An AI agent is not a chatbot or using ChatGPT manually. It is software that acts on its own, chains steps and makes decisions.
Last week a client asked us if we could automate the classification of emails in their inbox. They received between 80 and 120 emails a day between customer enquiries, orders, incidents and spam. Their team spent the first hour of the morning sorting that out before they could start working. Today an agent does it in two minutes while the team has coffee.
What sets an AI agent apart from a chatbot
A chatbot answers questions. An agent acts. The difference seems subtle but changes everything. A chatbot tells you "your order is on its way". An agent detects that the order has been stuck in the warehouse for two days, generates the apology email to the customer, notifies the supplier and opens an internal ticket, all on its own.
Agents chain steps, use external tools (your CRM, your store, your inbox) and make decisions within the limits you define. They do not replace human judgement in complex situations, but they eliminate the mechanical work that consumes hours every day.
Use cases with real returns for SMEs
Classifying and replying to customer emails. The agent reads each email, identifies whether it is a price enquiry, an order incident or an information request, generates a personalised initial response and escalates to a human only when it detects situations outside the script. Typical result: 60-70% of emails handled without human intervention.
Qualifying incoming leads. When someone fills in the contact form, the agent looks up their company on LinkedIn, estimates the size, checks whether they have interacted with your website before, assigns a score and books a call if they exceed the threshold. The salesperson arrives at the call already with context.
Generating product descriptions for the catalogue. If you have a store with hundreds of references, an agent can read the name, reference and technical attributes of each product and generate an SEO-optimised description in seconds. What used to be weeks of copywriting work.
Automatic weekly reports. The agent connects to your Google Analytics, your CRM and your sales platform, consolidates the week's data and sends the summary by email every Monday at 8am. Without anyone having to prepare it.
Accessible tools without coding
Make (formerly Integromat) is the most mature option for SMEs: it has native AI modules, a visual interface and a free plan that allows you to start at no cost. n8n is the open source alternative, more powerful and more technical. Zapier Central integrates AI agents directly into Zapier flows that many companies already use.
For more complex cases, Relevance AI allows you to build agents with memory and reasoning without code, at a reasonable entry price for small businesses.
How much does it cost to get started
A basic email classification agent with Make costs between €0 and €49/month in tools, plus 3-4 hours of initial setup. If that frees up one hour per day for your team, the return is immediate.
A more complex agent, with CRM integrations and advanced decision logic, may require between €800 and €2,500 in initial development plus the monthly cost of the tools. It is still worthwhile if the process it automates consumes more than 10 hours per month of human work.
Most importantly: start with the most repetitive and best-defined process you have. Agents work well when the rules are clear. Ambiguity is still human territory.
Have a project in mind?
Tell us what you want to achieve.