The first automation project is not a technical project. It is a credibility project: if it delivers a visible gain within weeks, everything that follows becomes easy to defend. If it bogs down, it vaccinates the organization against automation for two years.

Choosing which process to automate first therefore deserves more attention than choosing the tool.

The four criteria of a good first candidate

1. Frequency. A task performed ten times a day beats a task performed once a month, even if the second one hurts more. Frequency multiplies the gain and quickly produces the data that proves it works.

2. Clear rules. If you can explain the task to a new employee in thirty minutes with a one-page document, it automates well. If every case is an expert judgment call, it is not a first candidate; it is a second or third wave project.

3. Accessible data. The task relies on systems you control: your CRM, your email, your structured files. If step one of the project is negotiating access or cleaning ten years of data, your timeline just exploded.

4. A measurable gain. Before building anything, you must be able to complete this sentence: “this task costs X hours per week across Y people.” If nobody can complete it, measure first, automate second.

The three classic traps

Starting too big. The urge to “transform the whole department” produces six-month projects that die in status meetings. A good first project is almost disappointingly modest: one workflow, one trigger, one result.

Automating a broken process. If the manual process produces errors, automating it will produce the same errors, faster and at higher volume. Fix first, automate second.

Ignoring adoption. An automated workflow the team works around is a failure, even if it works technically. The people who perform the task today must take part in redesigning it; they know the exceptions the documentation ignores.

What a good first candidate looks like

A few typical examples in Québec SMBs:

  • Manual entry of information from an email or PDF into a system (orders, invoices, forms).
  • Producing a recurring document from existing data (weekly report, standard quote).
  • Sorting and routing incoming requests (client emails, tickets, applications) according to known rules.
  • The systematic follow-ups nobody has time for (payments, check-ins, renewals).

Each one checks the four boxes: frequent, rule-driven, data at hand, quantifiable gain.

Before choosing, know where you stand

The best process to automate also depends on your maturity: state of your data, process documentation, team appetite. Our AI readiness quiz gives you a picture in nine questions, free and without collecting anything.

And if you hesitate between two candidates, write to us: that is exactly the kind of question an initial audit settles in a few days.