Now Everyone’s an Einstein
AI. You couldn’t miss it at Dreamforce. It was everywhere — signage, sessions, and news coverage. As dawn rose on the first day of Dreamforce (Tuesday, Sept. 12), AI product announcements were already flooding social feeds as the news embargo broke hours before Salesforce CEO Marc Benioff’s keynote address at 10 a.m. PST.
Now everyone’s an Einstein
What should we expect? Is AI the new crypto? Let’s not forget the blockchain announcement from Salesforce just a few years ago — you know, in case you want to mint an NFT using Salesforce.
Benioff on AI: Don’t Be Left Behind
Marc Benioff AI keynote
Marc, the consummate ringmaster, kicked things off with a jarring approach I had not seen before. The messaging was uncharacteristically negative. And by “negative,” I mean unbridled fearmongering that AI could create the kind of dystopian society we’ve seen in films. Springboarding from that chippy picture, Marc segued into an urgent appeal for companies to get on board with AI, and fast — don’t be left behind! Projections point toward seismic shifts over the next few years. For example, three out of four companies will adopt AI by 2027, and 30% of an employee’s time will be freed up due to generative AI by 2030. Even now, 92% of businesses are seeing a return on their AI investments.
The AI opportunity
AI and Matters of Trust
While Marc’s position on AI adoption was clear, he also acknowledged that the technology is far from perfect. Large language models (LLMs) need data — lots and lots of data. Even with lots of data, AI still runs risks such as hallucinations, bias, and toxicity. To that point, I think Williams-Sonoma President & CEO Laura Alber nailed it at the keynote when she said, “There can be nothing artificial about the customer experience. You need it to be authentic.” That’s a tall order for AI.
AI trust gap
Exploring the Architecture
After making the business case for AI adoption , we drilled down to the details with some architecture slides — the Einstein Trust layer, Einstein 1 and its metadata framework, and Copilot and Skills.
Einstein Trust layer
Einstein 1 architecture
Of course, you can’t have a keynote without demos. Einstein AI can start an email, refine it with images, transcribe a phone call, suggest a campaign or relevant pdf, prompt a user to update a record, and summarize a weeklong Slack conversation while you were out of the office.
So, what’s needed to get started with AI? Did I mention data? For Einstein to have enough context to respond, you might need to connect it to more data sources. It also needs to be trained using Einstein Builders — Copilots, Skills, Prompts, Apps, and Flows — to help Einstein pinpoint the right answers. Otherwise, it might not be able to suggest anything to a user. In theory, a new service agent could leverage well-trained AI to apply equivalent knowledge of a service agent with years of experience. That’s the promise — but it requires a lot of behind-the-scenes work in order to train Einstein and make it a reality.
How Data Cloud works
What if the data Einstein needs isn’t in your CRM? That’s where the Data Cloud comes into play. With Data Cloud, you can purchase Mulesoft to connect to other databases, or you can now BYOL (bring your own data lake) and connect to Snowflake or Databricks.
Sales Slack Elevate
New Features in Slack
The keynote concluded with a demo of the new features in Slack. Years ago, Marc said Slack would be the new UI of Salesforce. Sales Slack Elevate made good on that promise with its ability to display core CRM sales objects in Slack. Impressive, but to me they have it backward. Wouldn’t it be more beneficial to have Slack channel conversations visible inside Salesforce? I suspect having Slack close out the keynote was necessary as there has been a lot of criticism around this $27.7-billion acquisition. You could feel it in the room as the Slack demo ended and the presenter said, “It’s OK to clap.” Awkward!
In closing, you had to be impressed with what Salesforce was able to have demo-ready by Dreamforce. The rumor was that Salesforce was releasing new software builds right up to the keynote. I’m sure there are a lot of software engineers behind the scenes we have to thank for what we saw that morning.
The keynote messaging was definitely targeted toward enterprise clients. We’ll have to wait to see if there is any real utility for smaller businesses that won’t have (or can’t afford) the data infrastructure needed to train and leverage Einstein 1. Did it meet the hype? I’d say Salesforce did a good job demonstrating they’re squarely in the AI game.
PS — This blog post was written without the assistance of AI. 😉
Shell with Einstein