What to automate first: a priority spreadsheet for small business
List your recurring tasks, rate them on four scales and get a ranked shortlist of what to automate first, with a warning where a person has to stay in the loop.
This spreadsheet is for owners and small teams who keep hearing that they should automate something with AI and do not know where to start. You list the tasks that repeat every week, enter how often they run and how long they take, and rate three things from 1 to 5: how costly a mistake is, how clear the rules are, and whether the data is already digital. The sheet calculates a priority score and a rank. The score compares your tasks with each other; it does not forecast hours or money saved. Fill it in before you pay for a tool or ask someone to build an automation.
What is inside
- A How to use sheet with the rating scales and the score formula written out in plain words
- A Tasks sheet with 12 example rows and six empty rows that already contain the formulas
- Calculated columns: minutes per month, priority score, rank and a short hint for each row
- A Status dropdown: Idea, Testing, Running, Left manual
- A One task card sheet: trigger, inputs, steps, human approval, errors, weekly check, owner and stop rule
- A filled-in example of the card for invoice reminders
How to use it
- Open the Tasks sheet and read two or three example rows to see the level of detail.
- Type your own recurring tasks into the rows marked Your task, or over the examples. Clear the examples you do not need so that they leave the ranking.
- Enter runs per month and minutes per run from last month's real work, then give the three ratings using the scales on the first sheet.
- Read the Rank and Hint columns together. Take the best-ranked row whose hint is Good candidate.
- Fill in the One task card for that single task and mark every line Settled before anyone builds anything.
- Update the Status as work moves on, and rate the list again every few months.
How to read the ranking
The score multiplies the time a task takes each month by two readiness ratings and divides the result by the cost of a mistake. A task moves up when it eats a lot of time, follows clear rules and already lives in digital form. It moves down when a slip would be expensive.
Read the rank as an order for discussion. A score twice as high does not mean twice the benefit, and two rows with close scores are equal for practical purposes. The Hint column carries what the number cannot: a task can sit near the top and still need a person to approve every result.
| Hint | Why the row gets it | What to do next |
|---|---|---|
| Good candidate | At least an hour a month, rules and data rated 3 or more, cost of a mistake 3 or less | Fill in the task card and test on a small batch |
| Only with human approval | Cost of a mistake rated 4 or 5 | Automate the preparation and leave the last step to a person |
| Prepare first | Rule clarity or digital data rated 1 or 2 | Write the rules down or collect the data in one table, then rate again |
| Leave manual | Under 60 minutes a month in total | Use a checklist or a saved template instead |
When not to automate
A good rank is not enough on its own. Leave the task manual, or postpone it, in these cases.
- The task is rare. If it takes under an hour a month, setting up and looking after an automation can easily be more work than the task itself. A checklist does the job.
- The rules are unclear. If two people on the team would handle the same case differently, a tool will not settle it for you. Agree the rules, write them down, then come back.
- A mistake is expensive and nobody checks. Payments, refunds, contracts and anything a customer cannot undo should not run unattended. Automate the draft and keep the approval with a person.
- The data is not digital yet. If the input is paper, voice messages or someone's memory, the first project is getting it into one table, not automation.
What the score leaves out
The sheet does not know what a tool costs, how long setup takes or how much attention the automation will need afterwards. Weigh those yourself before you commit. The task card helps: if you cannot fill in the lines about errors and the weekly check, the task is not ready.
It also cannot tell you whether the task should exist at all. Before automating a report nobody reads, ask whether you can simply stop producing it.
Questions
Does the priority score tell me how much time I will save?
No. The score has no unit. It only orders the tasks in your own list by how promising they look. What you actually save depends on the tool, the setup work and how much checking the automation needs, and the sheet does not estimate any of that.
Why does the top-ranked task say Only with human approval?
Because a costly mistake lowers the score but does not take the task off the list. A task can take so much time and follow such clear rules that it still comes first. In that case automate the preparation and leave the final decision to a person.
Can I change the formula or the limits?
Yes. They are ordinary cell formulas. If an hour a month is too low a bar for you, change the 60 in the Hint column. If you want mistakes to weigh more, divide by the cost of a mistake twice. Keep the same formula in every row so that the ranks stay comparable.
Do I need AI for the task that comes first?
Not always. A form, a template or a scheduled message is often enough. A language model is worth considering where the input is free text, such as customer messages or documents, and a person reviews the output.