If you've already read what an AI agent is, you have the definition down. This article is the next step: which one to buy, at what budget, and in which cases the cheap chatbot is actually the right answer.
Because it is, more often than any agency will admit.
In This Article
- Decision comparison
- When a chatbot is enough
- When you need an agent
- The real cost of each
- How to migrate without throwing away what you've built
Decision Comparison
| Criteria | Chatbot | AI Agent |
|---|---|---|
| What it solves | Narrow, frequently asked questions | Full conversations and actions |
| Off-script | Breaks down or bails out | Reasons and carries on |
| Executes in your systems | No, or very limited | Yes: CRM, calendar, orders |
| Setup cost | Low | Medium |
| Recurring cost | Fixed and low | Variable, usage-based |
| Maintenance | Manually update the decision tree | Keep the source documentation up to date |
| Main risk | Frustrating the customer | Over-answering without control |
When a Chatbot Is Enough
A well-built chatbot solves a lot if this is your reality:
- 80% of inquiries are the same five or six repeated questions: business hours, location, how to return an item, where's my order.
- You don't need to qualify anyone: whoever's messaging in is already a customer.
- The goal is to take load off the support team, not to sell.
- The budget is tight and you want something live this week.
We'll say this even though it doesn't work in our favor: if this is your situation, a €30/month chatbot gets you 80% of the result. Building an agent here means overpaying for capability you'll never use.
When You Need an Agent
An agent is worth it when the value isn't in the response itself, but in what happens next:
- Every conversation is different, and customers ask questions in their own words, not yours.
- You need to qualify leads before handing them to sales.
- You need to take action: check availability, book a meeting, create the record in the CRM.
- The average deal size is high enough that losing a lead actually hurts.
- Leads come in after hours and nobody follows up until the next day.
The litmus-test question is simple: is the problem that people ask you a lot, or that opportunities are slipping through your fingers? The first is a chatbot. The second is an agent.
Two Real-World Cases
Case A · Ecommerce with 400 Orders a Month
85% of messages are "where's my order?", "how do I return this?", and "do you ship internationally?". There's no one to qualify: whoever's writing in has already bought.
Decision: chatbot. With a twelve-answer decision tree connected to order status, it resolves most inquiries for under €50/month. An agent here would mean paying for reasoning nobody needs.
Case B · B2B Consultancy with an €8,000 Average Deal Size
60 leads come in each month through the contact form and WhatsApp. Each one asks something different, half of them aren't a fit for the service, and the sales team is two people who are already stretched thin.
Decision: AI agent. Here the value isn't in answering, it's in filtering. If the agent avoids five useless calls a month and recovers two leads that came in overnight, it pays for itself within the first quarter.
The rule we use to decide in 30 seconds: if the incoming message is from someone who's already a customer, chatbot. If it's from someone who could become one, agent.
The Real Cost of Each
| Chatbot | AI Agent | |
|---|---|---|
| Setup | €0 – €1,000 | €2,000 – €6,000 |
| Monthly recurring | €20 – €80 | €150 – €500 depending on volume |
| Time to launch | Days | 2 – 4 weeks |
| Monthly maintenance | 1 – 2 hrs | 2 – 4 hrs |
The annual cost gap runs around €4,000–€8,000. Against that, here's the right question: how many opportunities a year do you need to recover to cover it? With an average deal size of €3,000, that's two or three. If the volume of leads you're losing is higher than that, the math is already done.
How to Migrate Without Throwing Away What You've Built
If you already have a chatbot, you don't need to shut it down to try an agent:
- Keep the chatbot for narrow, frequently asked questions — it handles those well and cheaply.
- Route only sales-related traffic to the agent: the quote request form, the pricing page, WhatsApp.
- Measure both channels separately for six weeks: resolution rate, qualified leads, meetings booked.
- Decide based on the data, not the demo.
What is worth carrying over is the chatbot's conversation history: it's the best source for training and fine-tuning the agent.
Conclusion
Chatbot and AI agent aren't competing with each other: they solve different problems. The chatbot takes load off support, the agent recovers sales opportunities.
If your inquiries are repetitive and narrow, don't pay for an agent. If you're losing leads because you're not responding in time or not qualifying them, a chatbot won't fix that no matter how much you improve it.
Not sure which case you're in? Give us half an hour and we'll tell you.
