AI chatbots are no longer just for big companies. They’re becoming essential for small and medium-sized businesses that want to improve customer experience, support, and scale. With the right approach, a chatbot can help you sell more, support better, and make your content easier to access.
Why use an AI chatbot?
A chatbot can be a sales person, guiding potential customers toward the right products and options, making the buying process smoother. It can be a support agent, also assist existing customers by helping them get the most out of their purchase or by directing them to the right channels when issues arise.
It can also be a translator, even if your training data is in one language, a chatbot can translate and respond in real time, breaking down barriers for international customers or those who prefer to speak another language. And when it comes to accessibility, a chatbot can be a knowledgable guide, allowing the customer to use conversational interface making even the most organised knowledge bases easier to navigate.
Is your business a good fit for a custom AI chatbots
If you already have content like blog posts, FAQs, product specs, PDFs, or videos then you’re in a strong position to train a chatbot. Businesses that sell internationally or have a growing non-English customer base will benefit even more, as chatbots can handle multilingual conversations effortlessly.
If your product or service information is complex or scattered, a chatbot can simplify it into clear, direct answers, often with links for deeper learning. This makes your resources far more useful to customers.
Existing trainable AI chatbot solutions
Zapier has a product, currently in beta, which offers a simple setup and responsive configuration, but its free tier is very limited. Paid plans start at £20–30 per month and allow only 500KB of training data, which you’ll quickly outgrow.
Tidio integrates with existing chat systems and supports handoff to live agents. It offers generous training data storage, but customisation options are limited. Pricing starts at around £40 per month for 50 chats.
Botpress is the a super flexible option, with a powerful studio for complete customisation. However, it has a steeper learning curve for non-programmers. Its pay-as-you-go model means you only pay for AI compute, and most core features are free. Professional upgrades, like “always alive,” cost about £10 per month.
Pricing challenges
Some platforms, like Tidio, charge per chat session up front over the month, which means you could pay for chats customers never need. Others, like Zapier, charge mainly by training data size, making scaling difficult. Very few charge based on AI usage, this amount can vary depend on how much your audience interact with the bot, but generally works out cheaper. Botpress uses this method to charge, then upsell certain features to improve the experience should want to opt in.
How to train your chatbot
Start by defining its role and tone (i.e. “You are a polite, brief, and accurate AI Customer Support Assistant”). Decide what it should and shouldn’t do (i.e. “assist users with questions about products using only verified information retrieved from the Knowledge Base (RAG). You never generate answers from your own model knowledge.”), how long responses should be (i.e. “Try to answer all questions in just 2 sentences, if required, link to more than 2 resources at the end of the message.”), and how to handle questions it can’t answer (“If the information is missing, unclear, or insufficient, do not attempt to answer. Instead, provide a helpful response that acknowledges the limitation and offers alternative.”)
Then add your unique training data—blog posts, FAQs, product details. Avoid private or sensitive information. This is your unique element, perhaps no other chatbot has ever seen this data before, and no other chatbot can answer questions with this data anywhere else, which makes this tool so powerful, and a reason to get customers coming back to you.
You can also reference any data that perhaps you yourself haven’t written, but you know to be correct. Perhaps you can’t release this type of content as a blog post due to copyright infringement, but your AI chatbot can read these articles are reply to queries using that learned information.
Challenges (and solutions) for training an AI chatbot with Botpress
Workflows – at the back bone of Botpress is a very powerful workflow engine. Immediately you will be presented with a studio of tools and options, a user interface that you have likely never seen before. This can be overwhelming; but most options have full documentation, and even videos explaining what they mean.
Upgrades – Botpress studio itself is always being updated. this can make the documentation quickly out of date. The AI landscape is also always changing, the technology underpinning the chat bot is often in need of being updated to the latest versions. This can be tricky, but generally, allowing the upgrade won’t bring about any huge problem if you have not made extensive configuration changes.
Tips to reduce costs
Set up a multi-option welcome message with 2 or 3 of your most frequently asked questions. These questions can provide a plain text response that you have already written, rather than generating an answer each and every time. If the AI chatbot just responds to such an input with pre-defined text, you do not incur any AI spend.
Limit the response length in your instructions/prompts. Most sales bots do not need to give very robust answers, in fact, most customers do not want to wade through extensive answers. you can instruct the AI chatbot to aim to answer all queries in maximum 2 sentences, with links to resources if required.
Real world example
Below is a transcript of an AI chat bot in action, with commentary, to explain what the AI chatbot did and why.
When disassembling your desk top, unscrewing and re-screwing into the same holes can slightly loosen them over time. To mitigate this, avoid using pilot holes if you plan to frequently disassemble, as this allows the screws to grip more material each time. For more detailed guidance, check out our pilot hole guide .
Contact us today for help putting any of this into practice
Training an AI chatbot isn’t just about technology, it’s about strategy. Start with clear goals, use the content you already have, and choose a platform that fits your budget and technical comfort level. Done right, your chatbot becomes a powerful extension of your team, helping you sell more, support better, and make your content easier to access.