What Problem Did This Sales Coaching Automation Solve?
Shane opens by framing the core problem: sales coaching has traditionally been subjective. When a rep reports back on a call, the account of what happened is their own recollection, not an objective record. For years, the team followed the Sandler sales process and brought in outside consultants to coach reps, but the feedback was only as accurate as what the rep remembered or chose to share.
How Did They Build the Sales Coaching Custom GPT?
The team took two steps. First, they began recording and transcribing every sales call to create a reliable data source. Second, they built a custom GPT inside ChatGPT and trained it on their specific sales process and desired outcomes. Reps could then paste a call transcript into that model and receive structured feedback covering what they did well, what they missed, and what qualifying actions to take next. The output used the actual words spoken by both the rep and the prospect, not a summary from memory.
Why Use a Custom GPT Instead of Standard ChatGPT?
Hunter explains that a standard ChatGPT session treats every query as a fresh search. You have to re-explain the format, the context, and the desired output each time. A custom GPT stores all of that upfront. Once it is trained on your call log format and your evaluation criteria, you simply attach the transcript and submit. The model already knows what to do. This also means the model continues learning from each interaction rather than discarding it.
Shane adds a related consistency point: relying on end users to remember and correctly type the same prompt every session introduces variability. The custom model removes that dependency entirely.
What Are the Privacy and Compliance Considerations?
Shane flags that any business using ChatGPT with real customer or call data should be on a private instance, not a public one. The public learning model behaves like a broad web search and may incorporate your inputs into shared training data. A private instance keeps proprietary information contained, and a custom GPT within that environment narrows the scope further to only the parameters you define.
What Other Use Cases Work Well for Custom GPTs?
Hunter shares two additional examples. The first is personal: he uses a custom GPT to generate a weekly grocery list based on his meal preferences, showing the range of flexibility available. The second is business-focused: building custom GPTs trained on the documentation for specific software tools your team uses. Instead of searching vendor docs or calling support, a rep can query the custom model directly and get a targeted answer.
How Can You Extend a Custom GPT With CRM and API Integrations?
Shane describes where the team has taken their original sales coaching model after roughly 18 months of use. They built a front-end web interface that connects to HubSpot via API. This means the model can now pull live CRM data, allowing the team to ask questions like which company a contact is associated with or whether a prospect already exists in the system. The pattern they recommend is consistent: start with a basic, functional model and layer on integrations as the use case matures.
Hunter confirms that their team can also build these custom models for clients who want the same type of functionality tailored to their own business processes and connected to the tools they already use.
What Should You Define Before Building a Custom GPT?
Shane closes with practical guidance for anyone starting out. Before building, think through how many people will use the model, how they will interact with it, and what consistency you need across those interactions. From there, define how the prompt should be structured, what parameters the model should follow, and what the output should look like. Getting those decisions made upfront determines how useful the model will actually be in practice.