The teams that are getting massive value out of gen AI today aren't the ones who bought the shiniest tools, the highest costing ones. They're the ones who took maybe two, three months, six months to clean up their data model. To clean up their permission sets, clean up the business logic.

Take a back seat and have a look. If a machine looks at this data, what will it do? What will it understand and start replicating at a higher speed or scale? Do I have some of those gaps? Do I have some contradicting data sets or processes that will amplify? And that's when I'll notice.