Using Custom GPTs to Create Innovative Client Deliverables
Ray Poynter. 7 April 2025
Reports, dashboards and PowerPoint decks are depressingly static deliverables, focused on the creator’s perception of the world, rather than being user centric. In this post I will highlight a powerful alternative that leverages AI. I am talking about using Custom GPTs to deliver a responsive and intelligent tool to the user.
A Custom GPT
OpenAI’s Custom GPTs are personalised GPTs, configured for a specific purpose. In the context of the use case I am talking about, the GPT comprises information from a research project, a set of instructions, and linkage to ChatGPT and its Code Interpreter & Data Analysis capability.
The information uploaded could be a report, could be the raw data, or it could be all the information relating to a project. The process of uploading information and linking it to the LLM creates a RAG, Retrieval-Augmented Generation model.
The instructions define how the LLM interacts with the information. For example, you want the LLM to prioritise the information uploaded over information it was trained on. Of all the steps involved in creating a Custom GPT, creating the right instructions is perhaps the one requiring the most attention (usually coupled with plenty of trial and error).
An Example
The video below shows a GPT that has been built with a publicly available data source, the World Bank Report on Social Protection 2025.
If you have a ChatGPT account, you can access the GPT in the video by clicking here. One constraint of OpenAI’s GPT’s is that you need to have an OpenAI account to access it This deliverable is therefore only of interest to clients and end-users who have OpenAI access.
Want to Learn More about Using GPTs for Market Research and Insights?
We are running a course on 9 April 2025 that will cover this use of GPTs and several other ways that market researchers and insights professionals can use them. You can book a place on our course by clicking here.