting the basics, at any hour.
- Recurring reports. The same export, the same spreadsheet, every week. See Automated reports.
- Standard documents. Contracts, proposals and receipts filled in with data that already exists.
What to do before building
Once the task is chosen, three steps prevent most frustrated projects.
- Map how it is done today. Automating a confusing process only makes the confusion move faster. The steps are in Map the process before you automate it.
- Record the starting number. How many hours per month, how many errors, how much waiting time. Without it you will not know whether it was worth it.
- Define what happens on an exception. Every automation meets cases the rule did not foresee. Decide beforehand where they go and how the person is notified.
The first automation has two goals: solve a real problem and teach the company to live with automations. That is why it should be small, visible and easy to measure.
Common mistakes when choosing
- Starting with the tool. Buying the platform and then looking for something to do with it reverses the order. Process first, tool second, and often it is one you already pay for.
- Starting with the hardest one. The big project drags on, the team loses confidence and automation gets a reputation as something that does not work.
- Automating with no owner. Someone has to be responsible for checking that it is running and for adjusting it when the rule changes.
- Ignoring the people who do the work. The person who does the task knows the exceptions. Without them, the automation covers the simple case and breaks on the rest.
An example to picture it
Imagine a real estate agency with six people in administration. Management wants a bot that answers questions from property owners. In the survey, something else turns up: every day, two people spend the morning copying data from approved proposals into the contract system and into finance's spreadsheet. The rule is clear, the volume is high and the error is easy to fix. That is the quick win. The questions bot goes into the queue as a project, after the knowledge base is organized.
Where to start
Gather the team for an hour and list each person's repetitive tasks, with estimated frequency and duration. Apply the three questions and mark the three that landed in quick win. Pick one, map it and measure before building.
If you prefer a structured starting point, delpra™'s Operational Maturity Assessment estimates how many hours a year your team spends on manual work and points to the top three priorities for your operation.
Written by the delpra™ teamProcesses, automation and AI so operations scale efficiently.