Stop Thinking About AI for Your Workflow. Start Thinking About Your Tasks.

I’ve been writing about technology for nearly thirty years (pardon me while I go turn into a pile of dust)—and using it for more than forty years (see previous parenthetical). One of the truisms that comes with the benefit of seeing back through time is that I’m rarely surprised by the discourse around (and I’m sorry to use this term) disruptive technologies.

The latest technology to stumble into the public square is Generative AI.

The general discourse around these technologies as they roll out is rarely nuanced. (Y’all may not remember Desktop Publishing, but that was absolutely going to ruin things, and how dare these ‘zines exist!) A cabal of true believers will tell us that everything will change, while a murder of defenders of the past speak only of its evils.

And so it goes—this time—with Gen AI.

The discourse will change in the coming years, as the technology seeps deeper into our way of life. (The one thing I do hope we’ve learned—although I have my doubts—is to proceed with a sense of caution when it comes to the impact on culture and life.)

Which brings us to here, an inflection point in the discussion about Artificial Intelligence and its use within the workforce.

Researchers from MIT’s Sloan School of Management recently published a 120-page paper outlining a framework for evaluating workflows—and the tasks that make them up—to help teams determine where (or if) AI might be deployed and how that might (or might not) help the team. It’s a deep dive into the nuance of work, one that project managers who have done Work Breakdown Structures have lived with for a while.

For the TL;DR crowd, I created a summary page for each section—and included a bit about its importance in my Comms/Project Management world. (See how Claude Projects and GenAI made your life a little easier there.)

The general idea that resonated for me (and please read to get the actual idea) is that you shouldn’t—or can’t—think about using AI for your workflow. Instead, you should evaluate each individual task within a workflow to determine whether or when it can be improved with an existing tool or one that will soon exist. This gives teams the ability to be thoughtful about how to improve their workflows.

While I was an academic for almost twelve years, I’m a practitioner at heart. (Go, Journalism and Storytelling.) I’ve always tried to operationalize research and turn it into a practical tool that helps me solve problems. That’s what research is (sometimes) really good at doing.

So, I turned to Claude Projects again and set out to vibe code an interactive template that uses that framework and turns it into a usable tool for building a workflow and evaluating the potential use of AI for any task within it.

If you’re working towards understanding when, why, and why you can (or shouldn’t) use AI in your workflows, this framework is pretty great.