Insights
The value isn't in the model. It's in the redesign.
We've been building AI into actual workflows for the past year. Not proofs of concept. Not demos for leadership. Actual tools that people use on a Tuesday to get their job done. And here's the thing nobody at the conferences is saying: most of the value we captured had nothing to do with the AI.
The AI was the excuse to finally look at how work actually flows. And what we found was that the processes were broken long before anyone tried to automate them.
Chewing through documents. This is the unsexy win and it's the biggest one. We had a process where someone spent two days a month pulling data from compliance documents, cross-referencing it with operational records, and building a summary for leadership. We replaced the extraction and cross-referencing with AI. Took it from two days to forty minutes. The person who used to do it now spends that time actually analyzing the data instead of assembling it. That's real.
Catching misalignments earlier. Hillary's side of the house. We built a tool that flags when the demand signal and the supply plan start diverging — not at the monthly review, but in near-real-time. The AI isn't making the call. It's surfacing the question a week earlier than anyone would have noticed it in a spreadsheet. That extra week is worth a lot when you're making production commitments.
Learning that adapts. Dave's side. Instead of one static leadership curriculum for everybody, we're routing people to different content based on what they're actually working on. Not a Netflix recommendation engine — that metaphor needs to die — but a system that knows a plant supervisor in their first 90 days needs different support than a senior leader preparing for a cross-functional initiative. The AI handles the routing. The humans built the content.
Anything that required understanding why people do what they do. We tried to use AI to predict adoption patterns for a new process rollout. It was confidently wrong in ways that would have been embarrassing if we'd acted on it. Turns out the reasons people resist change are political and emotional, and no model trained on historical data is going to capture the fact that the VP sponsoring this initiative burned trust with the plant team two years ago.
Also: chatbots for internal support. We built one. It answered questions correctly about 70% of the time. That sounds decent until you realize the 30% it got wrong eroded trust so fast that people stopped using it within a month. Accuracy has to be near-perfect for internal tools or people just go back to emailing Karen in HR who always knows the answer.
If you're thinking about AI for your operations or your people development programs, start with the process, not the model. Map how work actually flows. Find the parts that are manual, repetitive, and time-consuming. Ask whether the process itself makes sense or whether you're about to automate something that shouldn't exist in the first place.
The AI is a tool. The redesign is the strategy. Most organizations are buying tools and skipping the strategy, and then wondering why the ROI isn't there.
← Back to Insights