Automation

AI agent or plain automation? How to choose without overpaying

An AI agent sounds trendy, but a plain Make or n8n scenario is often enough. A simple cheat sheet: when to pick automation, an AI step, or an agent.

Bartosz CzerniawskiFounder of dajtoAI
Published: 5 min read
A wide office: a hand-drawn diagram of boxes and arrows splitting into two paths on a white wall, one person standing beside it and another seated at a table
Contents

An AI agent is a program in which a language model decides for itself which steps to take and which tools to use to reach a goal. A plain automation does the same thing every time, along a path someone designed in advance. That difference decides cost, risk and whether the solution will run stably at all.

Below is a simple cheat sheet for choosing. Position as of 14 November 2026.

What’s the difference in practice?

Take an example from a training company: someone signs up for a course, and the system should save their data, send a confirmation, add them to a list and remind them before the date. Every step is known and repeats identically. That’s plain automation. A scenario in a tool like Make or n8n does it cheaply and without surprises, as I describe in the guide to automating sign-ups.

Now a different case: “prepare information about this company before a sales call”. You don’t know in advance which pages will be needed, what follows from them, or how many steps it will take. Here an agent makes sense: it decides where to look and when to stop.

Anthropic puts it this way: in a workflow, models and tools are orchestrated through code paths defined in advance, while in an agent the model directs its own process. Their advice is simple: start with the simplest solution and add complexity only when needed.

Three levels, cheapest first

Instead of “automation or agent”, think of a ladder with three rungs.

Level How it works When it’s enough Example
1. Plain automation Fixed steps, no AI Data is structured, rules are clear Course sign-up → confirmation → list
2. Automation with an AI step Fixed path in which AI does one specific thing Input is messy (text, emails) but the way forward is known Customer email → AI recognises the topic → goes to the right person
3. Agent The model picks steps and tools The task is open-ended, steps can’t be written down Preparing a company brief from many sources

Most processes in companies fit levels 1 and 2. Lead handling is a classic level 2: the path is fixed and AI classifies the messages.

Five questions that help you choose

  1. Can you write the steps on a sheet of paper? If so, automation is enough.
  2. Is the input structured? Forms and fields: level 1. Free text: level 2.
  3. What does a mistake cost? The higher the price of an error (money, client, law), the less autonomy.
  4. Do I need to know why the system did what it did? A plain automation can be replayed step by step. An agent is harder.
  5. How often does the task repeat? Rare, one-off research is handy to give to an agent. Repetitive daily processes are better locked down in a scenario.

If you answer “yes, it can be fixed in advance” to most of these, you don’t need an agent.

How much does an agent cost, and why is it often more?

I won’t quote figures, because they depend on the tool and the scale, but the mechanism is constant. A plain automation has a predictable cost: this many steps, this many operations. An agent usually makes more model calls, because it plans, checks and corrects on its own, and you don’t know the number in advance. Run time grows too, along with the risk that it does something you didn’t plan for.

Then there’s the cost of upkeep. An agent needs testing on many examples, monitoring and regular prompt fixes. A scenario in Make or n8n usually just works until the system on the other side changes. I covered choosing a tool in the comparison of Make, Zapier and n8n.

How do you limit the risk if you do need an agent?

  • Minimum permissions. The agent reads email but isn’t allowed to delete it. It has access to one folder, not the whole drive.
  • A human approves irreversible actions. Sending an email to a client, a payment, a price list change: preview first, then “approve”.
  • Limits on steps and budget. An agent that loops can burn money. Set a maximum number of steps and a daily cost cap.
  • Logs. Record what the agent did and why, so you can check later.
  • Checking results. Agents hallucinate too. The checking rules are in the post on AI hallucinations.

Add a company AI policy that says what data must not go into tools: the post on company AI policy.

Do it yourself or get help?

You can build a plain automation or a simple AI step yourself in Make or n8n, if you like that kind of thing and have time to test. Help is useful when the process touches clients, money or personal data, or when you want to check which level you really need. We handle such projects under automation for businesses, with implementation from PLN 3,900 net (+23% VAT).

Summary

An AI agent isn’t the next step after automation, just a different tool for different jobs. When the steps are known, use automation. When the input is messy, add an AI step. Keep agents for open-ended tasks and give them permissions carefully.

One thing to do today: pick one process you want to automate and try to write its steps on a sheet of paper. If you can, you don’t need an agent.

Sources

  1. Anthropic, Building effective agents, 2024.
  • AI agents
  • automation
  • Make
  • n8n

Written with the help of AI, edited and checked by a human.

Bartosz CzerniawskiFor 10 years he made films, campaigns and content for businesses. Today he builds AI tools and processes.More about me
Call · 30 min · free

Facing a similar problem?

Book a 30-minute call.

Book a free 30-minute call(opens in a new tab)
All posts
Newsletter

AI in business, straight to the point

What works, what it costs and what to avoid. Straight to your inbox, no spam. The newsletter is in Polish.

The data controller is Moon Lion Media Bartosz Czerniawski. Sign-ups are handled by Kit. More in the privacy policy.

Manage cookies

Essential · always on

They remember your choice and keep the site secure. The site won't work without them.