Find the repetitive work worth automating

Record triggers, actions and handoffs

For each repeated task, write what starts it, who carries it out, which information they need and what completion looks like. “Customer admin” is too broad. “A customer emails an appointment change; the office updates the booking and sends confirmation” is specific enough to evaluate.

Include every handoff. If the task takes three minutes in the inbox and another four to update a calendar, those are both part of the workflow. If someone has to chase missing information, record that separately rather than hiding it inside an average. Repeated chasing often points to a better form or clearer policy, not a more powerful AI system.

Use the free workbook with real observations

The Hampala workflow audit workbook has an Instructions tab, a Task log and a Summary. Enter the task name, owner, repetitions per month, minutes per repetition and estimated exception percentage. Monthly hours are calculated from frequency and time; priority hours discount those hours by the exception percentage. The Summary totals those screening figures.

The workbook includes clearly labeled illustrative tasks so you can see how the formulas behave. Replace them with your own observations, or clear their names and inputs. Use consistent monthly units. Do not enter weekly frequency in a monthly field or combine two employees' time while counting only one employee's repetitions. Small unit mistakes can completely reverse the ranking.

Track exceptions as carefully as minutes

An exception is a case that does not follow the usual path: a disputed invoice, missing booking details, an unusual delivery request or a question outside the approved policy. Count exceptions instead of relying on memory. A workflow with many exceptions may need a clearer process before it can be automated safely.

Keep exception frequency separate from consequence. A rare bank-detail change is still high risk. A frequent missing postcode may be easy to resolve. The priority score is only a time-and-predictability screen; write data sensitivity, customer commitments, financial impact and setup complexity in the notes. A high score never overrides a serious safety concern.

Further reference: NIST: AI Risk Management Framework

Rank candidates without claiming savings

Monthly hours equal repetitions multiplied by minutes, divided by 60. Priority hours equal monthly hours multiplied by one minus the exception percentage. In an illustrative example, a task repeated 80 times at ten minutes each occupies about 13.3 hours. At a ten-percent exception rate it scores approximately 12 priority hours.

That is not a promise of twelve hours saved. The process may still require review, the tool may not handle every ordinary case and configuration takes time. Use the figure to compare candidates with the same method. Then inspect the top candidates manually: is the input stable, does a suitable feature already exist, can a person verify the output, and can the process be rolled back?

Look for simplification before automation

Suppose three colleagues each maintain a different spreadsheet for the same bookings. Connecting all three may preserve the underlying problem. A single agreed record with clear ownership can remove duplication before any integration begins. Likewise, a better enquiry form may eliminate a repeated clarification email without requiring AI at all.

For the strongest candidates, sketch a simple before-and-after process. Name which steps disappear, which become automatic and which remain human. Ask the people doing the work to check the sketch. They often know the exceptions an owner misses, and a workflow that ignores their practical knowledge will create workarounds rather than reliable time relief.

Turn the audit into one testable decision

Choose one task and write a small pilot brief: its trigger, allowed inputs, expected output, prohibited actions, review owner and fallback. Attach several representative examples, including an exception. Agree how you will measure the test: time per task, correction rate, completed cases and unresolved cases are useful starting points.

Keep the audit available after the pilot. Revisit it when volume changes, staff change or an upstream tool changes its behavior. Do not re-audit the whole business every week. A periodic review of the shortlist and exception notes is enough to keep priorities honest. The outcome should be a better operational decision, not another spreadsheet that nobody uses.

Turn your shortlist into a 30-day pilot plan →

Check your units with a worked example

For an illustrative reminder task, forty monthly repetitions at six minutes each occupy four hours. If one in ten cases needs special handling, the exception input is ten percent, not 0.1. The priority figure is 3.6 hours. It is a comparison value only; it omits review, setup and the consequences of a missed reminder.

Ask another colleague to check a few entries. Confirm frequency refers to the same period and that time includes the full task. If an estimate has a wide range, record that in the notes and test whether changing it would change your shortlist. A ranking that flips under small changes needs more observation.