From Grind to Greatness with AI Agents

The Subtle Frustrations of Modern Workspaces

What everyday frustrations do you think could be solved with AI agents?

I was sitting at the Ymnig AI office, nestled in the heart of Södermalm, Stockholm, near Zinkensdamm. However, a recent mundane experience highlighted an often-overlooked aspect of modern workspaces. I descended the staircase to the communal printer, hoping to retrieve a single sheet for brainstorming. I was in fact working on writing this blog article. My plan was simple: jot down ideas in the traditional pen-and-paper method before diving into digital tools. Yet, this seemingly straightforward task turned unexpectedly frustrating.

The printer, usually a reliable ally in my work, was on the brink of a paperless state. With just a few sheets left, a dilemma presented itself: Should I use these last sheets, potentially disrupting someone else's workflow, or should I retreat empty-handed? Choosing the latter, a sense of annoyance crept in as I walked up the stairs empty-handed. My next thought was whether I should notify Frida, our office manager, but it seemed like a bother (find her email, write mail to her, handle response, etc) so I ended up not doing it, instead feeling a bit bad. While this annoyance and guilt was fleeting, replaced by the seamless digital alternatives I later employed, a combination of ChatGPT and Miro, it sparked a thought. What if these minor, yet cumulative, annoyances in our professional lives could be effortlessly managed? This moment of inconvenience led me to ponder the potential of AI in transforming our interaction with everyday tasks.

Thinking about this in a scenario with AI Agents

In an alternative future scenario, where AI agents seamlessly integrate into our work lives, this trivial yet bothersome series of events could have unfolded differently. Picture this: Instead of leaving the printer paper status unchanged, feeling guilt and frustration, I casually mention the situation to my AI assistant. With a simple voice command, "Please notify the unit manager, Frida, about the printer's low paper supply," the issue is out of my hands. The AI assistant, understanding the task's nature, swiftly composes and sends an email to Frida.

On the receiving end, Frida, too, is equipped with an AI assistant. Her assistant filters through her emails, identifying this particular message not as another item in an overwhelming inbox but as a specific, actionable task. It then autonomously updates her to-do list, adding the task "Refill printer paper in unit X" to the next trip here in her schedule. This interaction is smooth, requiring minimal effort from both parties, yet ensures that the small but crucial detail of refilling the printer paper is handled.

The Capabilities Required for Such AI Agents

For such a scenario to become reality, AI assistants need a suite of capabilities. Firstly, they need to be able to send, receive and reply to emails. Secondly, these assistants must be adept at interfacing with various digital tools and platforms. Frida's assistant, for example, should seamlessly integrate with task management systems, like a kanban board in Trello. This capability ensures that any actionable item is not just noted but is systematically incorporated into the user's workflow. Lastly, AI assistants should have a level of decision-making ability. They need to differentiate between tasks that require direct human intervention and those they can autonomously handle. This balance is crucial to maintain both efficiency and personal oversight.

At Flitig AI, we are at the forefront of developing such AI agents. Our focus is on creating agents that act as digital coworkers, helping out with the boring parts of the work, capable of handling a range of tasks from email management to workflow integration or basically any task with an exposed API. For users, the requirement is simple: beyond the more simple tasks that work straight off the bat, typically the work to set the agent up entails clearly articulating how the agent should think when doing the task (agent prompt engineering). Our AI agents are devised to handle tasks autonomously or seek approval when necessary, offering a blend of independence and control.

A standout feature of our agents is that they are designed so that the user can interact with them as they would interact with a human coworker: voice, chat, Slack, email, or directly in your workflow, e.g. via comments on cards on a Trello board. Users simply tell the AI what to do, when and where it makes sense to do so. This natural mode of interaction eases the user experience significantly, making it more intuitive and less time-consuming than traditional digital interfaces.

Envisioning a future with AI agents everywhere

Envisioning a future with AI agents everywhere begins with appreciating the cumulative value of resolving numerous small, everyday frustrations. These minor inconveniences, often overlooked, collectively form a significant source of unnecessary stress and inefficiency in our work lives. By deploying AI agents to manage these tasks, we not only streamline workflows but also liberate human creativity and focus for more complex and rewarding endeavors. This vision doesn't rely on grandiose technological leaps but on the thoughtful application of AI in alleviating the small frictions that pepper our daily routines.

In the workplace, the potential applications for AI agents are as diverse as the tasks that populate our to-do lists. Imagine an environment where digital agents handle the relentless influx of emails, prioritizing messages, drafting responses, and flagging critical items for human attention. Documents are not just read but analyzed, with suggestions for improvements and consistency checks conducted in real-time. Project planning and calendar synchronization become a breeze, as AI agents recommend optimal schedules based on team availability and project deadlines. Meeting notes are meticulously captured and organized, actionable items are distributed, and follow-ups are automated. Beyond these, AI agents navigate the labyrinth of administrative bureaucracy, filling forms and updating systems without human intervention, turning what was once a tedious task into a seamless process.

However, while the general domains where AI can add value are becoming increasingly clear, the vast landscape of specific use cases is just beginning to be explored. This exploration demands creativity, innovation, and a willingness to rethink how we approach our work. As we look to the future, the question that sparked this reflection remains open: