What is an AI chatbot?

An AI chatbot is a conversational interface built into your app that can understand natural language questions and respond intelligently — without a human on the other end. Unlike the scripted chatbots of five years ago (which followed rigid decision trees), modern AI chatbots use large language models like Claude or GPT-4 to understand context, handle follow-up questions, and give genuinely useful answers.

For a business, this typically means one of two things: a customer-facing assistant that handles common questions, or an internal tool that helps staff search documentation, fill out forms, or navigate complex processes.

A tradie supply company in the Southern Highlands, for example, might build a chatbot that lets customers ask "what's the right drill bit for concrete?" and get an accurate, product-specific answer — available at 2am, without a staff member.

When does your app need an AI chatbot?

• Your support team spends significant time answering the same questions repeatedly
• You have a large knowledge base (product catalogue, policies, FAQs) that users struggle to navigate
• Your app serves users across time zones or outside business hours
• You want to qualify leads before a human picks up the conversation
• Staff need to retrieve information quickly from internal documents or manuals
• You're building a product where personalised guidance is part of the core value proposition

How much does it cost?

Adding a working AI chatbot to an existing app typically adds 13–27 hours of development — roughly $2,000–$6,000 AUD at Australian boutique agency rates.

What drives cost up:

• Training on your data — a chatbot that knows your specific products, policies, or documents requires retrieval-augmented generation (RAG) setup, which adds significant work
• Conversation history — storing and using past messages so the bot remembers context across sessions
• Multiple channels — deploying on web, mobile, and WhatsApp simultaneously
• Handoff to human — detecting when a conversation needs escalating to a live agent

What keeps cost down:

• Using a hosted AI API (Claude, OpenAI) rather than running your own model
• Starting with a focused scope — one use case done well, rather than everything at once

How it's typically built

Most app chatbots are built using a hosted AI API (Anthropic's Claude or OpenAI's GPT models) rather than training a model from scratch — that would cost ten times as much. The developer writes a system prompt that defines the bot's behaviour and knowledge boundaries, connects it to your data (either via RAG with a vector database, or by injecting relevant content into the context window), and builds the chat UI.

For apps that need the chatbot to access live data — checking an order status, booking an appointment — the AI is connected to your app's existing APIs via tool calls.

Australian businesses should consider data sovereignty: make sure your provider stores conversation data in a way that complies with the Australian Privacy Act, and that you're not inadvertently sending sensitive customer data to offshore servers without appropriate safeguards.

Questions to ask your developer

• What model are you using, and how does it handle confidential data? Some models log conversations by default — this matters if users share personal information.
• How do you handle hallucinations? A good implementation constrains the bot to your data sources and makes it say "I don't know" rather than invent an answer.
• Can it be updated without a code deploy? You want to be able to update the bot's knowledge base as your products and policies change.
• What's the ongoing API cost? AI API usage is metered — understand what you'll pay per month at expected usage volumes.
• How do you measure if it's working? You need analytics on question types, escalation rates, and satisfaction scores to improve the bot over time.

---

See also: AI text generation · Document extraction & analysis · App cost calculator