Article

AI agents for SMEs: what they can and cannot do

An honest explanation of what an AI agent does, when it delivers the most value and where a person still makes better decisions.

Written by Loek Delahaye, founder, Delahaye Solutions · 10+ years software architect and CTOPublished:
Short answer
  • An AI agent is software that works through multiple steps on its own to reach a goal, without you having to intervene each time.
  • Agents deliver the most value on tasks that currently take multiple steps and always follow the same pattern: lead follow-up, document processing, reporting.
  • What they cannot handle well: nuanced customer conversations, creative judgments and high-stakes decisions still require a person.
  • Start with a simple automation. Once that proves the process is predictable, an agent is the logical next step.

An agent is more than sending a prompt

There are three levels of AI use that business owners mix up. The first level is sending a prompt to ChatGPT and doing something with the answer yourself. The second level is an automation: a fixed trigger (a new invoice, a completed form) starts a fixed action (writing to a system, sending a confirmation email). The third level is an AI agent: software that receives a goal and decides for itself which steps are needed to reach it. It can use multiple tools, process the result of one step in the next, and adjust course if something is off. That distinction matters, because it determines which level fits your situation.

What agents can do

Tasks where an AI agent delivers the most value

Agents are strongest on processes that currently take multiple steps, always follow the same pattern and leave little room for subjective judgment.

1. Lead follow-up from start to finish

A new lead comes in through your website. The agent reads the form, looks up the company name, assesses whether it is a serious inquiry, sends a personalised first response and automatically schedules an introductory call. You only receive the warm lead with context, instead of doing every step yourself.

2. Document processing with exceptions

An invoice arrives by email. The agent reads the key data, checks it against the corresponding order, writes everything to the accounting system and flags discrepancies. If something is off, it raises the alarm instead of blindly continuing. That is more than a standard automation can do.

3. Research and preparation

Before a client meeting you need to know who you are dealing with. An agent searches the company website, recent news and public information, and delivers a concise briefing before you walk into the meeting. That currently takes fifteen minutes of manual work per client.

4. Internal workflows and onboarding

Registering a new employee or contractor requires multiple actions in multiple systems: requesting IT access, notifying HR, sending a welcome email, creating a task list. An agent can carry out all those steps in the right order as soon as the registration comes in.

5. Reports with context

Not just pulling numbers and putting them in a table, but also comparing them with previous periods and naming deviations. An agent fetches the data from your systems, composes a concise summary and sends it every Monday morning, so you start the week with the right information instead of opening a spreadsheet.

The limits

What an AI agent cannot do well

An agent is not a replacement for human judgment. For complaints from upset customers, complex negotiations or decisions with major business consequences, you want a person. Agents also make mistakes, more often than they appear to. They are most reliable on tasks where you can check the outcome before it has consequences. Processes that you cannot precisely describe yourself are also ones you cannot explain well to an agent. A vague process produces a vague outcome. The rule of thumb: if you can write the step-by-step plan on one page and the answer is always the same, an agent fits well. If constant nuance is required, a person fits better.

Frequently asked questions

Frequently asked questions about AI agents for SMEs

What is the difference between an AI agent and an automation?
An automation carries out a fixed step on a fixed trigger: a form submitted, a confirmation email sent. An AI agent receives a goal and decides itself which steps are needed to reach it. It can use multiple tools, adjust based on intermediate results and handle slightly varying situations. An agent is essentially an employee you give a task and who works out how to handle it themselves.
Which tasks are most suitable for an AI agent?
Tasks that currently take multiple manual steps, always follow roughly the same pattern and require little subjective judgment. Think of lead follow-up, document processing, summarising reports, research for client meetings and internal onboarding workflows. The more predictable the process, the better an agent performs.
Is an AI agent reliable enough to run independently?
For low-stakes tasks with reversible outcomes: yes, provided you spot-check results. For tasks with major business or financial consequences: no, a person must assess the outcome first. Always start with a period of active oversight before you let an agent run fully unsupervised.
What does it cost to have an AI agent built?
That depends strongly on the complexity of the process, the systems that need to be connected and the desired level of oversight. A single automation starts from 1,500 euros with us. A full agent that combines multiple systems and acts adaptively is typically a larger project. We make an estimate up front so you know what to expect.
How do I know whether my business is ready for an AI agent?
If you have a repetitive process that involves multiple steps and you can write down each of those steps precisely, you are ready for an agent. If the steps are unclear or vary significantly per case, start by documenting and standardising the process first. That is a useful step in itself, even without AI.

Want to know if your process is a good fit for an AI agent?

Book a free call. We look at your situation and advise honestly whether an agent is the right choice, or whether a simpler automation already does enough.

Book my free call →

Free and without obligation · Reply within one business day · Honest advice, even if automation turns out not to be the best choice