Every business can gain from combining generative AI and LLMs with company data, and Retrieval Augmented Generation (RAG) has been the most common technical approach to this. After all, if you could provide ChatGPT or another LLM full knowledge of your company – your products and services, clients, employees and expertise, past projects, and other valuable information – the potential is huge. Instead of scouring the intranet or paging through documents in search of a piece of information, employees would have a tool that could instantly answer questions related to the company – accessing organisational knowledge would be transformed.

Making huge steps towards this is absolutely possible, but the question is how. Microsoft provide a variety of approaches for “AI on your data”, from no-code Copilot GPTs, low-code Copilot Studio, mid-range Azure OpenAI “on your data”, to raw building block Azure AI Search – and even the latter is simplified by various GitHub “ChatGPT accelerator” solutions which can be used. Choosing the right approach can seem like a minefield – do you want to bring in data from Microsoft 365, Azure, SQL, a SaaS app, or simply a public website? Are you trying to provide a Copilot against a small knowledge base or a more expansive ecosystem of sites? Do you want to pay by user or by AI consumption? Should the experience surface in Teams, a Copilot plugin, or be embedded in an intranet or internet site?

This session aims to be a navigator through the Copilot and AI on your data maze, informed by battle scars from implementing all these forms of AI.