How to Identify & Prioritize Use Cases for AI
Figuring out where to integrate AI into your organization is a challenging exercise, as the viability of each use case requires granular understanding about the business process, people, and systems involved. However, it's also essential to consider where AI can assist in streamlining and enhancing how you work.
Instead of attempting a top-down mandate, using a bottom-up approach will yield more meaningful and tangible use cases. Let's explore how to do this.
Create a cross-functional team that at its core includes people from operations and technology. Each of these folks should have a working knowledge of generative AI, even if their use is limited to their personal lives (i.e. non-work tasks).
Engage a lead stakeholder from each department to act as a liaison between your team and those that execute business processes and use systems within that department. The lead stakeholder will help you:
Identify a list of processes and systems their team is executing
Contextualize sensitivities their team may have related to AI (and digital transformation initiatives as a whole)
Explain to their team why the listening tour is happening
Observe how a team executes a business process from start to finish using real data if possible. The best way to observe how each stakeholder performs the work is in-person where they would typically be, as it is a familiar environment; moving a person into a conference room with a different technology setup or attempting to narrate their work over Zoom disrupts their flow, which might leave out key details.
Capture a list of ideas where AI could play a role in each business process, which can be as simple as asking "could we do that with AI?". Each idea represent a use case and should have traceability to a specific part of the business process along with tangible examples, as those will assist in vetting feasibility.
Note that you get bonus points (ask your manager for these, not me) if you can also:
Attribute value to each use case, which can be as simple as noting how much time is spent and how often that use case occurs
Find accidental white space in the process, which can be as simple as describing the process to AI and asking it to find holes (like the Swiss cheese metaphor)
Groom and prioritize the list of use cases by identifying overlaps, which could be the same task performed by multiple groups or different tasks that rely on the same underlying data or knowledge. To prioritize these use cases, consider these factors:
Determine the feasibility (that is likely driven by technical and legal constraints)
Assess whether rolling back to the current state is possible if AI does not work as expected
Estimate the value realized if AI can improve the process
Another opportunity for bonus points (again, ask your manager, not me) is to create a prototype that shows how AI could work to satisfy a use case. You don't need to be fancy, just indicate the assumptions made and use the tangible examples that you captured.
If a use case cannot be satisfied by AI today, indicate why and keep it on your list. AI offerings and capabilities are changing a weekly basis, so something that is impossible or not feasible today might be possible next week.
TLDR: Create a team with ops and tech, go on a "listening tour" of each function within your organization, prioritize based on impact and rollback.