What to automate first: a 30-minute audit for any business

Why the first project matters most

Most businesses have dozens of things they could automate, and that is exactly the problem. Without a method, the first project tends to be whatever the loudest person is annoyed by, or whatever a vendor demoed last week. A poor first pick costs more than money. If it drags on or breaks, your team stops trusting automation and the next good idea gets pushed back a year. A good first pick does the opposite: it pays back quickly and builds the confidence to do more.

The audit below takes about 30 minutes and needs nothing more than a spreadsheet and someone who knows how the work actually gets done, which is often not the owner. It works for a five-person agency and for a 200-person operations team alike. You will end up with a short ranked list, one clear first project, and a written definition of success you can hold any builder to.

Step 1: list your repetitive tasks (10 minutes)

Open a spreadsheet and write down every task your team does more than a few times a month that follows roughly the same pattern each time. Do not filter yet, and do not think about technology. Walk through a normal week department by department: sales, customer service, operations, finance and admin. Ask the people doing the work, not just their managers, because the most tedious tasks are often invisible from above. For each task, note roughly how often it happens and how long one instance takes.

• What do you copy from one system and paste into another?
• Which emails or messages get nearly the same reply every time?
• What documents do you read just to type their details somewhere else?
• What reminders, chasers or follow-ups do you send by hand?
• Which reports do you rebuild from scratch every week or month?
• What gets forgotten when someone is on leave?

Step 2: score each task on four factors (10 minutes)

Give every task a score from 1 to 5 on four factors, then multiply the four numbers. Multiplying matters. A task that scores high on three factors but 1 on rules clarity ends up with a low total, and that is correct, because a task nobody can describe is a poor place to start, however painful it is. The maximum is 625. In practice, anything above about 150 deserves a serious look, and anything under 60 can usually wait.

Be honest about rules clarity, because it is the factor people overrate most. Test it by asking: could I write the steps on one page so a new hire would get it right nine times out of ten? If not, the task may still be worth automating later, but it needs process work first. AI copes with more ambiguity than traditional rules, yet it still needs clear examples of what a good outcome looks like.

• Volume: 1 means a few times a month, 3 means daily, 5 means dozens of times a day.
• Time per instance: 1 means under a minute, 3 means about 10 minutes, 5 means 30 minutes or more.
• Error cost: 1 means a mistake is harmless, 3 means rework or an annoyed customer, 5 means lost money, a lost deal or a compliance problem.
• Rules clarity: 1 means it depends on gut feel, 3 means mostly rules with some exceptions, 5 means a new hire could follow a written checklist.

A scoring example

Here is how the scoring might look for a property management firm with 20 staff. The numbers are illustrative, but the pattern is typical. Supplier invoice entry comes out on top, not because it is the most painful task, but because it happens often, takes real time, causes costly errors and follows clear rules. Tenant complaints carry the highest error cost of all, yet they score lowest because every case is different and judgment matters.

Notice that two tasks tie at 120. When that happens, prefer the one with a single owner and fewer systems involved, because it will be faster to build and easier to measure. Also notice that the complaints task is not thrown away. It becomes a later project, perhaps an assistant that drafts replies for a person to approve, once the simpler wins have built trust and freed up time.

Step 3: pick one and define success (10 minutes)

Take the top scorer and check it against two questions. Does it have one person who owns the process and can make decisions about it? Does it involve no more than three or four systems? If both answers are yes, that is your first project. If not, move to the next task on the list. Then write a short definition of success before anyone builds anything. This one page is what keeps the project small, measurable and on budget.

• Baseline: how long the task takes today, how often it happens and how many errors you see.
• Target: one number, such as 'invoices entered within 1 hour of receipt' or '20 hours saved a month'.
• Scope: which cases are in, which exceptions go to a person and which systems are touched.
• Owner: the person who approves outputs during the first weeks and signs off on launch.
• Review date: when you will compare results with the baseline, usually 30 days after go-live.

Red flags that push a task down the list

Some tasks score well on paper but make poor first projects. Watch for these warning signs, and if a top candidate shows two or more of them, either fix the underlying issue first or move to the next task. None of them mean a task can never be automated. They mean it is likely to take longer, cost more or produce results that are hard to trust in the early weeks.

• The process changes every month or lives in one person's memory.
• The data sits in spreadsheets with no consistent format, or in someone's personal inbox.
• A key system has no API and no export, so connecting to it means fragile workarounds.
• Nobody can say how long the task takes today, so success cannot be measured.
• The output goes straight to customers or moves money with no review step planned.
• Several departments must agree on how the process works before anything can be built.

What a good first project looks like

A strong first project is boring in the best way. It handles one process from start to finish, ships a working version in two to four weeks, and saves time you can measure in hours per month. It keeps a person approving anything that touches money, customers or legal decisions. It is built on accounts you own, with documentation your team can follow. And when it runs well, it makes the next project much easier to justify.

For the property firm above, that means invoices arrive in a shared inbox, an AI step reads the supplier, amount, date and property, and the workflow creates a draft bill in Xero. Anything unusual, such as a new supplier or an amount above a set limit, goes to the finance lead for approval. Staff stop typing and start checking. After a month, the team compares hours and errors against the baseline and decides what comes next.

Frequently asked questions