Dreamforce is now settling into the rear view mirror. I’m pretty much caught up on everything that was waiting for me when I got back, so here’s my thoughts (which I know you’ve all been desperate to hear).
As you may have seen on the socials, it was mostly announcements about AI, which I’ll be calling AInnouncements for the duration of this issue, and maybe for longer if it doesn’t get a completely negative reaction. Starting with the relaunch/rebrand of all things AI as Einstein1. Spare a thought for AI Cloud, which must have been the shortest lived product name Salesforce ever had - quite the achievement given how keen they are at changing names. It’s also interesting to see that the suffix ‘1’ is back in favour at Salesforce, pretty much a decade after everything was rebranded Salesforce1, only to be changed back the following year. Maybe the name will stick, but I’m not convinced. I get the need to move away from “cloud” style messaging for Einstein, as it’s not a product in its own right, more a collection of features that get added to each product. Tacking ‘1’ on the end doesn’t scream supercharged performance though, so I won’t be surprised when the inevitable renaming occurs.
I’m also feeling the same way about Einstein Copilot - not because it’s a bad name, but given Github/Microsoft are already using it, I’m surprised that Salesforce went for it in the first place. Personally I’d prefer fun names like Henchman, Accomplice, or even Chairman of the Council of People's Commissars, but I can’t see those catching on in enterprise software. In keeping with the 1 suffix, Sidekick was a popular suite years ago so maybe we’ll see that return. The concept is again sound - a single term to refer to the many assistants that will be striped across the entire product set so they don’t have to be explained individually. If you are curious as to the individual capabilities and the licenses that you need, treat yourself to this 18 page cheat sheet(s!) that covers all of it.
The big AInnouncement prior to Dreamforce was in August with the general availability of Einstein Studio - a single control panel for all of your AI models that also allows you to train up a new model from the information that you hold in Data Cloud. There was a frisson of excitement that this might include generative AI, but it’s still a bit early for that and these are around classification and predictions. What’s interesting about this is we are seeing some consolidation of the user interface around all things Einstein AI, while also seeing a proliferation of models rather than a single mega-model that knows everything. At Dreamforce we saw this renamed to Model Builder and rolled into Copilot Studio along with Prompt Builder and Skills Builder.
What I do think is interesting right now is that a lot of research is going into smaller LLMs that can run on end-user devices - laptops and mobile phones - rather than requiring giant GPU farms. To use Salesforce’s marketing against them a little, this would truly democratise generative AI. Fantastic for consumers, but a headache for enterprises that like the idea of centralised AI with appropriate controls and guidelines. Salesforce are very much going for the trusted and controlled AI approach, and at this moment who would bet against them?
If you can’t wait or afford to access the Salesforce integrations, it’s easy enough to integrate directly with OpenAI through something like the Salesforce CLI, although bear in mind you won’t have anything like the Einstein Trust Layer protecting you:
Finally, if all this talk of AI is giving you FOMO but you don’t know where to begin, checkout the BrightGen webinars I’ve hosted on the topic: