AI-First
Business automation means having software do the repetitive tasks your teams handle by hand: data entry, follow-ups, reconciliations, email sorting. It pays off when it targets a specific, measurable process, and it gets expensive when you buy it as a concept. I train SMBs on these tools, and I see the same pattern every time: the gains come from small, well-chosen workflows, not from big programmes.
- 📊 The numbers, 67% of repetitive tasks are reportedly automatable, yet fewer than 30% of SMBs have acted.
- 🎯 Process first, map where your time goes before you pick n8n, Make or an AI agent.
- ⚠️ Inflated promises, agencies making $300,000 a month are mostly a myth.
- 🚀 A quick first win, one measurable workflow tested in a few days beats a big project.
The open questions are what the word actually covers, what the market figures really say, and how to avoid paying for a sales pitch. Here are the five steps of that reasoning, from the definition to my verdict.
What does business automation cover today?
Business automation is the use of software, and increasingly AI, to run business tasks and processes without human intervention. According to IBM, it aims to integrate and streamline an organisation's processes under centralised control, going beyond automating a single isolated tool.
In an SMB, the practical version is more modest: a chain of "if this happens, then do that", as described by francenum.gouv.fr in its guide to automation for small and mid-sized businesses, published in November 2025 and updated in July 2026.
Why is the topic back in the spotlight?
Because the average company no longer runs one piece of software, it runs twenty. The creator of the Emprende Aprendiendo channel recalls how it happened: before the 2000s, a company installed one big program to handle invoices, customers and inventory. Then came subscription software (Salesforce, Dropbox, Mailchimp), each running on its own servers, with no connection between them.
The result: teams copying data by hand from one tool to another. Platforms like Zapier, Make and n8n connect these tools through their APIs. The same creator gives the textbook example: a contact leaves their email, and the automation invoices them in Stripe, creates a record in Notion and adds them to Mailchimp, without a human touching anything.
What does AI add to these workflows?
AI adds the ability to handle what couldn't be automated before: a free-text email, a badly structured PDF, a phone call. The same video cites an ElevenLabs voice agent that calls a prospect back as soon as their form comes in, to qualify the lead.
That's where I draw the line between automation and simply "using ChatGPT". AI in a chat window stays a personal tool. AI wired into your email, your CRM and your back office becomes part of the production line. Oliver Rettschlag, managing director of PolePosition Automation, puts it his own way: ChatGPT subscriptions are no substitute for professional automation.
How many tasks can really be automated?
According to plateya.fr, citing a 2023 McKinsey Global Institute study, 67% of repetitive business tasks are technically automatable. The same article claims French small businesses and SMBs could save up to 20% of their operational time, yet fewer than 30% of them have started.
I treat these figures as an order of magnitude, not a promise: "technically automatable" doesn't mean "worth automating in your business". The McKinsey Global Institute's research remains the benchmark for sizing the potential, but the decision rests on your own workflows.
Why do so few SMBs take action?
The obstacles plateya.fr lists are lack of time, fear of complexity and a lack of in-house resources. I'd add one I see on the ground: nobody can list what's actually wasting their time. Without that list, you can't know where to start, and the project never gets past good intentions.
The francenum.gouv.fr guide brings good news: no-code platforms no longer require any particular IT skills, which puts automation within reach of small businesses. I've covered the logic of this step in process first, tool second, and the surge in demand in this overview.
The obstacle is no longer technical. It's diagnosis.
Which approach should you choose: rules, no-code or an AI agent?
SMB automation falls into three broad families: rules built into your business software, no-code platforms such as n8n, Make or Zapier, and AI agents that read, decide and act. The choice depends on the task: predictable and repetitive, or variable and text-heavy.
The table below compares these approaches and their limits, without prices I couldn't source: rates change too fast to pin down in an article.
| Approach | Use case | Main limitation | Choose it when |
|---|---|---|---|
| Business software rules | Reminders, notifications, statuses in a CRM or invoicing tool | Locked inside a single tool | The workflow fits in software you already pay for |
| No-code platform (n8n, Make, Zapier) | Connecting two or three tools with fixed logic | Breaks if a data format changes | The task is repetitive and the data is structured |
| AI agent | Reading a free-form email, sorting, drafting a reply, qualifying a lead | Can make mistakes, needs human review | The input varies and a human checks the output |
| Custom development | Critical or high-volume workflow, specific business rules | Lead time and maintenance cost | No off-the-shelf tool covers the need |
SOURCE: author's synthesis based on cited transcripts, francenum.gouv.fr and IBM · UPDATED 10/2026
When does an AI agent make sense?
An AI agent is a program that reads a piece of information, decides what happens next, carries out an action in your tools, then reports back. It makes sense when the input isn't predictable: customer requests in plain language, mixed document types, emails to classify.
My rule of thumb is simple. If an "if, then" rule will do, don't use an agent: it costs more and it's less reliable. If the input changes every time, the agent brings what a rule can't. I go deeper into what to hand over to one in AI agent automation for SMBs, and into the contractual side in the clauses no vendor shows you.
Should you hire a provider?
