As more white-collar workers enter positions involving AI, a new way of showcasing one's professional skills has begun to gain popularity: many professionals no longer just emphasize their proficiency in using AI, but openly display how many AI agents they manage, treating their "digital teams" as proof of their capabilities and efficiency.
Individuals start to display the number of proxies they have.
Foreign media reports that some respondents assign names, responsibilities, and divisions of labor to AI agents. The person in charge of AI consulting company, Mason Scurry, stated that his company has over 30 AI "employees" responsible for task tracking, request processing, and model cost auditing, among other tasks. The Meta communications manager, Maayan Sarig, also mentioned that he manages 6 agents, of which only one is a real human colleague.
Senior marketing expert Derrick Hicks stated that he set up 16 professional agencies and additionally created a system to coordinate these agencies. For some consultants and freelancers, the number of agencies they represent is also used to demonstrate their delivery capabilities and to attract potential clients.
Enterprises are adopting it at an accelerating pace.
This trend is not limited to social media platforms only. Reports mention that Mark Zuckerberg, the CEO of Meta, is developing AI agents to assist him in his work, and he is also advancing the AI version of personal avatars to interact with employees. At the corporate level, "agents" are increasingly being regarded as an indicator of the progress of AI.
- AlphaSense: 2,175 financial report conference calls last quarter mentioned agents
- Salesforce: Over 25,000 companies use Agentforce
- Wal-Mart: Has established four 'Super Agent' frameworks
However, the article points out that what people refer to as “AI agents” is not the same thing in different contexts. Some of these are merely automated tasks that run at set times, responsible for capturing information, invoking models, or transferring data between different services, and they are not the same as systems that make continuous decisions.
Maintenance costs have also increased.
Automated Consulting Group stated that they use approximately 25 proxies, but many of them only perform a single task, such as scanning LinkedIn, organizing meeting minutes, recording customer emails, or summarizing subscription content into weekly reports. He believes that this number sounds large, but it does not mean that there are 25 “digital employees” working continuously.
Allen also mentioned that the setup speed of proxies is very fast, so it's easy to increase the number, but subsequent maintenance is not easy. He even set up a dedicated proxy to check if other proxies are running normally and to issue alerts in case of failures. According to him, there are almost daily issues with some aspect of the process.
A report from Work AI Institute, which is under Glean, indicates that office employees spend over 6 hours per week “taking care of robots,” including providing additional context, checking outputs, and correcting errors. This means that while AI agents save time, it may also introduce new management tasks.
The measurement standard remains actual output.
Many respondents believe that what is truly important is not the number of agents, but whether they can save time, improve quality, and solve actual business problems. Sarig indicates that if an agent's performance cannot consistently be faster or better than that of a manual process, it will be disabled. Hicks also states that the number of agents only makes sense when they can alleviate business bottlenecks.
