Recently, Salesforce CEO Marc Benioff (Marc Benioff) bluntly criticized Microsoft's Copilot AI at the company's annual Dreamforce conference, comparing it to Clippy, a digital assistant that failed in the last century.

“We all know now that Microsoft Copilot is basically the new Microsoft Clippy, and customers aren't getting value from it.” Benioff said in an interview with Bloomberg.

Microsoft launched Clippy in the 1990s. This is a cartoon paperclip assistant that provides suggestions to users. It was criticized for its strong distraction and poor practicality, and was eventually removed by Microsoft in the early 2000s.

Benioff pointed out that many customers complained that Copilot wasn't as good as OpenAI's ChatGPT, even though the two used the same technical foundation.

Tech giants are betting on Agent

In stark contrast to this, Benioff is full of confidence in Salesforce's latest AI agent product, AgentForce.

AgentForce is a type of AI agent (Agent) that aims to provide companies and their employees with direct productivity improvements without the need for companies to develop and adjust AI models themselves.

“We've attracted tens of thousands of customers and told them they don't need to develop AI solutions themselves,” Benioff said in an interview.

In fact, many global tech giants such as Microsoft and Workday are putting agent systems at the core of their development strategies, with the goal that they can not only help, but also complete tasks on behalf of users.

Microsoft CEO Satya Nadella (Satya Nadella) said that the ultimate form of an agent is an “enterprise orchestration layer,” where employees can generate and use these agents by talking to the system to complete various complex tasks.

Meanwhile, companies such as Salesforce are targeting areas such as automated customer service, aiming to use AI agents to handle more complex tasks or even replace some jobs.

Can AI agents actually be implemented?

Although many people are optimistic about the future of agents, this technology is still facing many challenges.

One of the most prominent issues is that generative AI systems are prone to “illusions”, that is, generating inaccurate information in uncertain situations, which leaves users with reservations about completely relying on agents to perform tasks.

Barry Briggs (Barry Briggs), the former chief technology officer of Microsoft, said he was “cautious and nervous” about the widespread use of AI agents, and believes that such systems must operate in a workflow where employees have the final say.

Furthermore, in the past, software companies charged customers based on the number of employees using the software, but as agents replace part of the workforce, this charging model will also be under pressure.