Which tasks should an SME automate first?

Wanting to automate tasks in a small business is easy. Choosing which ones is much harder. People often start with whatever annoys them most, or with what a tool promised in a demo. The result: a project that costs more than it saves, or an automation nobody uses after three months.

Here is the method we use at the start of every engagement to decide where to begin.

Step 1: list what repeats, not what annoys

For one week, ask each team to note the tasks they do every week that always follow the same path. Not one-off frustrations: routines.

Typical examples in an SME:

• copying information from a quote into an invoice;
• consolidating Excel exports for the Monday meeting;
• sorting incoming emails and forwarding them to the right person;
• preparing a new hire's arrival (contract, access, equipment);
• searching for prospects and filling in their details by hand.

The goal is a concrete list, written in the team's own words: "we type the same information three times between the quote and the invoice" rather than "optimise invoicing".

Step 2: score each task on four criteria

Give each task a score from 1 to 3 on these four criteria:

Add up the scores. The highest-scoring tasks are your first candidates. A task that takes little time but is costly when it fails (an invoice that never goes out) can rank above a time-consuming task with no real consequence.

A real example. For Manon Sassy, a casting director, the heaviest task was preparing client presentations: at least 4 to 5 hours per casting. Automated, it now takes 15 minutes at most. Read the case study

Step 3: rule out the false good candidates

Some tasks look perfect and are not:

• Processes that change every month. Automating an unstable process freezes it at the wrong moment. Stabilise it first.
• Tasks where every case is unique. If each answer needs real thinking, automation can prepare the ground but will not do the job.
• What an off-the-shelf tool already does well. If your invoicing software can already chase unpaid invoices, switch the feature on before building anything.
• Tasks nobody wants to change. Automation imposed without the people who do the work ends up bypassed.

Step 4: start small, but in production

The first project should be small enough to finish in a few weeks, and visible enough for the team to see the benefit. The classic mistake is to start with a big overhaul, or the opposite: a prototype that never leaves testing.

A good first project:

• targets a clearly identified weekly task;
• is built on the team's real files, not demo data;
• keeps a human in the loop when customers are involved (an email, an invoice);
• is documented so the team can change it without depending on a contractor.

The candidates we see most often

In most businesses with 10 to 150 employees, the same areas come out on top of the grid:

• quotes & invoicing, where information is re-keyed several times;
• financial reporting, rebuilt by hand every week;
• inbound requests handled in a shared inbox;
• lead generation, when prospecting only happens "when there's time left".

If invoicing comes out on top for you, note that France's e-invoicing reform requires structuring that data anyway by September 2027: we cover it in this article.