Automation: where to start when you're starting from zero

Most small business owners share the same feeling: they know their team is losing time on repetitive tasks, but they have no idea where to start. Too many tools, too many promises, no clear method. The good news: getting started requires neither a big budget nor technical skills. Here is the approach we recommend, in three steps, with the classic traps to avoid.

The trap of trying to automate everything

The natural reflex is to list everything that could be automated: emails, quotes, follow-ups, accounting entries, order tracking. Then to look for the big project that would handle it all at once.

That is the best way to never get anywhere. A company-wide project involves every department, requires a large budget before the first result, and stalls at the first surprise. The automation projects that fail are rarely the ones that were too small: they are the ones that tried to do everything at the same time.

The right approach is the opposite: one task at a time, starting with the one that pays off fastest. A first visible win within a few weeks builds trust, teaches the team how an automation behaves, and makes the next project easier to decide. The three steps below are there to find that first task and validate it without risk.

Step 1: measure where the time goes

Before talking about tools or artificial intelligence, observe what actually happens in your company. The simplest format: one week of honest observation.

For one week, each person concerned writes down the repetitive tasks they perform: how many times per day or per week, and how many minutes each time. A shared spreadsheet is enough, no sophisticated tool required.

The calculation that matters is frequency x duration. A 10-minute task repeated 15 times a week adds up to more than 10 hours per month. At a fully loaded hourly cost of 25 to 40 € in a small business, that is 250 to 400 € per month going into a single task, often without anyone having put a number on it.

Two tips to make the measurement useful:

Involve the team from this step onwards. The people doing the tasks are the ones who know where the time goes. Involving them in the measurement also prepares the adoption of the future automation.
Write down the real numbers, not the flattering ones. An honest measurement beats any hallway estimate. It is the reference you will use to judge the pilot.

Step 2: choose your first process

Your measurement week produces a list of tasks with numbers attached. Not all of them make good first projects. Three criteria help you choose:

Enough volume. A custom project starts at 1,500 € fixed price. A task that only represents 2 hours per month will never pay it back. Aim for a process that weighs several hours per week, or at least around ten hours per month.
A stable, describable process. If you can explain the steps to a new hire in ten minutes, that is a good sign. If the way of working changes every month or depends on who handles it, stabilise first, automate second.
Accessible tools. The automation has to connect to your existing tools: email, CRM, invoicing, spreadsheets. Most modern tools offer an API or at least exports. Without either, integration costs more than the gain.

If several processes pass all three criteria, rank them with a simple profitability calculation: time saved x hourly cost, compared to the project cost. We covered that calculation in detail in how to calculate the ROI of an automation.

Step 3: a small, measurable pilot

Do not deploy the automation on the whole process at once. Run a pilot, with four rules:

A restricted scope. One part of the process, not all of it. For email handling, for example: classify the messages and prepare draft replies, without sending anything automatically.
Real data. Test on your actual cases, not on ideal scenarios. That is where the exceptions show up.
A short duration. After a serious assessment, a working pilot goes live in 2 to 4 weeks. If someone offers you six months of project work before the first concrete result, be wary.
A measured gain before scaling up. Compare the time observed during the pilot with the measurement from step 1, and apply a prudence discount: keep 70 to 80 percent of the theoretical gain, because an automation never captures everything (edge cases, supervision, exceptions).

If the pilot delivers, widen the scope. If it disappoints, adjust or stop: you will have spent the price of a reliable answer, not the price of a large failed project. Our case studies show the orders of magnitude we see on this type of first project.

The classic mistakes of a first project

Four mistakes come up again and again on first projects:

Automating a vague process. If nobody can describe the steps precisely, there is nothing to automate. Automating vagueness just produces vagueness, faster. Clarify the process before handing it to a machine.
Aiming too wide. Connecting five tools and covering three departments in the first project multiplies the failure points. Shrink the scope until the project becomes simple to describe.
Forgetting team adoption. An automation the team works around returns nothing. Involve the people concerned from the initial measurement, show them the pilot, and adjust based on their feedback.
Ignoring running costs. AI API calls, subscriptions, maintenance, supervision: expect a few dozen euros per month for most SMB use cases. Modest, but not zero, and it belongs in the profitability calculation.

The takeaway

Starting from zero is not a handicap, it is actually the best position to do things properly: one week to measure where the time goes, three criteria to choose the first process, a 2 to 4 week pilot to validate the gain on real data. No grand project, no gamble: a step-by-step progression, with numbers at every stage.

That is exactly what we do during the free 30-minute assessment: identify with you the most profitable process to automate first, and give you a numbers-backed estimate before any commitment. See how our method works.