AI Integration into Workflow
AI has woven itself into the very fabric of our lives, especially through the everyday practices we often overlook. As we move forward, we are seeing more and more work being done in workflow integration, which looks not just at tools themselves but at the ways in which we interact with them on a regular basis.
Picture this: you're drowning in emails, trying to catch up with the flood of content. AI could help by skimming through everything, pulling out the most salient points, and summarizing them for you. Or imagine you’re facing a mountain of documents. Instead of sifting through it all, why not let AI craft a tidy summary to point you in the right direction?
The beauty of AI here lies not in offering separate functionality but in enhancing what is already present. Consider rewriting paragraphs in an authoring tool or taking notes in a Zoom meeting. Imagine a system tracking the commitments made during conversations and organizing them for you. We are now at a point at which development teams look around at what we're doing in business settings informally and target the places where language models could make the most difference.
The power of this approach is that, by its nature, this view of integration is aimed at getting value out of the models without relinquishing control. By looking at existing workflow, we can identify needs and pain points. I am trying to catch up on my email. I want an improved version of what I have already written. I want someone, or something, to take notes for me during meetings.
In law, for example, incorporating AI might look like this: Imagine you’re reviewing a contract. You’re still reading and considering the context yourself, but a system is beside you, pointing out patterns and pitfalls that others have encountered previously. The workflow doesn’t drastically change, but your abilities are amplified so that you are more effective.
It’s about identifying needs by looking at our work and building systems that address those needs rather than perform the entire task.
Of course, we need to be careful. Summaries are fantastic—unless you’re left with the impression you shouldn’t read your documents. If, for example, you rely solely on an AI-generated summary without a human-centered veracity check, you could lose the safety net of catching and correcting mistakes.
The goal isn’t mere timesaving; it’s about using time wisely, sharpening our edge in tasks that demands our judgment or critical thought. Improving performance is the focus, not reducing effort. It’s the art of doing better, and AI, when integrated thoughtfully, is the brush with which we can paint this new picture of productivity.
Kristian Hammond
Bill and Cathy Osborn Professor of Computer Science
Director of the Center for Advancing Safety of Machine Intelligence (CASMI)
Director of the Master of Science in Artificial Intelligence (MSAI) Program