Which Business Tasks Should You Automate First?
By CodexierPublished 6 min read
The wrong way to start automating is to pick a tool and look for something to do with it. The right way is to look at a normal week, find the task you repeat most often, and ask what it costs you when it goes wrong. This guide gives you a scoring method that works for a two-person consultancy as well as a thirty-person workshop, and says honestly when automation is the wrong answer.
List a normal week task by task
For one week, keep a plain list of every task that comes back more than once: the same customer question, an order re-typed from the webshop into the accounting system, a reminder before an appointment, a quote that went quiet. Note how often it happened and how many minutes each instance took. Do not skip the small ones; a three-minute task done twelve times a day is a bigger target than an hour-long task done monthly. Ask the people who actually do the work, because the person in the inbox knows which tasks are truly repetitive.
Score volume, error cost and risk
Give every task three scores from one to five. Volume is how often it happens. Error cost is what a mistake costs: a wrong invoice amount is expensive, a late newsletter is not. Risk is how bad it would be if the automation misbehaved unnoticed for a week. Add volume and error cost, subtract risk. The highest total is your candidate.
| Task | Volume | Error cost | Risk | Total |
|---|---|---|---|---|
| Copy webshop orders into Fortnox | 5 | 4 | 2 | 7 |
| Send booking confirmation and reminder | 4 | 3 | 1 | 6 |
| Follow up quotes after five days | 3 | 4 | 1 | 6 |
| Approve supplier invoices | 3 | 5 | 5 | 3 |
Example scoring for a small retailer. Your numbers will differ; the method is what matters.
Supplier invoice approval scores low even though it is frequent and expensive to get wrong. Anything that moves money out of the company, touches personal data in a new way under GDPR, or makes a decision a customer could dispute should keep a human in the loop. It can be assisted by automation, but it should not be your first pilot.
Pick one pilot and define what done means
Take the top-scoring task and write one paragraph defining done before anyone builds anything. For example: when an order is marked paid in the webshop, an invoice is created in Fortnox with the right VAT rate; if the customer number is missing, the order is flagged in a shared inbox; success means fewer than one manual correction per fifty orders after four weeks.
- Trigger: the exact event that starts the flow, and where it comes from.
- Action: what is created, sent or updated, in which system, with which fields.
- Exception path: what happens when data is missing. A flow without one fails silently.
- Owner: the one person who reviews exceptions every day.
- Measurement: minutes saved per week and corrections per hundred runs.
Keep it narrow: one trigger, one action, one exception path. Every extra branch added before the first one works doubles the time it takes to learn whether the idea was good.
What a first pilot costs in time
The build is usually the smallest part. The real cost is documenting the rule, cleaning the data it depends on, and watching exceptions during the first month. A rule that lives in one employee's head takes longer to extract than to implement.
| Phase | Your time | What happens |
|---|---|---|
| Mapping | Two to four hours | Write the rule, list the fields, find the edge cases |
| Build and test | One to two hours of review | Flow built against test data, exceptions checked |
| First month live | Ten minutes a day | Exceptions reviewed, rule adjusted, savings recorded |
| Decision | One meeting | Keep, extend or switch off based on the numbers |
If you want the build done for you, our workflow automation service is a fixed-price package covering mapping, build and the first month of adjustments. Current prices are on the pricing page.
When automation is the wrong answer
Some tasks look repetitive but are not. If the rule changes every time someone senior looks at it, you do not have a process yet, and automating it freezes a rule nobody agreed on. If the data the flow depends on is unreliable, you will produce wrong records at speed.
- Do not automate a task with no written rule. Write it, run it manually for two weeks, then decide.
- Do not automate something you are about to change, such as a pricing model under review.
- Do not buy an automation package, ours included, if the task takes under an hour a week in total. The maths will not work.
- Do not start with anything that moves money, deletes data or messages customers without a human check.
The decision rule: a task that scores seven or higher, has a written rule and takes more than an hour a week is worth a pilot. If none qualifies, make a new list in three months. A free 15-minute call is enough for us to tell you which side of that line you are on.
Frequently asked questions
Do I need AI to automate my business tasks?
Usually not for the first pilot. Most high-value tasks in a small business are rule-based and are handled by ordinary workflow tools and integrations. AI becomes useful when the input is unstructured, such as a free-text email or a phone call, and something must interpret it before a rule can apply.
What about GDPR when automating tasks with customer data?
Automation does not change what data you may process, but it can change where the data travels. If the flow sends personal data to a new tool, you need a processing agreement with that supplier and a note in your records of processing. Keep the pilot inside systems you already use where possible.
Can we automate several tasks at once to save time?
You can, but you learn less. The first pilot shows how your data really behaves and how much attention exceptions need, which makes the next automations faster and safer. Three at once means three sets of surprises at the same time.
Not sure which task to start with?
Bring your week's list to a 15-minute call. We score it with you, tell you which task is worth a pilot, and say plainly if none of them are yet.
Book a free 15-minute call