Differentiation Manifesto

1. 360° Customer View & Contextual Conversations

Benefit: Every customer interaction (calls, tickets, chats, emails) is logged in one place — the HubSpot contact timeline.

Value: Agents see the full history instantly, respond smarter, and avoid repeating questions like "Have we spoken before?"

2. Seamless Ticketing + Calling Workflow

Benefit: Aircall calls automatically create or update tickets in HubSpot, and link to the right contact/deal.

Value: No need for manual logging → faster resolution, better SLA tracking, and fewer things falling through the cracks.

How does it work?

While Large Language Models (LLMs) like ChatGPT or Claude strut around showing off their impressive party tricks, they have one rather embarrassing weakness: they're not exactly what you'd call reliable fact-checkers, resulting in nonsense or generic responses. Not ideal for handling frustrated customers.

This is where RAG, Retrieval-Augmented Generation, comes into play. In Hubspot's Service Hub, you can share certain Information such as Documents (PDFs), websites, blogs, as well as Knowledge Base Articles and Files, supercharging the LLM behind the Chatbot with valuable context. All the information provided is stored in a so-called multidimensional vector database, which makes it digestible for a connected LLM. And that makes a huge difference. So, the information provided to the AI Agent is Key for its perception. 

You won't have to worry about that in detail, but if you are interested, here's how it works:

1. User types in a message.
2. Chatbot searches the database for related information
3. Database returns data relevant for this request. e.g. if the User is asking for a return process, that could be a website describing that process.
4. Question and the context information, e.g. the website describing returns + parameters regarding tonality, etc., is forwarded to the Language Model.
5. LLM responds with an answer to the Prompt.
6. Answer is forwarded to the User.

Again, the Key here is Context. The more contextual information that can be shared with the LLM, the better and accurate the response.