Chatbots answer questions on your website, in WhatsApp and in other channels, at any time of day. Yet many organisations get stuck on the same questions: what exactly is a chatbot, how does it work, what does it deliver and where does it go wrong?
This guide sets out the basics and refers, per topic, to an article that goes into more depth. If you want to know straight away how to have a chatbot set up on your own customer questions, read our page about AI chatbots.
Chatbots in short
A chatbot is software that holds a conversation with a visitor on its own, through text or speech. It answers questions, collects details and passes a conversation to a member of staff when needed.
The word is a contraction of chat and robot. The first chatbots worked with fixed scripts; modern chatbots use language models and also understand questions phrased differently from what was expected.
You will also come across the term virtual assistant. That usually means a broader assistant that carries out tasks in several systems, such as booking appointments or looking up a status. What the difference is and which forms exist is covered in what is a chatbot.
What happens between question and answer
A chatbot first recognises what the visitor means, then finds the matching answer and then decides the next step. That next step can be an answer, a follow-up question or a handover to a person.
With a rule-based chatbot that route is fixed. An AI chatbot uses natural language processing to recognise the intent, even with typos or an unusual phrasing. The answers come from a knowledge base with information about your products and processes.
How intents, knowledge base and integrations with your CRM work together is covered in how do chatbots work.
Buttons or free text
Not every chatbot needs AI. A click bot works with fixed buttons and leads the visitor through a conversation step by step.
That works well for predictable questions: opening hours, the status of an order or a simple appointment. As soon as visitors want to ask their question in their own words, a click bot gets stuck and an AI chatbot is the better fit.
When to choose which variant is covered in what is a click bot.
What a chatbot delivers for your organisation
The biggest gain from a chatbot is availability: visitors get an immediate answer in the evening and at the weekend too. Chatbots also take over recurring questions, so your team keeps time for questions that really need attention.
A well set-up chatbot does more than answer questions. It can ask about the visitor’s wish, record contact details and pass an enquiry with context to your sales team. That is how a question becomes a lead.
Which benefits you can expect and where they depend on the set-up is covered in the benefits of chatbots.
Where chatbots fall short
A chatbot is only as good as the information it runs on. If the knowledge base is incomplete or out of date, it gives incomplete or wrong answers.
With emotion, complaints and complex advisory questions, a chatbot quickly falls short as well. Visitors drop out if they are stuck in a loop or cannot reach a member of staff. A clear handover to a person, with the question and the conversation attached, prevents that.
How to safeguard the quality of answers and which pitfalls to avoid is covered in the disadvantages of chatbots.
Chatbot, live chat or both
A chatbot automates the conversation; with live chat a person answers the question. The choice depends on the kind of questions and on what a missed conversation costs you.
For predictable questions in large numbers a chatbot is efficient. For advice, bespoke work and valuable sales opportunities a person works better. In practice many organisations combine both: the chatbot handles the first questions and a chat specialist takes over where needed. What that looks like is explained on our page about live chat with chat specialists on your website.
How to weigh that up per situation is covered in chatbot vs live chat.
Building with a tool or with a specialist
You can build a chatbot yourself with a tool, or have it built and managed by a specialist. The difference lies less in the technology than in the work afterwards: filling the knowledge base, maintaining integrations and keeping conversations improving.
Building it yourself fits if you have the time and knowledge in-house and the questions are simple. If you do not have that capacity, or the chatbot also has to qualify leads, handing over the build and management is often the faster route.
Which costs and trade-offs are involved is covered in building a chatbot yourself or outsourcing it.
Making a chatbot smarter
A chatbot is never finished. After go-live you learn from real conversations which questions are asked, which answers do not land well and where visitors drop out.
Training means feeding those insights back into the chatbot: adding new questions, sharpening answers and adjusting the moments of handover. How to go about that is covered in chatbot training.
What works in practice, and what is coming
The chatbots that deliver the most have one clear goal. They answer the questions visitors really ask and know when to refer.
What you see changing: chatbots are becoming active across several channels at once, such as website chat and WhatsApp, and increasingly carry out tasks instead of only answering. Examples of what works and what does not, with the trends you can expect, are in the chatbot examples.
Where the chatbot stops and the person begins
In the conversations we hold at Bconnect we see the same pattern again and again: AI handles frequently asked questions well, but anyone who wants more leads from their website also needs people who ask follow-up questions and bring the conversation to an appointment or quote.
That is why we work with three models. With Operator Assisted AI a chat specialist holds the conversation, supported by AI. With AI First + Human Handoff the AI holds the conversation and a specialist or your own team takes over where needed. AI Only suits predictable questions and fixed processes.
If the AI cannot get any further, it announces the handover and the conversation goes to a person with the full context. Our chat specialists are available from 08:00 to 23:30, seven days a week, and the AI covers the hours outside that. How a lead carries on from there is covered in the guide on lead generation.