ChatBot FAQ's

A rule-based chatbot follows a fixed decision tree — it only responds to specific keywords or button clicks and breaks down the moment a user says something unexpected. An AI chatbot uses a large language model (LLM) to understand what the user actually means, regardless of how they phrase it. It can handle ambiguity, maintain context across a conversation, and respond sensibly to questions it has never seen before. All of the chatbots we build are AI-powered.

A standard chatbot responds to what you say. An AI agent reasons about what needs to happen and then takes a series of steps to make it happen. For example, a chatbot might tell you your order is delayed. An AI agent would look up the order, identify the delay reason, check whether a replacement can be dispatched, and either take action or escalate to a human — all without being explicitly told each step. Agents are suited to more complex, multi-step tasks where the path to an answer is not always predictable.

In most cases, no — and that's not usually the goal. Chatbots are most effective at handling high-volume, repetitive interactions that would otherwise consume your team's time: answering the same FAQs, collecting information from customers before a call, or handling out-of-hours queries. This frees your staff to focus on work that genuinely requires human judgement and relationship. Many of our clients find that a well-built chatbot makes their team more productive rather than smaller.

We primarily build on models from OpenAI (including GPT-4o) via Azure OpenAI Service, which gives enterprise-grade security and compliance. We can also work with other providers such as Anthropic Claude or open-source models deployed privately, depending on your requirements around cost, data residency, or integration with existing Azure infrastructure.

There are two main ways. The first is a system prompt — a set of instructions we write that tell the AI who it is, what it knows, and how it should behave. The second is a document store (knowledge base) — your existing documentation, FAQs, product information, or policies are processed and stored so the chatbot can search them in real time and base its answers on your actual content. These two approaches are usually combined for the best results.

RAG stands for Retrieval-Augmented Generation. Instead of relying solely on what the AI model was trained on, RAG lets the chatbot search your own documents for relevant information before generating a response. The result is a chatbot that answers from your actual content — current, accurate, and specific to your business — rather than guessing or making things up. It also allows the bot to cite sources, so users can verify the answers they receive.

Yes. One of our core services is database and system integration. The chatbot can be connected to your CRM, order management system, booking platform, product catalogue, or any system that exposes a database or API. This means the chatbot answers with live data — not static information that may be out of date.

A well-designed chatbot should never pretend to know something it doesn't. We build explicit fallback behaviours into every bot — typically the chatbot will acknowledge it cannot answer, offer an alternative (such as a link to a relevant page or a way to reach a human), and optionally log the unanswered question so you can improve the knowledge base over time. Graceful failure is a first-class part of the design, not an afterthought.

A straightforward FAQ or information chatbot can typically be designed and deployed within two to four weeks. More complex implementations — those involving database integrations, custom workflows, or AI agents — generally take four to eight weeks depending on scope. We will give you a realistic timeline during the design phase, before any work begins.

We build on Microsoft Azure, which means your chatbot and its data are hosted in secure, UK-based datacentres where required. Azure's compliance certifications cover a wide range of regulatory requirements including GDPR, ISO 27001, and NHS Digital standards. We can also deploy into your own Azure subscription if you prefer to retain direct ownership of the infrastructure.

Yes. We use Azure OpenAI Service, which means your data is not used to train OpenAI's public models. Conversations are not shared with third parties. All data is encrypted in transit and at rest. We advise on and implement appropriate retention policies, access controls, and logging so you have full visibility of how the chatbot is being used and what data it is handling.

Yes. Where we implement a document store (knowledge base), we put in place processes and tooling to make updates straightforward — uploading a new document or replacing an old one re-indexes the content automatically. For system prompt changes or workflow updates, these are managed through a simple change process. We also offer an ongoing maintenance retainer if you would like us to handle updates on your behalf.

Pricing depends on the scope of the build. There are typically two components:

• A one-off build fee — fixed price, agreed after the design phase, covering design, development, testing, and deployment.
• Ongoing running costs — cloud hosting and AI API usage, which scale with the volume of conversations your chatbot handles. We provide a usage estimate during the design phase so there are no surprises.

Get in touch for a free initial consultation and we'll give you an honest, no-obligation estimate based on what you actually need.