Which process should an SME automate first?
Not the most impressive one, the one that hurts every week. The five-step method applied to three real cases: a firm, a trading business, a restaurant.

Article prepared with the help of artificial intelligence from our monitoring, then reviewed and approved by Christophe Delebarre before publication.
We get the question at every first meeting, and the answer often disappoints: you do not start with AI, you start with a process. A single one. The one that costs time every week, depends on one person, and whose result can be measured. The rest comes later, and comes better.
The three criteria of a good first process
- It is repetitive and frequent. A task that comes back every day or every week, with the same steps. A yearly task does not need automating, however painful it is.
- It rests on data that already exists. Invoices are in the ERP, requests in the mailbox, meetings in the calendar. You do not automate what lives in someone's head.
- Its result can be measured. A delay, a count, an amount. If you cannot say how long it takes today, you will not be able to say what you gained tomorrow.
A fourth criterion, often forgotten: the person doing this work today must want to be freed from it. Automating a task someone is attached to is the surest way to see the automation sit in a corner.
The five steps, in practice
Our method is one loop: identify use cases with the teams, propose the simplest solution, prototype and measure, train then deploy, monitor and move on to the next. Here is what it looks like in three typical situations, the ones we meet most often in first meetings.
An accounting firm: client reminders
Fifteen staff, three hundred client files. The chosen process: chasing missing documents and unpaid invoices, which took an assistant half a day a week, with omissions. The assistant reads the ERP, prepares reminders according to the delay and the client's history, the assistant validates the batch in ten minutes. Measure set from the start: average collection time and number of reminders sent. What to expect after one quarter: a lower average delay and no file "forgotten" any more. The next process, to choose with the team: client meeting minutes.
A building materials trader: quote requests by email
Thirty price requests a day land in a shared mailbox, with mixed attachments. Two sales people sort them every morning. The chosen process: sorting and pre-qualification. The assistant classifies each request, extracts quantities and references, checks the customer record in the ERP and prepares a draft reply or quote. The sales person reviews and sends. Measure: time between receipt and first reply. The realistic target: from half a day to under an hour, and that is what customers notice first.
A restaurant: trips across the dining room
Here the automation is not software. The process that hurt: the trips between the pass and the tables, with two waiters on a sixty-cover service and no extra to be found. The simplest solution that works: a leased delivery robot, tried over two services before signing. Measure: hours of extras per month. We told the story in our article on the cost of a service robot.
The mistakes that cost the most
- Starting with the most complex process because it is the most strategic. It involves more data, more validations, more people, and it fails more often. The first process is there to learn, not to win everything.
- Choosing the tool before the process. A platform subscription is not a plan. The plan is the list of processes, in order, each with a measure.
- Deploying without training. Teams must know what the automation does, what it does not do, and what they validate. Otherwise they work around it.
- Not measuring. Without a starting figure, renewal is decided on impressions, and impressions fade with time.
When you do not know where to start
That is precisely the role of the AI Diagnostic: a framing workshop with management, interviews with the teams to map processes and frictions, an ideation workshop, value and feasibility scoring, then a roadmap with the first prototypes and their quantified criteria. In a few weeks you know what to automate, in what order, and with what success criteria. For a first idea in five minutes, the AI Radar places your organisation on six axes.
