Salesforce AI: Overrated or Game-Changing?

Salesforce AI up to this point has been quite underwhelming and I dare to say overrated. With the announcement of Agentforce at Dreamforce this month, I think it will remain overrated and not worth the increased costs for the majority of Salesforce customers.

Disclaimer:
I have only read about the Agentforce release and have yet to implement it on a project.

Why I feel Salesforce’s AI products are overrated (as a Salesforce developer with the Salesforce AI Associate and AI Specialist certifications):

• Agentforce — seems to be largely a rebrand of Einstein (products such as model-builder and prompt-builder), with at least 80% of it already existing but now more grouped together (which I guess is a decent positive).
• From what I’ve played with, it’s not worth paying for unless you check a lot of boxes already. It’s not a gamechanger for organizations until it’s nearly available to utilize by default for most license types without having to purchase upgrades (on top of already existing expensive licenses).
• Pricing is quite steep. I believe almost all of the ‘cool’ features at this point require Data Cloud, which I don’t think is cheap. Data Cloud is both consumption-based (and currently very hard to estimate) and has a higher barrier to entry (you’re starting at a minimum tier of cost annually, rather than purely pay-as-you-go). There’s still some ambiguity around the actual cost for Agentforce. @ $2/conversation, it’s pretty high, but it’s also unclear whether there will be an attached subscription-based cost as well (will we need to buy Einstein for Sales/Service to get access to Agents?).
  • A lot of the features thus far make sense behind the Data Cloud paywall because almost everyone is just going to utilize OpenAI models, making it more of a fancy OpenAI integration wrapper.
• Prior to Data Cloud & Agentforce, you could already gain a lot of leverage using Einstein chatbots (with some machine-learning capabilities) to answer repetitive questions, eliminate question needs, or stall for available agents. So if that was most of your Salesforce use already, and you don’t have Data Cloud, it doesn’t make sense to pay up and roll out Data Cloud.

Where Salesforce AI / Agentforce makes sense to me & how it should be available to customers:

• A big Ecommerce or other support-type sales environment where you’re already reaching the point where you need to bring together all your data anyway, and you already have an existing Salesforce org, AND you have good data people but no ML engineers or data scientists… THEN Data Cloud and Agentforce make sense. You can potentially leverage it in 3 different directions all at once.
  • If you already employ some combination of data scientists and ML engineers, implying you’re already probably paying for Big Data services, and your employees have learned all sorts of open-source tools and suites, I just don’t see Agentforce being anything special for you yet.
• Salesforce AI features should be made more accessible. For example, if you bought CRM Analytics (CRM_A) from Salesforce and have a certain number of licenses, you should have access to Model Builder as well.
• OR, if you are a large user of Service Cloud, your existing chatbots should be “supercharged” with Agentforce once you turn it on and do some simple build-outs.

Conclusion:

While I’ve played with all prior Einstein AI features (and passed all the Salesforce AI certifications), I haven’t yet implemented Agentforce. Once I do, I will make another blog post to see if my hunch is proven correct. Also, stay tuned for AI implementation options for small to mid-size businesses.