Salesforce recently launched Agentforce, a no code platform that helps enterprises create personalised AI agents with simple prompts.
At his keynote address at Dreamforce, Marc Benioff made a clear point-AI agents are key to scaling AI adoption in enterprises. During the event, which is held in San Francisco, he took a jab at companies like Microsoft, noting that enterprises have found it challenging to realise the value of copilots- likening AI chatbots to Microsoft Office's rule-based agent from the 2000s-Clippy.
Benioff's criticism of copilots-a term popularised by Microsoft- might seem unexpected, given that the company, not so long ago, introduced its own version of copilot-called Einstein Copilot.
"The copilot world has been kinda hit and miss world. Customers have told me we got these copilots, but they are not performing as we want them to," he said in his keynote address at Dreamforce 2024.
Now, rebranded to Agentforce-of course-the platform is designed to help enterprises streamline business processes, deliver personalised insights, and boost productivity and all of that.
According to Benioff, copilots were able to assist with basic tasks, but their inefficiencies are what gave birth to AI Agents. Dubbed as the next iteration of deep learning models, AI agents are designed to go further and take actions on behalf of users.
( Salesforce CEO Marc Benioff during his keynote address at Dreamforce)
Interestingly, this isn't the first time Benioff has criticised Microsoft. During a recent earnings call, he highlighted how enterprises have squandered considerable resources on third-party models, often being sold as APIs or in the cloud, attempting to fine-tune them for their specific use cases.
Salesforce's approach is to democratise AI adoption, and Benoiff's message to customers at Dreamforce has been clear- buy, not build.
To further encourage this, Salesforce is mulling a consumption-based pricing model for its AI agents- somewhere around $2 per conversation.
Benioff revealed that companies such as Saks, Wiley, and Wyndham have already started leveraging the Agentforce platform to build customised AI agents for their specific needs.
A few Indian heavyweights, such as Mahindra and Mahindra, Tata Consumer, and TVS, which are all Salesforce customers, were present at the event to gain firsthand experience and 'know-how' of these AI agents.
While speaking to AIM, all of them revealed that they are excited by the prospect of Agentforce. "Attending Dreamforce provides insight into what's next for that platform. While Agentforce may sound innovative, it's actually about leveraging Salesforce's existing capabilities that are now being productised," Saurabh Khullar, chief customer experience and digital officer TVS, told AIM at Dreamforce.
"The rationale for buying rather than building is simple: if you build it, you'll need to continually enhance it and create the necessary ecosystem. In contrast, when you purchase a solution, you can rely on the vendor to keep developing the platform," he added.
Powering Agentforce is Atlas-an advanced AI reasoning engine designed to simulate human-like thinking and planning processes. Atlas can autonomously analyse data, make decisions, and complete complex tasks across various business functions.
Atlas enables businesses to deploy customisable agents tailored to their specific needs. By deeply integrating with Salesforce's Data Cloud and other integrated systems to process vast amounts of information in real-time, Atlas ensures that the AI agents in Agentforce make decisions based on the most current and relevant data.
While speaking with journalists at Dreamforce, Clara Shih, CEO of Salesforce AI, revealed that while the first iterations of these AI agents will automate simple routine tasks, going forward, they will be able to handle complex tasks such as handling payments.
"I believe that agents will have the ability to handle payments on behalf of their consumer and yes, of course, it means that the vendor, the supplier, will have to have a bunch of safeguards. Just like the internet created an entirely new category of cybersecurity, which is, by the way, a huge, multibillion-dollar, very fast-growing industry, AI agents will also require their own new cybersecurity stack," Shih said, answering a question posed by AIM.
While Atlas might be the reasoning engine, Data Cloud provides the underlying foundation of Agentforce, something the CRM leader announced at the previous Dreamforce.
Data Cloud is a customer data platform (CDP) that consolidates data from various sources into a single location. According to Salesforce, it is one of the fastest-growing innovations they've introduced, if not the fastest, reflecting significant organic growth.
For AI agents to work at their best, they need access to all enterprise data points, and Data Cloud enables that. What data cloud does best is bring all unstructured data together and make it accessible in one place. However, not all Salesforce customers are Data Cloud customers.
"The ability to search and process unstructured data is crucial for agents. Business conversations often begin with questions about policies, such as handling a delayed shipment. Access to relevant knowledge articles -- typically unstructured information -- is essential; without it, agents cannot perform effectively in resolving customer issues," Param Kahlon, EVP & GM, automation and integration at Salesforce, told AIM at Dreamforce.
(The new Agentforce Mascot at Dreamforce)
Interestingly, Salesforce's hard pivot to AI agents comes at a time when every major Software-as-a-service (SaaS) and AI company is betting high on them.
Microsoft, too, recently introduced Copilot agents in its Microsoft 365 Copilot platform, which is designed to help businesses customise AI for their specific needs.
However, the automation being promised is akin to what Robotic Process Automation (RPA) promises -- a type of business process automation that relies on software robots.
Explaining the difference between RPA And autonomous AI agents, Kahlon told AIM, "RPA agents were designed to automate repetitive, tedious tasks, such as transferring data between systems when APIs aren't involved. In contrast, autonomous agents process information more like humans, adapting to situations and making decisions based on changing conditions, enhancing efficiency and effectiveness in workflows."
He further adds that autonomous agents also does not mean the end of RPA technology. While RPA's will exist and continue to automate tedious tasks, AI agents will start to automate much more complex tasks.
"I believe RPA will coexist with today's agents. By utilising RPA bots, autonomous agents can effectively update records or orders, making them more versatile and capable of interacting with various systems seamlessly," Kahlon added.
