Most automations that fail in an SME don't break down: they automate the wrong thing, or a poor organisation. Trying to go too fast, copying the neighbour, neglecting edge cases and forgetting to measure are the four reflexes that turn a promised time saving into a real loss of time. Here are the most common pitfalls and the method to avoid them before you sink hours and money into them.

Automation delivers on its promises when it aims right. According to consultancy McKinsey, fewer than 5% of jobs can be fully automated, but nearly 60% of jobs contain at least 30% of automatable tasks. The lesson is simple: you don't automate a job, you automate specific tasks. SMEs that forget this launch into overly ambitious projects, and that is the leading cause of failure. The problem is almost never the technology. It lies in the starting decision: what to automate, in what order, and why.

Mistake no. 1: automating before clarifying the process

This is the most expensive mistake, because it is invisible at the start. You take a shaky process, full of useless steps and unwritten exceptions, and you hand it to the machine. The result: the machine repeats your flaws faster and on a larger scale.

This sentence from Bill Gates should be posted above every automation project. Before automating anything, lay the process out flat: who does what, in what order, with what information coming in and going out. Nine times out of ten, you discover steps to remove outright. Simplify first, automate second. It is also the right moment to check that your tools talk to each other: if your software cannot connect its data together, no automation will last over time.

Mistake no. 2: aiming too big, too fast

The second mistake is the 'big bang' project: automating everything at once, overhauling the whole organisation in a single quarter. These projects collapse under their own weight. Consultancy Gartner thus forecast that at least 30% of generative AI projects launched in companies would be abandoned after the proof-of-concept phase by the end of 2025, often due to an overly broad scope and poorly anticipated costs.

An SME has neither the time nor the budget for a project that takes six months before producing the slightest result. The right approach is the opposite: choose a single task, simple and frequent, automate it cleanly, and bank a visible win within a few days. That first success funds and legitimises the next. The automation of an SME is built with small bricks, not big building sites.

Mistake no. 3: forgetting the exceptions and the safety net

A real process is never perfectly regular. There is always the customer who pays in two instalments, the non-standard order, the supplier who replies differently. When an automation does not provide for these edge cases, it does not stop politely: it sends a wrong invoice, forgets an order or floods a customer with reminders in a loop.

Three safeguards prevent disaster. First, list the known exceptions and decide, for each, whether the machine handles it or hands it to a human. Next, set up a failure alert: you must be notified when something gets stuck, not learn it from an unhappy customer. Finally, assign a clear owner to each automation. An automation without an owner is an automation no one monitors, and which therefore ends up letting you down.

Mistake no. 4: measuring nothing, neither before nor after

If you don't know how much time a task used to cost you, you will never know whether the automation paid off. Yet this is the most widespread mistake: people automate 'by feel' and are unable to say what they gained. But the figures are the argument that unlocks the next budgets. A survey by software vendor Smartsheet showed that employees spent nearly a quarter of their working week on manual, repetitive tasks: that is exactly the pool of hours you want to quantify, then reclaim.

The method comes down to two measurements. Before: honestly time the task over a typical week, and multiply by a realistic hourly rate to get a cost in euros. After: re-measure the residual time (an automation rarely removes 100% of the work) and compare. The same is true for the most concrete gains, such as automating your quotes, invoices and reminders: without a starting figure, you are flying blind.

Mistake no. 5: resting everything on a single person

The last mistake, more insidious: the automation that exists only in one person's head. The day they go on holiday, change roles or leave the company, no one knows any more how it works or how to fix it. The 'black box' that used to save time becomes a risk.

The remedy is unglamorous but decisive: document. A simple diagram of the flow, the list of tools involved, the credentials stored in the right place, and the steps to follow when it breaks. A maintainable automation beats a brilliant automation that no one understands.

Frequently asked questions

Which task should you start with to avoid going wrong?

With a repetitive, frequent task with stable logic: chasing unpaid invoices, booking appointments, entering data between two tools. To start, avoid tasks that call for judgement or are full of exceptions: you will come back to those once the mechanism is running smoothly.

How long before you see a return?

For a simple, well-targeted automation, the gain is visible within a few days to a few weeks. If a project promises a return only after several months, it is probably too big: break it down into smaller bricks that each pay off quickly.

Do you need an in-house IT specialist to automate your SME?

No. Most automations useful to an SME connect tools you already have, without bespoke development. What matters is not coding, it is clarifying the process and choosing the right starting point. The occasional support of a specialist mainly serves to avoid the pitfalls described here.

Does artificial intelligence change the game for SMEs?

It widens the field of what can be automated (sorting requests, drafting a first reply, preparing a document), but it erases none of the mistakes above. On the contrary: an AI plugged into a vague process amplifies the vagueness even faster. The method stays the same, the need for rigour is simply greater.

Priority action plan

1. Map out the process you are targeting, above all, and remove the useless steps: you automate a clean process, never a shaky one.

2. Measure the real cost of the task today (time over a typical week × hourly rate) to have a quantified point of comparison.

3. Choose a single task that is simple and frequent to start with, rather than a big project that takes months to produce a result.

4. Plan for exceptions and a failure alert, and assign an owner: no automation should run without supervision.

5. Re-measure after a few weeks, adjust, document, then extend the method to the next task.

Automation rarely fails because of the technology. It fails because people skip the thinking step that precedes it. By avoiding these five mistakes, you turn a risky tool into a lasting advantage, and you reclaim the hours your competitors keep wasting.

Want to automate without falling into these traps? Describe your processes through our contact form, and together we'll pinpoint the tasks to automate first and the ones best left alone.