AI Agents — The Age of The AI Employee

Our relationship with AI is about to change. Until recently, we have communicated with Language Models, like GPT and LLAMA, as informational sources, teachers, and even therapists. However insightful the back and forth conversation with an AI Chatbot might be, the workload to apply the AI generated insight still falls on the human. AI Agents are about to flip that on its head.

To bring more tangible understanding to the article we will follow a team of AI Agents whose goal it is to write Articles for a fictitious Gossip Magazine called “In The Gen”.

What really is an AI Agent

AI Agents are not your ordinary chatbots; they are language models assigned to single or multi-agent AI roles — be it a Manager, PR Specialist, Writer, and more. Imagine creating an AI agent for almost any job role. Like their human counterparts, these agents can collaborate in teams to achieve objectives and complete goals, moving beyond mere dialogue to working together towards a shared aim.

For example, if the fictitious Magazine “In The Gen” wanted to create a Team of AI Agents with the objective of writing featured Quiz’s to include in their magazine they could put together a team of Agents comprising of a “Market Research Agent”, “Director”, “Manager”, and “Quiz Writer”. Each AI agent would have a different role on the team, just as a human team would.

Defining the Role of the Agent

The key to seamlessly integrating an AI Agent into a team is defining its role or roles. This ensures the agent operates within its designated scope, adhering to policies and ethical standards. Roles are shaped through “System Instructions” tailored to the agent. Depending on the Agent framework you choose to work with, System Instructions can be provided to the Agent each time a message is passed to the agent (per message instructions), once at the start of the session (persistent instruction), or a combination of the two (hybrid approach).

The components of an effective system instructions include (but are not limited to):

• Specifying the tasks and responsibilities of the agent
• Detailing the decision-making authority of the agent
• Specifying any rules or policy that the agent must abide by
• Language or tone preferences

For our Quiz-writing team at “In The Gen”, the Manager Agent’s instructions might read:

“ You are the manager of a team writing Quiz’s for a Magazine Company called “In The Gen”. You report to the Director who either assigns you work or reviews your work. You manage the Writer whose job it is to write Quiz’s. Your job is to critique these Quiz’s and ensure they align with company policy and retain a playful tone. If the section does not pass your review send it back to the Writer for revision with constructive feedback. Before writing your response consider if you need to address the Director or the Writer. If you are assigning a task or giving feedback then talk to the Writer. If you are sharing the Writer’s draft then talk to the Director.”

Note: System Instructions should be tweaked during experimentation depending on result

A Glimpse into an Agent Workflow

Consider this workflow: A user requests a quiz, targeting a specific audience, which is passed to the Director. The Director consults the Market Research Agent to pick a topic. Once selected, this topic is shared with both the Director and Manager, who then defines the quiz’s requirements for the Quiz Writer. Following reviews and revisions to ensure quality, the final quiz makes its way back to the user. This process exemplifies the collaborative and iterative nature of content creation with AI Agents.

Enhancing Human-AI Collaboration

As we delve deeper into the era of artificial intelligence, the dynamics of the workplace are undergoing a profound transformation. The advent of AI Agents marks a pivotal shift from viewing AI as mere tools to recognizing them as active participants in the workforce.

While we grow accustomed to trusting our AI counterparts, we retain control over the level of human oversight and interaction. By configuring how often the agents are required to touch base with the user for input or feedback, we can significantly influence AI-generated outputs. I firmly believe that until we hold our AI Agents to the same standards as our human employees, integrating human collaboration into agent frameworks will be critical.

Before anxiety sets in about the significant changes ahead, it’s important to recognize that as AI Agents assume more responsibilities within teams, the roles of human workers evolve positively. Humans are liberated from repetitive and mundane tasks, allowing them to focus on strategic thinking, creative endeavors, and complex problem-solving. This shift does not diminish the value of human labor; rather, it enriches it by adding layers of creativity and strategic oversight that AI has yet to replicate.

In the context of our example with “In The Gen,” AI Agents like the Director can work alongside a human user who approves the research and quiz topics that the Market Research Agent suggests. Approval from the user is required before the Manager Agent creates the quiz requirements for the Writer Agent to develop the draft. This scenario serves as a microcosm of the broader potential of human-AI collaboration. This partnership, when navigated thoughtfully, heralds a new age of work where the combined strengths of humans and AI elevate our capabilities, creativity, and collective achievements to new heights.

Thanks for reading! Interested in discussing further or collaborating on an article? Follow me on LinkedIn!

As for more insights on the future of AI, be sure to check out Neudesic’s 2024 Technology Vision.