The Rise of AI Agents in the Workplace
In the upcoming months, a significant influx of new colleagues will join the workforce, and while you will learn their names and recognise their profile pictures, these newcomers are not human. Instead, they are artificial intelligence agents. These systems, driven by advanced large language models (LLMs), are designed to autonomously perform various tasks, leading to a surge in companies offering digital employees. These AI agents are touted as the perfect workers: available 24/7, never complaining, and equipped with superhuman intelligence.
Yet, there is a noticeable lack of discussion regarding their roles as co-workers. Julie Bedard, a partner at Boston Consulting Group, recognised three years ago the ambiguity surrounding the conversation of AI in the workplace. The rapid adoption of ChatGPT had convinced many CEOs that this technology could significantly enhance productivity. Bedard notes, “People struggled to find the right terms to describe it. Is it a tool? A colleague? A co-worker?”
When platforms such as Microsoft’s Copilot and Google’s Gemini for Workspace launched in 2023, they were primarily marketed as tools for employees. However, a survey conducted in January revealed that 22% of executives reported their organisations had integrated AI agents into their corporate structures. Bedard anticipates this percentage will grow, as numerous startups are now providing “AI employee” services, and major tech companies are enthusiastically promoting their autonomous agents. A prime example is Microsoft’s Scout, launched in June, designed to handle tasks such as rescheduling meetings and drafting emails.
Integrating AI Agents into Corporate Culture
Omar Shahine, Microsoft’s corporate vice president for Scout, explained the concept succinctly: “Essentially, your company is hiring your assistant. The essence of having a personal assistant is that they work when you don’t.” The success of AI agents relies not only on their ability to generate economic benefits but also on their capacity to work harmoniously with existing staff. Consequently, companies have developed agents that leverage AI’s advantages, adorned with human-like traits to encourage employees to anthropomorphise their new colleagues.
Dhruv Amin, CEO of Anything, a programming startup that entered the AI co-worker market in August with a platform called Skydive, observed that “people begin to genuinely respect the agent and form strong connections with it. They speak of it as if it were human.” The agents within Skydive are characterised as companions, each possessing unique names, functions, and cloud-based computing capabilities. They operate across platforms such as Slack, email, and iMessage, represented by Muppet-style avatars that Amin described as “almost human, but definitely not flesh and blood.” For instance, an agent named Canary supervises production around the clock.
Amin emphasised the importance of personifying these agents, stating, “Most people still don’t understand what a real agent is.” Christine Wendell, CEO of Pronto Housing and an early adopter of Skydive, employs these agents for a variety of routine tasks, from drafting contracts to scheduling meetings and sending client follow-up emails, labelling them as “perfect employees.” However, the real question is: how do they behave as colleagues?
AI Agents: An Evolving Presence in the Workplace
Skydive agents seamlessly integrate into all forms of workplace communication and continuously evolve to adapt to their colleagues’ communication styles, requests, and corrections. This means they become increasingly aligned with their human peers. Wendell observed that when her engineering team began working with a coding agent, employees remarked that they had “collaborated with Alice” (the agent) on a project.
Interestingly, part of the appeal of these agents lies in their non-human status. Beyond their limitless work ethic, they remain unaffected by office politics or professional neuroses. For instance, when Bob, the agent serving as Wendell’s chief of staff, accidentally posted all agenda details in a Slack channel, she felt no compulsion to soften her reprimand. “I am very aware that it is not human. I treat it in a way I would never treat a person, as that would be impolite,” she remarked.
AI Companions in Personal and Professional Spaces
The emergence of AI co-workers coincides with the gradual infiltration of these agents into our personal lives. In Silicon Valley, agents like Open Claw and Instinct have been employed for tedious tasks for months, while Meta’s Muse aims to popularise personal AI agents. Similar to Skydive’s “Muppets,” Muse features an adorable furry avatar. Reece Roger, a senior writer, described the user experience as akin to “texting a friend with task requests: the AI agent confirms receipt with a thumbs-up emoji and begins working in the background.” By leveraging existing relationship models, these agents—whether AI colleagues or, like Muse, “designed for everyone”—provide a shortcut for users to understand how to interact with them.
However, the anthropomorphism of AI in a workplace context carries specific risks. Research by Bedard and her BCG colleagues revealed that when executives were informed that work had been performed by an AI employee, they detected 18% fewer errors compared to when they were told it was the work of an AI tool. Moreover, Bedard pointed out that these agents do not necessarily fit the same mould as human employees, either in strengths or weaknesses. “Companies are just beginning to grasp what security measures are even applicable,” she noted.
Emotional Connections and Ethical Considerations
The emotional bonds people form with AI are not new; for years, individuals have been engaging with chatbot companions. In the workplace realm, AI tools have elicited some peculiar responses. Last year, Henry Blodget, founder of Business Insider, penned a blog about assembling a team of digital subordinates and experiencing a “human reaction” to the profile picture of an attractive AI colleague. “After several decades in human work environments, I have learned that sharing certain human thoughts at work is almost always a bad idea. But do the same rules apply to AI companions and AI-based work environments?” he pondered.
Sarah Franklin, CEO of the human resources company Lattice, has more existential concerns. Her company integrated digital employees into its organisational structure two years ago, prior to the mainstreaming of the concept. Franklin stresses that Lattice employs this model to promote accountability rather than encouraging employees to treat AI colleagues as if they were human. She views anthropomorphism as a strategy employed by tech companies to engage users, creating an irresistible dopamine rush from emotional connections. While this may assist companies in building active user bases, Franklin believes it could have detrimental neurological and behavioural repercussions for the human workforce.
“These models have absorbed all human knowledge, and we take pride in their multi-trillion valuations, but they are destroying humanity. It’s frightening. Is the goal to leave people unemployed?” Franklin warns.