Articles · Operations & automation · September 30, 2026 · 5 min read
Choose your first business automation
Measure one repeated task, include review and upkeep, and decide whether automation is worth trying. Two worked examples and a practical test brief.
Start with one ordinary week of work. Write down a task each time you repeat it: copying an approved request into a job sheet, checking whether a form is complete, or preparing a routine report. Record the time it takes and the exceptions that make you stop. You need a useful baseline before you need another subscription.
Pick a task with a clear finish
Microsoft’s automation guidance favors repeatable, well-defined work, existing software where it fits, and a return that survives the cost of maintenance. Our practical extension is to write down the human review and exception work before estimating the benefit.
- Name one trigger: an approved form arrives, a job reaches a particular status, or a report is due.
- Name one result: a complete draft in the right place, ready for a specific person to review.
- Keep the decisions that need judgment with a person. Write down what happens when required information is missing.
- Check the software you already use for a suitable built-in feature. Confirm the actual plan, permissions and supported connection before assuming it works.
An example worth investigating
Illustrative inputs, not customer results: an office prepares 120 routine requests per month, taking 8 minutes each. A proposed flow would still need 3 minutes of human review per request and 2 hours of upkeep each month. That releases 8 hours: 120 × (8 − 3) ÷ 60 − 2.
At an assumed $30 per hour, those hours represent $240 of capacity. Subtract an assumed $60 monthly tool cost and the modeled net value is $180 per month. An assumed $900 setup would take 5 months to offset at that rate. These amounts are example inputs, not MAE prices or vendor quotes. Time released only becomes cash savings if spending actually falls; otherwise it is time available for other work.
An example to leave manual
Now suppose the task happens only 20 times a month and takes 4 minutes. The flow still needs 3 minutes of review each time and 2 hours of monthly upkeep. It adds about 1 hour 40 minutes of work. With the same $30 hourly value and $60 tool cost, modeled net value is negative $110 a month. There is no payback on those assumptions.
A clearer form or short checklist may be the better improvement. Revisit automation if volume grows, the process becomes more consistent, or maintenance becomes demonstrably smaller.
Write the test before building the flow
- Use a fictional request first. Check that the exact approved fields reach the intended destination.
- Submit the same test request twice. Check whether one record or an unwanted duplicate is created.
- Try a missing required field and an unavailable destination. A visible failure is better than a false success message.
- Name the person who approves outward messages or other consequential actions. A prepared draft is not proof that anything was sent.
- Confirm how to pause the flow and finish the job manually. Include that fallback in the instructions.
Keep a one-workflow brief
The calculator produces a brief you can copy and use with your existing software provider or whoever builds the flow. At MAE, we scope one workflow and test its actual destination before describing it as working.
Task: [one repeated job]. Trigger: [what starts it]. Current volume and time: [measured baseline]. Result and destination: [what should exist afterward]. Human review: [who checks what]. Exceptions and fallback: [what happens when it fails]. Costs and owners: [setup, monthly tools, upkeep, who pays]. Acceptance: [the evidence we will check].