Last week we hosted the latest in our global series of Executive Lunch and Learns at the Danish Consul Residence in New York City. A small group of senior CPG leaders sat down for what turned out to be an extremely open and honest conversation on where companies struggle to implement AI.
Every insight pointed to the same underlying truth: specificity beats ambition. The organizations making real progress are not the ones with the boldest vision. They are the ones who found one problem, defined a clear outcome, and used that win to earn the next conversation.
Here is what that looks like in practice.
The biggest bottleneck is not the technology
We went in expecting the conversation to centre on AI tools, platforms, and infrastructure. It did not. Almost every discussion came back to the same thing: change management.
The technology exists. The data, in most cases, exists. What does not exist in most large CPG organizations is a clear owner, a specific use case, and the organisational permission to start small and learn.
One framing stuck with us more than any other. You can have the best AI strategy in the world and it will die at the bench if the people running the workflows do not believe in it, understand it, or feel it makes their lives easier rather than harder. The people closest to the work will nod in the kick-off meeting but change nothing where it counts.
The lesson is not that change management is hard, the lesson is that most organizations are still treating it as a communications problem when it is actually a structural one. You cannot memo your way to adoption. You have to redesign the workflow so the new way is obviously better than the old way, for the person doing the work, not just the person approving the budget.
Burning money has burned trust
Several people in the room referenced the same dynamic. Large AI budgets were approved, ambitious pilots were launched, and the results did not materialize quickly enough to justify the investment. Leadership became cautious. The next round of funding got harder. The initiative stalled.
This is one of the most under- appreciated dynamics in enterprise AI adoption right now. Failed bets without clear context do not just waste money. They make the next attempt harder because trust has been spent.
The organizations making real progress are not the ones with the biggest budgets. They are the ones who defined the context tightly, picked a specific problem with a clear commercial outcome, and proved value before asking for more.
The money is in the boring stuff
There was a moment in the room where someone said something that landed harder than anything else that day. The money is going to come from the boring stuff.
It will not be in the flashy AI launch, or the headline-grabbing product, but the five week process that becomes five minutes. The ingredient screening that used to require a team of people and now happens before the first experiment is run. The food safety audit trail that used to live in a spreadsheet and now feeds a live dashboard to track warnings.
These are not exciting things to put in a press release. But they compound. Every project. Every team. Every quarter. And by the time the competition can see the lead, it is already locked in.
Incremental Wins Over Total Transformation
One of the most experienced R&D leaders in the room came in describing himself as an all or nothing person. He left saying that incremental is the smarter path.
This is a shift we are seeing across the industry. The all or nothing mentality, as in the belief that AI transformation has to be total to be worth doing, is one of the most common reasons large initiatives stall. You cannot prove value at scale without first proving it at the bench. You cannot get enterprise funding without showing commercial outcomes. And you cannot show commercial outcomes without starting somewhere specific.
Pick a few places first. Do not try to save a billion dollars in one go. If you try to do it all, you drown.
The organizations pulling ahead are not the ones with the most ambitious roadmaps. They are the ones who found one real problem, anchored it in science, defined a clear outcome, and used that win to earn the next conversation.
You need a real use case, not a vision
The single most important insight from the day, and the one we are taking directly into how we talk about AI going forward, was this:
The story is stronger than just saying we need AI. You need a real use case anchored in science with a clear commercial outcome. You need to be specific.
A general mandate for AI transformation does not move budget committees, does not convince technical gatekeepers, and does not build the internal trust needed to sustain momentum. A specific problem with a specific outcome does all three.
For one organization in the room, the problem was ingredient screening taking five weeks when it should take five days. For another it was the inability to draw correlations across disconnected data lakes. For another it was the cost of adding more people to solve food safety problems that technology should be solving.
These are not abstract AI challenges. They are operational problems with commercial consequences and clear solution paths. That is where the conversation needs to start.
What this means for how we work
We need to keep creating these rooms. There was a moment near the end of the lunch where someone said something quietly that validated just how important these events are: "We underestimate just getting together, talking intellectually, and hearing what others are doing."
That is what this series is for. And we are just getting started.
Reshape Biotech is the operating system for microbiology R&D and QC. We help the world's leading CPG and ingredients companies build the data foundation that makes AI possible and then put that AI to work. If you are navigating any of the challenges described above, we would love to continue the conversation.
Book a conversation with our team
