How do I know if a recurring task is worth automating?
A recurring task is worth automating when the time or expense you can recover justifies the cost to set it up and keep it running. Measure the work people do today. Subtract the time they will still spend on review, exceptions, and repairs. Give the remaining hours a value, subtract running costs, and compare the result with the setup cost. If the numbers look close, test the step that takes the most time before you buy anything.
What does the task cost today?
Imagine an office that sends its managers a weekly list of open requests. On Friday, a coordinator opens two systems, copies records into a spreadsheet, checks who owns each item, and sends the report. The file arrives at noon. The coordinator started at nine. That three-hour gap sounds like a good opportunity until you learn that it includes a staff meeting, a phone call, and a wait for a colleague to correct a record.
Measure the minutes the coordinator actually spends on the report. Count time spent finding the latest file, fixing an unreadable date, and answering a question caused by an unclear entry. Leave out the meeting. Record the wait for the correction separately. Cutting that wait could get the report to managers sooner, even if it frees very little staff time. Both outcomes matter, but they answer different questions about value.
Follow several runs. A quiet Friday with ten open items can give a misleading picture of a month-end Friday with fifty. Write down the number of items, minutes of work, and anything unusual. Ask the person doing the job which weeks your sample misses. If the office skips holiday weeks or only prepares the report during its busy season, use the actual number of runs in a year. An entry on a recurring calendar does not prove that someone did the work.
Which minutes can you actually recover?
Split the report into pieces you can observe. Collect the records. Match them to the right owner. Find items without an owner. Decide who takes each one. Send the result. A connection between systems might do the first two steps quickly. The decision about an unassigned request could still take a phone call. Someone may also want to read the list once before it goes to management. Give those minutes a place in the estimate.
Include work the change creates. Who checks that both sources finished updating? Who handles a failed export? If a new check takes three minutes each week, add it to the time per run. If maintenance takes an hour every few months, put it in the maintenance budget. Choose one place to count each piece of work. Charging for the same hour in both the weekly review and the annual support estimate makes two good options hard to compare.
Ask the person who receives the result what they use. A manager may need just the five overdue requests and their owners. In that case, a saved view in an existing tool could replace much of the report. Price that smaller change as an option. There is little value in automating the production of fourteen columns if the reader can make the same decision with three.
A worked example for a weekly report
The following figures are illustrative. Suppose the report takes 75 minutes of staff time and runs 48 times a year. After a change, the coordinator expects to spend 15 minutes per run on review and typical corrections. That saves 60 minutes per run. At 48 runs, it frees 48 hours a year. An internal planning rate of $35 per hour gives those hours a value of $1,680. Use a rate that makes sense for the person doing your work.
Suppose setup costs $900, including time for the coordinator to explain the process and check the first reports. Software and maintenance together cost another $480 each year. The estimated annual benefit after running costs is $1,200. Spread evenly through the year, that is $100 a month. At that rate the setup cost is recovered in nine months. The first full year would return $300 in estimated value after setup and running costs.
Here is the calculation you can reuse. Hours recovered per year = (current minutes per run − remaining minutes per run) × runs per year ÷ 60. Multiply the hours by your hourly value, then subtract annual running costs. Divide the one-time setup cost by the resulting annual benefit and multiply by 12 to estimate payback in months. If annual benefit is zero or below, the task has no payback from time savings under those assumptions.
Microsoft uses frequency, processing time, labor cost, and time to repair errors as inputs in its automation planning materials. The figures above are an example of how an owner can apply those inputs to one decision. See the Microsoft automation setup checklist.
Does an hour saved mean cash saved?
Ask what will happen during the hour the coordinator gets back. If the office pays overtime every Friday to finish this report, removing that overtime can reduce a bill. If a salaried coordinator uses the hour to call customers whose requests are overdue, the office has gained capacity. The same payroll may go out that week. You can still have a sound reason to invest if the calls are work the business needs to do.
Name the intended use before making the purchase. “Return the five oldest customer calls before lunch on Friday” gives the coordinator and manager a result they can check. “Be more productive” gives them little to work with. If there is no shortage of time and nobody can name a better use for it, use a lower value for the recovered hours or make the decision on a different benefit.
Sometimes the deadline matters more than the labor. A Friday report that regularly arrives after the managers meet cannot help them assign work at that meeting. Ask what changes if they receive it on Thursday afternoon. Track whether more overdue requests get owners before the meeting ends. You can use that operational result alongside the time estimate without inventing a dollar amount for a missed decision.
Which assumption would change the decision?
Run the example again with 45 minutes of review per report. Now the coordinator saves 30 minutes each time, or 24 hours a year. Those hours are worth $840 at the same planning rate. Subtract the $480 running cost and the annual benefit falls to $360. The same $900 setup now takes 30 months to recover. A half hour of extra checking has changed a purchase that pays back within a year into a much longer commitment.
Turn the calculation around if you have a firm deadline. In this example, recovering setup and running costs within the first year requires $1,380 of value. At $35 an hour that takes about 39.4 hours a year, or roughly 50 minutes saved on each of 48 reports. Starting at 75 minutes, the coordinator has about 25 minutes left for review and exceptions on an average run. That gives you a specific question to put to a trial.
Also ask how long the current process will exist. If your team plans to replace one of the source systems in six months, a nine-month payback deserves a closer look. Will the connection survive the move? How much of the setup can you reuse? Ask for the cost of rebuilding the affected step and compare it with the benefit you could collect before the change. The answer might favor a temporary spreadsheet improvement.
What should a small trial measure?
Choose an ordinary reporting week and one with a known problem. Have the coordinator prepare the normal report while the proposed workflow prepares a separate copy. Compare the actual rows, owners, and totals. Time how long it takes someone to verify the copy and correct errors. If the workflow generated a file in seconds but the coordinator had to trace every row to its source, use the full checking time in the estimate.
Look for errors that shift effort to someone else. A missing owner may be quick for the coordinator to skip and expensive for three managers to sort out later. Ask the receiving team to flag any row they cannot act on. Add that work to the trial record. If a revised version fixes the problem, run the same example again and keep the most recent measurement with the assumptions it depends on.
What belongs in the purchase decision?
Put the decision on a single sheet. Identify the job, its owner, the current time per run, the expected review time, the frequency, and the value per hour. Show setup and annual running costs on separate lines. Include the payback for the expected case and the slower review case. Write the main uncertainty beside the number it affects. A colleague should be able to change one input and explain why the answer moved.
Agree on what would make you stop. For the weekly report, that might be review time above the 25-minute budget or a report that still misses the meeting. Also agree on who will measure the first few live runs. When the next invoice for the tool arrives, the owner will have records of time used, reports delivered, and follow-up work completed. Those records make renewal a business decision you can explain. For a way to run that trial with past work, see how to test an AI workflow before giving it real requests.