Build a Quick Win
Most organizations approach AI like a construction project: foundation first, everything else after. In theory, that’s right.
In practice, the business runs out of patience before the foundation is finished.
Colin McQuay, Senior Director and Head of SRM at Nestle, describes the tension clearly. The data foundation matters. Getting pricing code to talk to promo math, mapping retailer hierarchies to internal logic, building the infrastructure that makes agentic workflows actually possible. All of it is real work that has to happen. But nobody has two years to do it before they have to show something. The business won’t wait. The budget won’t wait. And if you ask for two years before delivering value, you probably won’t get the two years.
A Director of Supply Chain Strategy at a large CPG organization puts the pressure in plain terms: the business doesn’t care about your data architecture. They want the answer. Give it to them in Excel if you have to, but give it to them. That pressure is real and it isn’t going away. The question isn’t whether to respond to it. It’s how to respond without sacrificing the foundation work that makes everything else scale.
The answer is sequencing. A senior global technology executive at a major CPG company describes this as finding the recurring workflow where the data is already clean enough to act on, automating that first, and using the credibility it earns to fund the harder infrastructure work underneath. Not a grand transformation. A foothold. One workflow that runs reliably, delivers visible ROI, and gives leadership a reason to keep investing while the foundation gets built.
McQuay’s team did exactly this. A messy back-and-forth ordering and pricing workflow between their organization and a major retail partner was consuming time and causing out-of-stocks. Two teams built agents on each side of the process. The manual exchange disappeared. The ROI was immediate and measurable. That win didn’t require a perfect data foundation. It required finding the one workflow where the data was clean enough and the problem was painful enough to justify moving now.
The foundation still matters. Build it. But don’t let the pursuit of perfect infrastructure become the reason you never deliver anything.