Easter Already?
The start of this year has been a busy couple of months for me, as we tackled the next phase of Credera/Omnicom’s acquisition of BrightGen - becoming One Credera. That is all firmly in the rear view mirror now, and while I’m still battling the muscle memory of typing my old email address everywhere, I’m able to focus a lot more of my time back where it belongs - Salesforce and related technology. So has anything been happening while I’ve been away?
Salesforce
Spring ‘24
The Spring ‘24 release of Salesforce went live in early/mid February with some fine new features like the Null Coalescing Operator. But the real excitement for me was in the beta and preview features - Scratch Org Snapshots, Zip Handling in Apex, and Dynamic Formula Evaluation in Apex all look really cool, and you can play around with all of them in scratch/developer orgs.
One item that didn’t go GA with the Spring ‘24 release was Prompt Builder, and I was expecting a delay at the end of February as there was so little noise about it on the socials, whereas Marc Benioff would usually be putting out all sorts of teasers. As usual, what do I know? Taking advantage of the extra day as we’re in a leap year, Prompt Builder burst into GA on 29th February, while Einstein Copilot entered beta at the same time.
TrailblazerDX
It felt like the dust had just settled on the Prompt Builder and Einstein Copilot news when we heard it all over again, and more, at the TrailblazerDX keynote. The demos were a lot better than I’d seen before, and put a great story on how Generative AI can help Salesforce customers at a high end resort hotel. Sadly the keynote is probably the closest I’ll get to Turtle Bay Resort, but we can’t have everything (where would we put it?).
We also saw a slight change in the messaging around the role of humans in the brave new world of AI - it’s now Human at the Helm, rather than AI needing all it’s work approved before it can be used.
There were a number of interesting announcements at TrailblazerDX - the Einstein Trust Layer becoming a little more accessible/configurable, Einstein for Developers now having far less onerous terms of use, and a wealth of roadmap information for Lightning Web Components and Apex. You can see more of this via the slide deck that I presented at the March event of the London Salesforce Developers community group - the recording will take a little longer but will be available on our Youtube channel in the fullness of time.
There’s a whole bunch of recorded sessions available on Salesforce+ and Youtube - all you need is a couple of hundred hours to work through them all!
Farewell Workflow/Process Builder
The day everyone knew was coming is here - Salesforce have announced the end of support for Workflow and Process Builder on December 31st 2025. These tools have been faithful servants over the years, but they are showing their age and limitations, and there are better options available now. While the end of 2025 might sound like a long way away, it’s only 21 months, so get cracking on your migration to Lightning Flow.
AI
There’s been tons of news on the AI front - new models abound, Google’s Gemini got off to a torrid start, but seems to have settled down. OpenAI teased content from Sora, its impressive AI video generator, but were unable (or unwilling) to explain what content it had been trained on.
We’ve been introduced to Devin - “the first AI software engineer”. The videos accompanying this are impressive - apps going from concept to production with little human effort required. Is this the end for developers? Not according to Reddit. Feedback is that while it is good at creating new apps, it struggles with existing ones because of the context required to understand an app, and the fact that this may not exist in the codebase. Like many LLMs - good at following instructions, not so good at solving problems.
The Stack Overflow podcast had a great episode recently, entitled “Is AI making your code worse?” - in this, Bill Harding, CEO of GitClear, discussed the quality of AI code and the impact on productivity. So far so familiar. The difference here was not limiting to the perspective of how quickly can AI turn your instructions into code. Rather they looked at how quickly AI generated code needed to be changed after it was committed (code churn), whether it was generating code that already existed elsewhere in the repository (churn code), and if it could act like a real developer and find some existing code that could be reworked to handle the new use case. Well worth a listen.
Air Canada probably had the most column inches in the AI space, where in spite of their best efforts they were held responsible for advice given by their chatbot. Tips to avoid this in future include the chatbot establishing at the start of every conversation that it may be worse than useless : "Experts told the Vancouver Sun that Air Canada may have succeeded in avoiding liability in Moffatt's case if its chatbot had warned customers that the information that the chatbot provided may not be accurate.". I’m not sure how quickly AI will take all our jobs if it starts every interaction advising that it can’t be trusted!
Me
I’ve been playing around with Prompt Builder and Copilot quite a bit recently:
As these features are either very new or in beta, they are a little more challenging than usual to work with, and more than a little picky about what Apex they will accept. The fact remains it’s a very exciting time to be in the technology industry, and I can’t wait to see what comes next!