Leveraging AI to accelerate your Salesforce project: Go from "Requirement" to "Ready for QA" in 15 minutes!

#Salesforce #AI #OpenAI #ChatGPT

Authors: Paco Fálder , Andy Forbes

Would you like your Salesforce projects to move each requirement to a user story, then to development, and then to "ready for QA" more quickly? Have you been hearing about AI and are looking for a use case with a clear and obvious ROI? There's no reason to wait any longer. With the right tools and processes in place, your Salesforce projects can go from requirement to ready for QA amazingly quickly – for some requirements, literally within 15 minutes. By embracing AI, you can reduce your ideation-to-production cycle from months to weeks or even to days. Let's explore how you can make it happen!

Let's start with a requirement that one of your company's Customer Service Agents (CSA) should be able to reassign cases to other CSAs or a Salesforce Queue where the case can be appropriately addressed.

As ChatGPT notes, you may want additional acceptance criteria. This can be achieved by having your team write the additional acceptance criteria themselves, fine-tuning the current prompt and running it again, or writing additional prompts focused on error handling and the user interface. For this article, though, let's run with the ChatGPT user story as-is.

The next step is to post the user story to ChatGPT and ask for the metadata and code.

Now things get interesting. ChatGPT is trained using content on the Internet; the content on the Internet is not always fantastic. Some architecture/design decisions in the above could benefit from some discussion. Like the earlier "requirement to user story" ask of ChatGPT, what ChatGPT produces when going from user story to code can be rewritten by your team, can be updated by fine-tuning the current prompt and running it again, or augmented by writing additional prompts.

Picking just one of the concerns, let's have ChatGPT rewrite what was produced to use Lightning Web Components instead of Aura.

So what can we conclude from this exercise?

Is what ChatGPT produces exactly right the first time? No. Is what ChatGPT produces a solid starting point for refinement? Yes. Can what ChatGPT produces be refined in whatever fashion – manually, via prompt refinement, or via additional prompts – your team believes will provide the most beneficial path to getting work product to the QA team? Absolutely!

Right now, today, it is clear that AIs like ChatGPT can reduce Salesforce development time. The increase in the capabilities of AIs is accelerating, and access to better and better general-purpose AIs like ChatGPT will shortly be available to everyone. Beyond the general purpose AI capabilities, there will be problem domain middle layers that focus on project discovery, writing epics and user stories, configuration and development, quality assurance, deployment, and operations.

We're headed to a world where AIs:

• Monitor the operations of a system and write requirements to improve the system,
• Take the requirements and write user stories,
• Take the user stories and do the configuration and customization,
• Take the user stories, configuration, and customization and perform quality assurance,
• Deploy the configuration and customization to training, staging, and production environments, and
• … the process repeats.

This change isn't going to happen overnight or even in 2023, but it is going to happen. Now is the time to start using the current AI capabilities and building an understanding of how to leverage AI for your Salesforce projects!

Be not the first by whom the new are tried, Nor yet the last to lay the old aside. - Alexander Pope