Sometimes. PolePosition Automation, for example, offers to build the strategy, implement it and then take over operations, and highlights its ISO 9001 and 27001 certifications to meet the requirements of the NIS 2 directive. For a workflow that touches sensitive data, that framework matters.
The word "automation" also covers hardware: in an interview on the Elgin Automação channel, Elgin's CEO explains that the acquisition of Laurenti, an 85-year-old family business, was driven by self-checkout systems in retail. That's not the turf of a services SMB, but it's a reminder that automation is also sold as turnkey products.
For a simple first use case, though, a business owner or operations manager can get started alone with a no-code tool. I recommend bringing in a provider when the workflow is critical, or when nobody in-house can maintain it.
What are the spectacular promised gains worth?
Most of the figures promised in the automation market can't be verified, and some are simply false. Two sources in this research show the two extremes: a warning from a practitioner, and a sales pitch.
Zubair Trabzada, who has run an AI automation agency for a year and publishes on the AI Workshop channel, is fed up with videos of 18- and 19-year-olds claiming $200,000 to $300,000 a month. His own numbers are more sober: over $100,000 earned in total, mostly through his teaching community, a record $30,000 in one month for his clients, and an average of $10,000 to $15,000 a month from consulting.
"It's not a lottery. It's not some joke where you jump in and make $200,000 a month."
Zubair Trabzada, AI Workshop, 2025
How do you spot a pitch that's too good to be true?
He puts the odds of reaching $200,000 to $300,000 a month at around 0.1%, and points out that the companies that get there have sales teams and long sales cycles. I share his reading, with one caveat: his account describes the agency market, not that of SMB clients. But it sheds light on something useful for you as a buyer: if a vendor shows off revenue or gains without detailing the work behind them, be wary.
At the other extreme, PolePosition Automation claims it can demonstrate savings of up to 80%. That's a ceiling stated in a sales interview, with no worked case or calculation method in the conversation. I'm not saying it's false. I'm saying the figure is worthless until someone shows you the original process, the cost before and the cost after.
My test comes down to three questions: which exact process, what time or cost before, what measurement after. A gain that can't be expressed in hours or euros on a named workflow isn't a gain. That's also why I'm wary of agencies that sell talk rather than systems that actually run.
Where should you start, and what's my verdict?
Start by listing your repetitive tasks, note how much time each one costs you every week, then pick the one that best combines impact, feasibility and simplicity. The first workflow should be testable in a few days, with a measurable result. The tool comes last.
Which first use case should you pick?
A good first use case is a frequent workflow with clear rules, where a mistake costs little. Typical examples: copying incoming form requests into a CRM, chasing unpaid invoices, sorting a shared inbox. For a detailed approach, where to start with automation without overengineering it lays out the method step by step.
Keep a human in the loop on anything that touches a customer or money, at least until you've measured the error rate. And turn down the eighteen-month "transformation" project: that's exactly what sinks integrations, as I explain in why your AI projects fail.
Should you automate now?
Yes, but start small. Here's my verdict: business automation is worth it today for any SMB losing hours every week to repetitive tasks, provided you start from a measured process rather than a trendy tool. Companies that wait keep their costs and lead times too high. Those that jump in without a method create noise and technical debt.
I don't believe in miracle gains, and I don't believe in holding back out of caution either. The value lies in integration with your real tools: email, CRM, documents, back office. A first workflow in production within two weeks tells you more about your potential than any study.
Frequently asked questions
What is business automation in simple terms?
It's software automatically carrying out tasks your teams do by hand: entering data, sending reminders, sorting, moving data from one tool to another. It runs on "if this happens, then do that" logic, increasingly supplemented by AI to handle unstructured text and documents.
Which tools should an SMB use for automation?
To connect tools with fixed logic, the no-code platforms n8n, Make and Zapier are the most widely cited. To handle variable emails or documents, you add an AI model or an agent. Before choosing, check whether your CRM, invoicing tool or email client already has a built-in automation feature.
What share of tasks can be automated?
According to plateya.fr, citing a 2023 McKinsey Global Institute study, 67% of repetitive business tasks are technically automatable. That figure measures technical potential, not return on investment: whether it pays depends on the task's volume, its stability and the cost of setting it up.
What's the difference between traditional automation and an AI agent?
Traditional automation follows fixed rules on structured data. An AI agent reads variable input, decides what to do next and acts in your tools, with a risk of error that requires human review at the start. If a simple rule is enough, traditional automation is cheaper and more reliable.
How do you measure the gains from automation?
Measure the time spent on the workflow before and after, over at least a few weeks, and convert the difference into euros. Add the cost of the tool and its maintenance. If the vendor can't give you that calculation for a named process, their savings figure isn't usable.
Vidéos YouTube
- The TRUTH About Selling AI Automations to Businesses — Zubair Trabzada | AI Workshop
- Automations (Definitive Guide 2025) — Emprende Aprendiendo
- KI im Unternehmen: Wie Automation wirklich Kosten spart — PolePosition Automation
- Elgin + Laurenti: A New Journey for Commercial Automation — Elgin Automação