Skip to content

Light version: pages open quickly, without the 3D graphics. You are seeing the light version because the 3D graphics would be slow on this device. The content is the same.

Turn on 3D

Preparing the graphics. This can take a moment on the first visit.

Henry Lõiv

Web design & development

Artificial Intelligence 8 min read

AI Agent, Chatbot or Automation: What Does Your Business Need?

A chatbot is an interface; an agent is one way to handle the work behind it. Follow a fictional customer enquiry to choose a practical system for your business.

Choose the system around the task. A chatbot gives people a conversational interface. A fixed workflow moves a task through steps you define. An AI agent is useful when the system must decide which permitted tools or steps to use as the task develops. These can work together: a chatbot can be the front end of either a fixed workflow or an agent.

For a small business, the useful starting question is: what should happen after someone asks for help? Answer that before choosing a platform, model or subscription.

Reviewed on 13 November 2026. The enquiry below is a fictional design example, not a client case study or a measured performance test.

AI agent vs chatbot vs automation: the distinction that matters

Anthropic distinguishes workflows with predefined paths from agents that dynamically direct their process and tool use. It also recommends starting with the simplest adequate solution. Those architectural definitions are useful even as products change. Source: Anthropic, Building effective agents

TermWhat it describesQuestion to ask
ChatbotA conversational interfaceShould people interact through chat?
Fixed workflowA predefined sequence or set of branchesCan we specify the next step in advance?
AI agentDynamic choices about steps and toolsDoes the model need to choose the next action or tool?

For this comparison, a fixed workflow can branch according to predefined rules, while an agent lets the model choose the next action or tool as the task develops.

The distinction is visible in actual tooling: n8n's Chat Trigger can connect to an agent or a chain. A chat interface therefore tells you little about the architecture behind it. Source: n8n, Chat Trigger

An agent's useful freedom is also bounded. In n8n, the AI Agent works with connected tools; selecting a tool is different from having unrestricted access to a business. Source: n8n, AI Agent

One customer enquiry, three possible systems

Imagine a small design studio receives this message:

We need a bilingual service website with a booking request form before our event. Can you tell us the next step?

The event date, available content and budget are missing. The studio wants a useful brief and a sensible reply, without promising a delivery date it cannot meet.

Option 1: a fixed enquiry workflow

The studio defines a sequence: collect the message, extract requirements, flag missing information, create a draft brief and place it in the enquiry queue. A person checks the brief and replies.

AI can help interpret the free text while the surrounding process stays fixed. The workflow does not need to choose among dozens of tools. It always produces the same kind of output for the same reviewer.

This is a strong starting point when incoming requests differ in wording but follow a consistent business process. The main design work is agreeing on the required fields and deciding what happens when a field is absent or contradictory.

Missing information should remain explicitly unresolved in the output. “Before our event” belongs under an unresolved deadline, not under a guessed launch date.

Option 2: a chatbot that collects a better brief

The studio adds a conversational interface. The chatbot asks for the event date, whether the customer has copy and images, and what the booking form should do.

Here, chat is useful because a customer can answer follow-up questions immediately. The resulting brief can enter the same fixed workflow described above. Nothing about asking questions automatically requires an agent.

The interface should make its purpose clear: it helps prepare an enquiry. If the studio has not connected a reliable source of availability, the conversation should not suggest that a particular delivery slot is confirmed.

Design the exit as carefully as the opening. A customer should be able to submit what they know and reach a person when the questions stop helping.

Option 3: an agent that investigates a variable request

Now imagine the studio receives enquiries that require different combinations of information. Some need a service description, others a project checklist, and others an authorised availability record.

An agent could choose which of those sources to consult, identify a conflict and prepare a proposed next step. The value would come from handling a genuinely variable route through information.

That extra freedom brings more decisions to review. Was the right source used? Was it current? Did the agent stop when two records disagreed? Did it prepare a draft or create a commitment?

For this fictional studio, I would begin with the fixed brief workflow, then add chat if customers benefit from it. I would consider an agent after identifying recurring enquiries that the fixed process handles poorly. That is a design recommendation, not a claim that this approach has been tested at the studio.

Choose by uncertainty, not by the label

Write down three recent types of request, using anonymised or invented examples. For each, describe the desired output and the information needed to produce it.

If the route is predictable, a fixed workflow is a reasonable candidate. If the difficulty is collecting missing information, a form or chatbot may solve it. If the system must investigate different sources and adapt its route, an agent becomes worth evaluating.

There can also be a hybrid: an agent prepares a recommendation inside a larger fixed process. The surrounding workflow can still determine where that recommendation goes and which checks apply.

