Project Catalog
The goal of this newsletter is to educate its readers in how enterprise applications like CRM can take advantage of AI capabilities. The posts tend to either be thought pieces (less often) or project pieces, where I walk through how to build a specific capability – and provide the readers with the finished results that they can experiment with themselves.
To date, the following projects have been completed and can be used as reference for reader projects:
1. Getting Started with A Simple OpenAI integration
Two posts that got you up and running calling OpenAI from Salesforce, starting from absolutely nothing. Focusing on the bare minimum needed:
2. GPT Life Insurance Recommender
A project that shows you how to get GPT to recommend the best type of Life Insurance for a prospective customer. This post gets a simple version up and running:
A bit later on, there was a discussion on how to get the results back as JSON rather than just as free text for a more powerful implementation:
3. Named Credentials to Store OpenAI Credentials
A discussion on how to safely and security manage your OpenAI API key inside of Salesforce to minimize the chance of mis-use.
4. Using Google’s LLM instead of OpenAI’s.
A demonstration of using Google’s PaLM 2 model instead of OpenAI’s GPT models.
5. Building a Sales Idea Generator
This demonstrates how to take a bunch of data about an opportunity from Salesforce and use it to have GPT generate ideas to help move the sales process along.
6. Building a Templating System for Data Binding
Some have called this “data grounding”, but the idea is to have a sufficiently powerful way to merge live data from Salesforce into your GPT prompts so that the results are specific to and appropriate for a particular opportunity, case, etc. Two posts walk through the design of such a templating system as well as an engine to handle the interface between Salesforce and OpenAI (all of which I wrote and is available for free in GitHub).
7. A Sales Email Writer
A demonstration of taking data about an inflight opportunity and having GPT write an email to a specific contact on a specific topic as directed by the user. A series of three posts discuss the construction and working of the demo.
