Key answer: An AI agent is a program built on a language model that completes multi-step tasks on its own – for example answering sales enquiries based on your price list, qualifying leads or preparing reports from your data. It differs from a chatbot in that it doesn't just talk, it acts – it uses your company's tools, data and systems. A simple agent starts at around €5,000; complex implementations integrated with multiple systems cost more and always require a workshop first.
Table of contents
The short answer
An AI agent is a program that completes multi-step tasks on its own: it answers sales enquiries based on your price list, qualifies leads, watches deadlines, prepares reports from data. It differs fundamentally from a chatbot – it doesn't just talk, it acts, using your company's tools, databases and systems. It's the fastest-growing topic in business AI, and also the easiest place to burn a budget on an impressive demo.
How an AI agent works
An agent combines three elements. First, a language model that understands instructions and context. Second, tools: access to the product database, the CRM, the calendar, email. Third, rules of engagement – a description of what the agent may do alone and what requires a human sign-off.
In practice it looks like this: an enquiry arrives, the agent recognises which product it concerns, checks the price list and availability in your systems, drafts a reply consistent with company rules, then either sends it or passes it to an employee for approval. The difference shows in numbers: an enquiry that used to wait until the next day gets its answer in a minute.
What companies use agents for first
- Sales enquiry handling: replies based on the offer, price list and availability; consistent and immediate.
- Lead and ticket qualification: the agent recognises whether a message is a complaint, an invoice question or a new lead, and routes it to the right person with a ready summary.
- Recurring reports: every Monday the agent gathers data from systems and prepares the report that used to cost someone half a day.
- Process watching: deadlines, missing documents, order statuses; the agent reminds the right people itself.
How much do AI agents for business cost?
A simple agent built on company data starts at around €5,000. Implementations integrated with multiple systems, with permissions to act inside them, start at around €12,000. Add running costs: model API fees (usually tens of euros monthly at moderate scale) and technical care. The most expensive mistake, though, isn't the build price – it's building an agent for a process nobody has ever described. That's why we start with a workshop, not with code.
Security: where to draw the lines
An agent gets exactly the permissions we grant it, and that is a design decision, not a technical detail. Healthy rules: company data never feeds public model training; irreversible actions (sending a discounted offer, changing an accounting record) go through human approval; every agent answer is logged, so you can check where it took its information from. When we implement agents, these boundaries are written down in black and white before the start.
Where to start
With the one process that hurts most and happens often. For most companies that's handling repetitive enquiries or a recurring report. A pilot with a measurable goal will show within weeks whether the agent earns its keep – and only then is it worth widening its permissions and scope. We described the full process in Implementing AI in a company, step by step.
Want to check whether an AI agent makes sense for your process? Write to us and we'll work it out at a workshop.