Avoid giving the system an extra capability merely because the platform offers it. Every connection should have a purpose you can explain in terms of the customer's task.

What does the system cost to operate?

A subscription price is only one part of the decision. Build a comparison around a complete enquiry, from arrival to a reviewed result.

Cost or responsibilityWhat to establish before building
SetupWho prepares the source material, connections and output format?
UsageWhich model calls, tool operations and retries are billable?
ReviewWho checks uncertain cases and how much context do they need?
MaintenanceWho updates service details, access and broken connections?
RecoveryWhat happens when a source is unavailable or a step fails?

Ask vendors or implementers to explain these items for your proposed workflow. An agent that makes several attempts may have a different operating pattern from a single extraction step, but the actual comparison requires the selected tools and observed usage.

For website enquiries, the related guide to AI chatbot website costs covers the wider budgeting discussion. A higher level of autonomy should have a clear business reason beyond making the demo more impressive.

Define what the system may do

For the fictional studio, I would write a short action policy before implementing the workflow:

  • Read: approved service descriptions and the information submitted with the enquiry.
  • Prepare: a brief, missing-information questions and a draft response.
  • Escalate: conflicting dates, unclear scope and unavailable records.
  • Commit: prices, delivery promises and final client messages follow the studio's chosen approval process.

Another business may authorise routine messages automatically. The important design decision is to make that authority explicit and connect it to the action's consequences.

Keep enough context for a reviewer to understand the result. A polished paragraph without the original enquiry, unresolved questions or source references makes checking harder. The output should help a person reach a decision, not merely look finished.

Run a small pilot with clear expectations

Start with a bounded task and representative examples. Include a complete request, a missing deadline, conflicting information, an unavailable source and a duplicate submission.

Write the expected handling before running the examples. For instance: a missing date should be flagged; a duplicate should not create two enquiry records; an inaccessible record should not be replaced with a guess.

Then assess the whole process. Was the brief accurate? Could the reviewer see what remained uncertain? How much correction was needed? Did the workflow recover cleanly? Record time and cost only when you have measured them.

The bilingual workflow design sheet included with this article provides a structure for these decisions. Use it before comparing product feature lists. For concrete starting points, see five practical AI workflows for work.

Use the workflow design sheet to describe the input, permitted actions, review and handoff before choosing a tool.

Frequently asked questions

Can a chatbot also be an AI agent?

Yes. Chat describes the interface. A system behind it may follow fixed steps, use an agent or combine both. Ask what the system can access, how it chooses actions and what authority it has.

Can automation include AI without becoming an agent?

Yes. For example, an enquiry process can always extract the same fields, create the same brief and send it to the same review queue. A language model can perform one step within that predefined process.

Does a small business need an AI agent?

Some tasks may justify one, but business size alone does not answer the question. Start with a repeated problem, a clear output and someone responsible for the result. Evaluate an agent when adapting the route adds practical value.

Which option is cheapest?

There is no universal answer. Compare setup, usage, review, maintenance and recovery for the same task. A low subscription price is not a complete estimate of what it takes to produce a dependable result.

Sources and further reading

Product documentation was checked on 13 November 2026. Examples and implementation recommendations are original illustrative material.

Putting AI to work on your website?

Choosing a model is one part of a working feature. If your business needs a support assistant, content workflow or connection to an existing system, I can help define and build the interface, integrations and checks it needs.

Explore AI integrations for websites and applications⁠

Written by Henry Lõiv

Keep reading

All articles⁠
  1. Best AI agents in 2026: Grok Bot vs Dots and EU availability Compare AI agents by the work they can handle and whether you can use them in Estonia or the EU. Artificial Intelligence 5 min read
  2. Claude Code vs Codex for Frontend Development: A Practical Comparison A documentation-based comparison with a runnable frontend exercise and a blank scorecard. No measured winner is claimed. Artificial Intelligence 9 min read
  3. Is Your AI-Built Website Ready for Real Customers? A practical website review using a clearly fictional example, with a downloadable launch checklist. Artificial Intelligence 9 min read

Hi!

I really love the sea, yachts, sunsets and starry skies.

Through this little 3D world, I can share a small part of myself with you.

Thank you for opening this bottle and reading my letter.

Henry

Let’s talk about your project

Tell me about your project

Tell me what you’d like to build or improve. You don’t need a finished brief — we’ll work out the scope and next steps together.

01 How can I help?
02 How many pages do you need?
03 Which extras do you need? Choose all that apply
04 When does it need to be ready?
05 Project details
06 Your contact details

Enquiry received

Thank you!

Your enquiry has been received. I usually reply by email within 2 working days.