Getty Images/iStockphoto
Salesforce Agentforce: An answer in search of a question?
They used to be named Einstein Copilots. Now they're Agentforce autonomous agents. Salesforce's latest upgrade might be more than words.
Salesforce's Agentforce autonomous AI platform went live last week after much Dreamforce fanfare in September.
Now that customers can actually use the much-hyped Agentforce platform, they must consider the updated generative AI bot's practical applications and business value for their organizations.
"We believe that our computer scientists have delivered something that's extraordinary, but it doesn't matter until customers use it and get value from it. It just doesn't matter," Salesforce CEO Marc Benioff said at Dreamforce in September. "It doesn't matter what the benchmark is, or it doesn't matter [if it's a] very cool science project … if I can't bring value to my actual business in my life in a material way."
TechTarget Editorial asked a broad swath of Dreamforce attendees, including, customers, analysts, partners and executives the question: Is this truly a big-bang feature that users need?
Customers get CEO mandates
Across the country, many IT departments are feeling pressure from senior execs to bring generative AI tools to CX and expect CX leaders to figure out how. One such company is cloud security vendor F5; generative AI from Coveo is already live within its Salesforce Service Cloud, said Laurel Poertner, F5 senior director of digital services.
The goal of the Coveo implementation is to answer customer questions while they're using the software with better search and summary of technical help. If it doesn't answer their questions in line, the generative AI also will help F5 agents more quickly solve those issues. She said that searches for information have gone down by half at F5's service site since they turned this on about a year ago, she said.
"We were fighting the urge to say, 'I've got this tool that has this capability. Go figure out what to do with it,' rather than, 'I have a need to solve a problem. What will help me solve this problem?'" Poertner said.
As far as Agentforce goes, it's early days. F5 will likely figure out how to update its decade-old, rules-based service chatbot, whose responses are "kind of old school," she said.
At global publisher Wiley, the focus for potential Agentforce applications will be to personalize service at a more granular level. ChatGPT has raised expectations for AI, said Kevin Quigley, director of process improvement. The company signed up for the Agentforce pilot program to meet customer demand for better personalization.
In early tests, generative AI tools have shown that they can at least do a better job of doing that but not go too far.
"Our [AI] philosophy, we kind of call it internally in the Customer Service Group, is 'give up easily,'" Quigley said. "In other words, we don't want the risk to be too high. We want to prioritize the customer experience. And if we feel that's at risk for any reason -- be it a capability limitation, a subjective scenario, whatever -- we give it over to a human pretty easily."
At Wyndham Hotels and Resorts, Agentforce brings an attractive feature for companies like itself that are heavily invested in Salesforce Sales Cloud, Experience Cloud, MuleSoft, Tableau and Data Cloud: tight integration. Wyndham's customer data and franchisee data commingled in Data Cloud can paint a more complete picture of a hotel guest. Creating Agentforce bots to improve CX makes business sense to chief commercial officer Scott Strickland.
"Just dropping Agentforce on top makes sense for us, because it just plugs in," Strickland said. "Whereas if I took any other AI -- ChatGPT or any of the others -- and tried to plug it in, I'm going to have to teach it a little bit, and there's going to be a little bit of integration."
Analyst first impressions
Agentforce isn't the "big-bang" innovation that Salesforce CEO Marc Benioff characterized it as, according to both Opus Research founder Dan Miller and Deep Analysis founder Alan Pelz-Sharpe.
"If you're looking for the ability to route questions quickly via a decision tree to existing automation at scale -- and you can find value in that -- then sure, it's useful," Pelz-Sharpe said.
"It takes [Einstein] Copilot from assistant duties toward an easier-to-measure and therefore easier-to-calculate, value role that connects customers with automation. That will help move projects from the whiteboard to checkbook for Salesforce and their partners," he said.
Miller sees Agentforce as no Big Bang but, instead, a 10-year slow burn starting with Siri and Alexa, which emerged as voice assistants. The next evolution was to domain-specific advisors, such as agent-assist tools in contact centers of the last few years. Now, we have full-blown autonomous agents from vendors such as Salesforce, Oracle, ServiceNow, Nice, Microsoft, Google and Zendesk that can complete tasks.
"It's been at least a decade in the making," Miller said. "The time is ripe now because there are short-term, high expectations for employees and supervisors to take advantage of LLMs and GenAI. Businesses of all sizes need it."
Pricing for generative AI, regardless of vendor, has been difficult to measure, said Liz Miller, Constellation Research analyst. Agentforce pricing -- starting at $2 per conversation and then volume discounts come into play -- differs from the typical subscription model Salesforce and other vendors charge for their cloud services.
This could lead to confusion in the marketplace, especially because it doesn't consider the Salesforce Data Cloud credits needed to run Agentforce and its concomitant setup costs.
"At some point we need to admit we don't really know," Miller said. "That might be OK for now, but there might be a real bottom-line impact we aren't prepared for. Let's say that thanks to agents, I can triple my capacity to engage in conversations about returns or simple purchases. Sounds great, right? But what does 3x capacity times $2 look like? Are we ready for that cost?"
Salesforce does offer 1,000 Agentforce conversations per month free to users -- along with some Data Cloud credit -- as part of Salesforce Foundations, a set of free features and services also launched at Dreamforce.
Salesforce execs chime in
At its heart, Agentforce is an app running in Slack. That's where all the agents are created, the data sources are mapped, and users launch their generative AI bots. It is one step closer to Slack becoming the main interface to Salesforce, as Benioff claimed it would be when the company paid nearly $28 billion for Slack in late 2020.
"We're bringing AI to sort of connect all the dots here, where customers can bring agents, customer data, team collaboration, any business process they have," said James Lancaster, Slack vice president of product. "We can use AI to help them work faster and smarter, all in one place."
At Salesforce Industries, Agentforce will help the vertical-specific software development team deploy more tools to more industries faster but also to more of Salesforce's clouds. In previous years, the group was more focused on Sales Cloud and Service Cloud, but the technology is such now that Industries can cover more of Salesforce's offerings.
"[Each vertical] gets all the razzle-dazzle of innovation that comes out of Salesforce. But they also get all that deep, dark industry functionality that helps them solve their most palpable pain points," said Jeff Amann, executive vice president and general manager of Salesforce Industries.
Agentforce's roadmap includes capabilities to streamline not just quick tasks but also a business's operations, said Adam Evans, executive vice president and general manager of Salesforce AI. This might include multiple agents working together for longer periods of time in deeper thought than the quick hits single agents give to instantly solve one-off customer questions.
"When serving a customer live, you don't want [agents] to say, 'Hold on. Let me think about that for five minutes versus something like, 'Help me design a strategy to integrate various systems,' which is something that you don't need to respond to right away," Evans said. "You want to get it right. You're going to see a lot more around general automation at scale. This technology allows us to go after a long tail."
Don Fluckinger is a senior news writer for TechTarget Editorial. He covers customer experience, digital experience management and end-user computing. Got a tip? Email him.