Salesforce CEO Marc Benioff is known for his enthusiasm for generative AI. But even he questions the most optimistic claims about the technology's capabilities. And, predictably, he blames Microsoft.
A month after Salesforce's high-profile tech conference Dreamforce, where Benioff repeatedly touted and praised AI, especially in relation to Salesforce products, he is currently on a roadshow to set expectations. And he is negatively criticizing its main competitor and strongest rival, Microsoft. Recently, he has positioned himself on the podcast “Fast answer“, which is hosted by former Fast Company editor-in-chief Bob Safian.
Speaking about the potential of AI, he said: “I have never been more excited about another topic at Salesforce, perhaps in my entire career.”
The individual also noted that “incorrect information has been communicated to customers about enterprise AI, and possibly AI in general.” I believe that Microsoft has not adequately served our industry and the AI research being conducted.
Benioff has been critical of Microsoft Copilot's accuracy and usability. He even compared Copilot to Clippy, Microsoft's widely criticized talking paperclip cartoon from the 1990s that was designed as an assistant for Microsoft Office users.
We may have heard of AI priests and priestesses at prominent LLM and Microsoft companies claiming that AI is helping cure cancer and tackle climate change. There is also talk of the need to connect to nuclear power plants to access specialized data centers. Benioff denied the veracity of these claims.
He was reaffirming his stance regarding a comment conducted at X, in which he stated that “Large Language Models (LLM) are not a direct path to Artificial General Intelligence (AGI), and that much of the current potential of AI is being exaggerated.”
A dig was made at OpenAI CEO Sam Altman. Earlier this summer, Altman raised the possibility that AI-enhanced healthcare technology could help discover cancer cures in the future. Speaking at the Aspen Ideas Festival, he said: “[…] Maybe a future version will help discover cancer cures.” As reported by the magazine Newsweek.
Benioff also referenced research conducted by Gartner on Microsoft Copilot. According to a Gartner Report In a report published in April titled “Copilot’s Top Challenges for Microsoft 365,” it was revealed that only 25% of organizations running Copilot pilots are currently planning a full-scale deployment. This is actually quite a positive number considering the recent introduction of the technology and the slow pace of adoption by companies. In this regard, Gartner concluded that the increase in Copilot’s quality will be reflected in an increase in its adoption by companies.
According to one market research In a survey conducted by Forrester, Microsoft could point to a number of key benefits for small businesses that employ Copilot. Based on a survey of 266 small businesses, Forrester found that using Copilot resulted in a slight increase in revenue, while decreasing operating costs and streamlining the process of onboarding new employees. However, it is important to note that funding for this research came from Microsoft. The analysis of this scenario is left to the free interpretation of each person.
But Benioff is right that the AI we have today, while sometimes impressive, is not ready to replace human workers in most situations. It’s true that the podcasting capabilities of Google’s NotebookLM, which can generate entertaining AI hosts to explain content, are highly effective. But it’s hard to see how this move will directly help reduce the menial labor that affects emotional well-being and is common in most corporate jobs.
Benioff is also right to point out that GenAI excels at working with businesses, especially in the area of AI agents. That statement ties in nicely with Salesforce’s Agentforce product, which he recently promoted. A number of established and emerging technology companies are developing AI agent technologies. These efforts range from building applications for particular use cases to providing platforms that allow companies to design their own solutions. Notable examples include OutRival, Atlassian’s Rovo AI, and Sierra, a startup co-founded by former Salesforce co-CEO Bret Taylor and former Google executive Clay Bavor.
AI agents are being widely used in the customer service space, filling roles such as chatbots on websites and assistants in field services. Examples include Zingtree, Talla, and Neuron7. Salesforce offerings are also available here.
AI agents are proving effective in the sales arena, particularly in activities such as prospecting, cold calling, and sending emails with high bounce rates, which are considered entry-level activities in a sales structure. Examples include AI Regie.ai, AiSDR, Artisan, and 11x.ai.
Benioff speculated that over the next 12 months there will be more than a billion agents operating from Salesforce, based on testing by about 10.000 customers at his technology conference.
However, it’s also worth noting that there are other areas where LLMs are already valued and where Microsoft, specifically, has play. Programmers and software engineers are increasingly using them to help them test, debug, or generate code examples, including with Microsoft’s GitHub Copilot. Countless startups also offer AI coding assistants, such as JetBrains and Dragon, to name just a couple.
The close collaboration between Microsoft and OpenAI has made Azure, Microsoft’s cloud platform, a widely preferred choice for companies that employ deep learning language models (LLM) to develop their own generative artificial intelligence (GenAI) applications. As a result, Microsoft’s influence over GenAI manifests itself in a wide range of functions that go beyond writing documents in Word, creating charts in Excel, or transcribing meetings in Teams.
It is safe to say that Artificial Intelligence is overhyped when those promoting AI-related products acknowledge that its importance is exaggerated.
Benioff stressed the importance of managing expectations while taking advantage of the current capabilities of artificial intelligence.
