How to Brief an AI Chatbot Developer
A practical guide for UK business owners on preparing a clear, detailed brief for an AI chatbot or agent project. What information you need, questions to ask, and how to
Why a good brief matters for AI projects
An AI chatbot or agent works best when it knows exactly what job it needs to do. Unlike a website, which your customers browse, a chatbot is a tool that responds to what people ask it. Get the brief right, and your chatbot will capture leads, answer common questions, and qualify prospects. Get it wrong, and you'll end up with a tool that frustrates users and wastes your time.
A strong brief saves development time, costs less, and delivers better results. It shows your developer what problems you're solving, who your users are, and what success looks like.
Understand what your chatbot actually needs to do
Before you sit down with a developer, decide what job the chatbot will do. This is the most important step.
Common jobs for AI chatbots and agents:
- Answer frequently asked questions without your team getting involved
- Qualify and capture leads by asking the right questions
- Support existing customers with product information
- Book appointments or demos
- Handle simple transactions or orders
- Route queries to the right team member
Some businesses try to build one chatbot that does everything. That rarely works. Pick one main job first. You can add more later.
Ask yourself: what's the biggest problem this chatbot solves? What would your team do manually if the chatbot didn't exist? That's your answer.
Gather the information your developer needs
Your developer will ask for specifics. Have these ready:
Where will it live?
Will it sit on your website? In your email? On WhatsApp, Facebook Messenger, or another platform? Different platforms need different setups. Most websites start with a chatbot embedded on key pages.
Who will use it?
Describe your typical user. A prospect who's never heard of you? An existing customer? A mix? What do they need the chatbot to help them with? What questions do they actually ask?
What should it know?
List the information or documents the chatbot needs access to. This might be your FAQ, product specs, pricing, past customer queries, or your company handbook. The better the source material, the better the chatbot performs.
If you don't have this written down, that's your first job. Spend a few hours documenting your common questions and standard answers.
What should it NOT do?
Be clear about boundaries. Should it handle complaints? Refunds? Sensitive data? Should it ever transfer to a human? Under what conditions? A good chatbot knows when to say "I'll get a human to help with that."
What's the tone?
Should it sound formal or friendly? Professional or casual? Confident or cautious? Give examples of how you want it to communicate. This matters more than people think.
Prepare conversations, not just questions
Instead of a list of topics, write out sample conversations. Show your developer a real exchange between a prospect and an ideal answer.
Example:
User: "Do you ship to Scotland?"
Ideal answer: "Yes, we ship to all UK addresses. Orders under £50 cost £3.99, over £50 ship free. How can I help?"
User: "How long does it take?"
Ideal answer: "Orders placed before 2pm ship same day. You'll usually receive it within 2 working days."
These examples help the developer understand the level of detail, the tone, and what counts as a good response. They're more useful than bullet points.
Agree on the outcome and how you'll measure it
What does success look like? Be specific.
- "Answers 80% of questions without escalation" is measurable. "Improves customer satisfaction" is vague.
- "Captures 20 qualified leads per week" is concrete. "Generates more leads" is not.
- "Reduces support emails about opening hours by half" is testable.
Agree on how you'll track this. Most chatbots have built-in analytics. You should review these monthly.
Set expectations about limits
AI chatbots are powerful, but they're not magic. They work best for questions with clear, factual answers. They struggle with nuance, complex emotion, or truly unique situations.
Be honest about your source material too. If your FAQ is old or incomplete, the chatbot will reflect that. The first month, you'll probably spot questions it gets wrong. This is normal. You'll fine-tune it based on real conversations.
Most chatbots improve over time as they learn from actual interactions.
Share this with your developer
When you're ready to brief, make sure your developer gets:
- A one-sentence statement of what the chatbot does
- Who uses it and what problem it solves
- Where it will live
- Sample conversations (3 to 5 realistic exchanges)
- A list of documents or FAQs it needs to know
- Tone and style guidelines
- Success metrics and how you'll track them
- Any topics or situations it should avoid
Put this in a simple document. A page or two is enough. Avoid lengthy PDFs. Clear beats comprehensive.
Next steps
A good brief takes a few hours to prepare, but it saves weeks of back-and-forth. If you're unclear about any of this, your developer should help you clarify it. That's part of their job.
Once your brief is ready, you're in a strong position to start. You'll know what you're building, your developer will know what you need, and you'll have a way to measure whether it works.
If you're ready to explore an AI chatbot or agent for your business, get in touch. We can walk through your brief, answer questions, and explain how we'd build it.