Guide
Which processes to automate first: a four-step method
How to go from a vague list of tasks that eat your time to one concrete first project, with a sheet to describe each process and a matrix to compare them.
Almost every business that wants to automate has the same problem: not too few candidates, but too many. Invoices, quotes, customer follow-up, reports, onboarding new staff… Everything looks automatable and everything looks urgent. Picking the wrong first one is costly, because it shapes how your team sees automation for years.
This guide gives you a four-step method to go from a vague list of “things that eat our time” to one concrete first project, with a sheet to describe each process and a matrix to compare them. You need no tools: a spreadsheet and a week of attention will do.
Step 1: take an inventory of repetitive tasks
During an ordinary week, ask everyone on the team to note down the tasks they repeat. It does not need to be precise: one line per task and a rough idea of how long it takes. These questions help surface the ones nobody mentions because they have become invisible:
- What do you copy from one place and paste into another?
- What information do you request or forward by email every week?
- Which spreadsheet do you update by hand?
- What do you check “just in case” because it once went wrong?
- Which task do you put off because it is tedious, even though it matters?
By the end of the week you will have somewhere between ten and thirty tasks. Group the ones that are really the same process seen by different people: “entering invoices” and “chasing missing invoices”, for example, are usually part of the same loop.
Step 2: describe each candidate on a process sheet
To compare processes you have to describe them the same way. Fill in one sheet per process with these fields:
| Field | What to note |
|---|---|
| Trigger | What starts it: an email, a form, a date, an order |
| Steps | What gets done, in order, and in which tool |
| Frequency | How many times a week or a month it happens |
| Time | How long each one takes on average, and how many people are involved |
| Exceptions | Which cases fall outside the norm and what happens to them |
| Errors | What tends to go wrong and what it leads to |
| Data | Where the information lives and whether it includes personal data |
| Owner | Who knows the process and could check the results |
Filling in the sheet is useful in itself: often it turns out that a “simple” process has six exceptions nobody had counted, or that two people do it differently. You need to know both before automating anything.
Step 3: score impact and ease
With the sheets in front of you, score each process on two axes, from 1 to 3.
Impact, or how much you gain by automating it:
- Hours a year it takes (frequency × time × people).
- Errors it causes and what they cost to fix.
- Speed: whether a customer or your team is waiting for this task to be done.
Ease, or how much it takes to automate it well:
- Clear rules: can someone explain the steps and the exceptions?
- Digital data: does the information arrive in emails, files or tools rather than on paper?
- Connectable tools: do the programs involved have an API, or at least import and export?
- Low risk: if one case goes wrong, is it caught and fixed without harm?
Add up each axis and place the processes on a four-quadrant matrix:
- High impact, easy: your first project comes from here.
- High impact, hard: important projects, but for once you have one running.
- Low impact, easy: small improvements that can wait, or that the team can build itself.
- Low impact, hard: drop them.
Example (hypothetical). A wholesaler scores four processes. Keying orders from email into the ERP: impact 3 (it happens daily and takes two people), ease 3 (orders by email, ERP with an API). Monthly sales report: impact 1, ease 3. Purchase forecasting: impact 3, ease 1 (it depends on the buyer’s judgement and on data spread across several spreadsheets). New customer onboarding: impact 2, ease 2. The first project is clear: orders. Purchase forecasting waits.
Step 4: check the winner is ready
Before going ahead, run through this list with the winning process:
- The process is agreed. Everyone does it the same way, or the right way has been decided.
- There is an owner. One person who knows the process, can check the first results and will decide what happens with exceptions.
- You can access the tools. You know which programs are involved and you have permission to connect them.
- You have a baseline. Current volume, time per case and errors. Without it you will not be able to show later whether it was worth it.
- The scope is small. Better to automate 80% of the normal cases well and leave the exceptions to a person than to try to cover everything from day one.
If any of these fails, fix it first or move to the second process on the list.
Signs a process is not ready yet
Some processes score well on paper but are worth holding back:
- It is changing. If you are switching ERP or CRM in the coming months, automate after the switch, not before.
- It depends on one person who does it “their way”. That judgement needs writing down first.
- The decision is sensitive. Approving credit or settling a major complaint can be supported by automation, but the decision should stay with a person.
- Nobody would miss it. If a task could simply stop being done, drop it instead of automating it.
Common candidates by area
If your inventory comes up short, these processes tend to score well in small businesses:
- Finance and admin: entering supplier invoices, reminders for overdue invoices, matching payments against the bank.
- Sales: getting leads into the CRM, following up on quotes sent, keeping deals up to date.
- Customer service: sorting emails and tickets, answering repeated questions.
- Operations: orders arriving by email, stock alerts, coordinating with carriers.
- People: onboarding and offboarding staff across every tool, holiday requests.
- Management: weekly reports that are assembled by hand today.
Our guide to AI process automation for small businesses has examples of what automation would do in each case and where AI comes in. And if the winner is invoice entry, how to automate invoice processing with AI walks through it end to end.
From ranking to first project
With the process chosen and the baseline measured, the next step is to turn it into a project: what exactly gets automated, with which tools, what a person reviews and what it costs. To estimate whether it pays off, start with the hours it takes today in the estimator on our home page and compare them with the cost, as we explain in how much process automation costs.
If you would rather do this exercise with someone, that is exactly what the free 30-minute audit is: we go through one process with you and tell you in writing whether it is a good candidate, what we would automate first, with which tools and what it would cost. See the rest of the method in how we work, what our AI automation and agents service covers, or write to us from the contact page.
Frequently asked questions
What makes a process a good candidate for automation?
It happens often, takes up hours, follows rules someone can explain, works with digital data in tools that can be connected, and carries little risk if one case goes wrong.
Should I start with the process that takes the most time?
Not necessarily. The best first project combines high impact with ease. A costly process full of exceptions or sensitive decisions is better left until you have one running.
How long does it take to build the list of candidates?
In a small business, a week of noting repetitive tasks and a few hours to fill in the sheets and score them are usually enough.
What if the chosen process depends on one person?
Write down how they do it first, with the steps and the judgement calls, and agree with the team on the right way. Automating judgement that lives only in someone's head usually goes wrong.