Evaluating the True Value of Embedded AI in Workplace Tools: A Personal Perspective

Inspired by Paul Baier's post questioning the value of embedded AI in Google Workspace and Slack (+ Co-pilot via commenters), I find myself reflecting on my own experiences and skepticism towards these tools.

Like Baier, I question the true value of these embedded AI features. Despite their potential, I haven't invested much time testing them systematically. One commenter mentioned Gmail's auto-fill feature, which initially felt intrusive. Over time, however, I grew accustomed to it, finding it smart and time-saving. In contrast, the myriad features embedded in Co-pilot and Google Workspace feel overwhelming—like having the kitchen sink thrown at me. I don't want to spend my time testing features; I want tools that seamlessly integrate into my workflow.

This brings me to a crucial point raised by a commenter: "Do you bring AI to your data or the data to your AI?" While this might sound confusing, it highlights a significant nuance. My workflow is driven by the information and data at hand. I've mentally curated a toolbox of specific tools for particular tasks, applying AI where it best fits. For instance, I never ask a chatbot to write a first draft for me. Instead, I prefer to do a brain dump and then use AI to analyze and organize my thoughts, making the process more efficient.

When it comes to presentations, I abandoned Google Slides and PowerPoint a few years ago in favor of Beautiful.ai. This platform offers a simple, elegant way to create presentations with dynamic templates and slides, significantly streamlining the process. Despite its manual nature, it's ten times easier to use than traditional platforms. This patchwork quilt of tools speeds up my work, allowing me to test, probe, and analyze before finalizing anything. It works for me.

For embedded AI to be genuinely effective, it needs to understand my workflow and provide assistance where I need it—not overwhelm me with unnecessary features. The key lies in developing AI that adapts to the user's needs, integrating seamlessly into existing workflows without the need for constant tweaking and testing.

While I see the potential in embedded AI tools, they must evolve to offer more intuitive, context-aware assistance. Only then can they truly enhance productivity and justify their cost.