The One-Sentence Distinction
A chatbot responds. An AI agent acts.
A chatbot is designed for conversation: a user sends a message, the chatbot replies. Even a sophisticated GPT-powered chatbot is fundamentally reactive: it waits, it responds, it waits again. Its world is the conversation window.
An AI agent is designed for action: it receives a goal, reasons about steps, executes tools (web browsers, APIs, databases, code runners), and drives toward a completed outcome. It doesn't need you to ask each next question. It figures out what needs to happen and does it.
We build both. Our AI chatbot development work covers assistants that answer from your own content, and our AI agent development work covers systems that act inside your CRM, inbox and databases. Many projects end up as a mix of the two, as explained below.
Side-by-Side Comparison
| Dimension | AI Chatbot | AI Agent |
|---|---|---|
| Primary purpose | Answer questions, hold conversations | Complete multi-step goals autonomously |
| Trigger | User sends a message | Goal assigned, schedule, or event |
| Output | Text response in a conversation | Actions taken, systems updated, outcomes delivered |
| Tool use | Optional, limited | Core capability: web, APIs, databases, code |
| Memory | Conversation window only | Short-term + persistent long-term memory |
| Autonomy | Low: needs user each step | High: runs independently to completion |
| Planning | Responds per message | Decomposes goal into sub-tasks |
| Build complexity | Low–medium | Medium–high |
| Typical custom build cost (our published ranges, excl. VAT) | £5,000–£15,000 for one use case; £15,000–£40,000 multi-channel | £5,000–£25,000 for a single-purpose agent; £25,000–£50,000+ multi-agent |
A Concrete Example: Customer Asking for a Refund
Same scenario handled by each type of system:
Chatbot Response
Customer: "I want a refund for order #4892."
Chatbot: "I understand you'd like a refund. To process this, please email our support team at support@company.com with your order number and reason."
Result: customer still needs to send an email. Human still processes the refund.
AI Agent Response
Customer: "I want a refund for order #4892."
Agent: [Looks up order #4892 → checks refund eligibility → confirms within 30-day window → processes refund via payment API → updates CRM → sends confirmation email] "Your refund of £47.99 has been processed. You'll receive it in 3–5 working days."
Result: fully resolved. No human involved.
The Hybrid Reality: Conversational Agents
In practice, the most effective business AI systems are hybrids: they have a conversational interface (so users can interact naturally), but they're powered by an agent underneath (so they can actually do things).
A customer-facing support system might look like a chatbot (users type messages, they get natural replies) but it's actually an agent that looks up orders, processes requests, sends emails, and updates records based on those conversations.
The distinction that matters is not the interface: it's the capability underneath. Can it take actions? Can it work without a human approving every step? If yes, it's an agent.
When to Use a Chatbot
Choose a chatbot when:
- The primary need is answering questions from a knowledge base
- Users need a guided conversation flow (e.g., lead qualification, booking enquiries)
- You need 24/7 first-response coverage for support tickets
- The interactions are mostly informational: no external systems need updating
- Budget is limited: chatbots are significantly cheaper to build and run
- Speed to market matters more than depth of functionality
When to Use an AI Agent
Choose an AI agent when:
- The task requires accessing and updating external systems (CRM, database, APIs)
- The workflow has multiple steps that require reasoning about what to do next
- You want the AI to work independently on a schedule, not just when a user is present
- The task currently takes a person real, focused effort each time it comes up
- High accuracy and auditability matter: you need to trace every action taken
- You're automating an entire process, not just a conversation within it
Technology Comparison
| Technology Stack | Chatbot | AI Agent |
|---|---|---|
| AI Model | GPT-4, Claude, Llama | GPT-4o, Claude Sonnet, Gemini |
| Framework | Botpress, Rasa, custom prompting | LangChain, LangGraph, CrewAI, custom Python |
| Integrations | Website widget, WhatsApp, Slack | Any API, database, browser, file system |
| Memory | Conversation history | Vector DB (Pinecone, Qdrant), SQL |
| Monitoring | Chat analytics | LangSmith, Langfuse, custom logging |
The Decision Framework
Use this to choose quickly:
Does the task require updating external systems?
No → Chatbot may be sufficient. Yes → Lean toward agent.
Does a human currently need to do multiple things to complete this task?
No (it's one step) → Chatbot. Yes (it's a workflow) → Agent.
Does the AI need to work when no user is present?
No → Chatbot. Yes → Agent.
Is the primary interaction conversational or goal-driven?
Conversational → Chatbot. Goal-driven → Agent (or hybrid).
Not sure which side your use case falls on? Book a free consultation and we will tell you honestly whether a chatbot, an agent or a simpler automation fits.