Anyone can demo.
Not everyone can deploy.
A new generation of AI companies can demonstrate what's possible in physical industries. Far fewer can become part of how a refinery runs, how a factory operates, or how a power grid stays balanced.
Crossing that gap is rare, and it's not because of raw capability. The companies that make it to deployment understand what enterprise software is for.
The Next Enterprise Software Cycle
Every generation of enterprise software has changed where organizational knowledge lives. The first systems digitized records. ERP standardized business processes. Cloud software connected workflows across organizations. Each generation let companies stop relying on individuals to coordinate, remember, or manage something, and moved that knowledge into software that outlived the people who held it.
SAP wasn't a better filing cabinet. Salesforce wasn't a better Rolodex. They became enduring companies because they captured knowledge that had lived only in paper, inboxes, and individual employees, then made it available across the organization. We believe the next cycle follows the same pattern capturing the one form of organizational knowledge enterprise software never reached.
The Largest Body of Knowledge Software Never Captured
The physical economy digitized its paperwork decades ago. Manufacturers run ERP. Utilities run asset management systems. Plants run on SCADA, historians, and maintenance platforms. Logistics companies digitized dispatch. Construction firms manage projects through software. Industrial companies have enormous sets of data. They've spent decades recording operations, logging every alarm, work order, inspection, maintenance event, and sensor reading.
But operations have never run on records alone. They run on operational expertise: the operator who recognizes a failure before the threshold is crossed, the technician who knows why two identical alarms require different responses, the supervisor who changes a production run because today's material behaves differently than yesterday's. That expertise rarely entered enterprise systems because it wasn't the kind of knowledge they were built to capture. It wasn't a transaction or a workflow. It was judgment, developed over years of operating real systems under real conditions.
For decades, enterprise software captured what happened. Experienced people understood why it happened.
The Next Layer
That boundary is beginning to move. Modern AI can increasingly learn from operational history and improve through real-world deployment. Combined with advances in sensing, connectivity, and computing, software can begin learning from both halves of an operation: the operational record and the expertise applied to it.
At the same time, the workforce that accumulated decades of practical experience is retiring faster than it can be replaced, taking with it knowledge that was rarely documented because it never needed to be.
AI enables fundamentally different kind of enterprise software, not software that records operations, but software that participates in them.
The Hard Part Isn't Intelligence
It's tempting to think this is a modeling problem. It isn't. The hard part begins after the model works.
When software manages records, mistakes create administrative work. When software participates in operations, mistakes shut down production, damage equipment, interrupt critical infrastructure, or put people at risk. That single difference changes everything.
That's why so many technically impressive industrial products never move beyond a successful pilot. Customers aren't deciding whether the software is intelligent. They're deciding whether it deserves a permanent role in work that carries real operational consequences. They grant that role the way industrial organizations always have: gradually, through operational evidence and repeated performance, with a hand kept near the switch until the system has earned its way to the center of the work.
The companies that define this generation will win because they design for that process from day one.
The goal was never a successful pilot. The goal is becoming part of how the operation runs. That's why large funding rounds alone doesn't win here.
Trust is the one asset a larger check can’t buy.
Why We Invest at the Beginning
If this cycle is defined by earning the right to participate in critical operations, the investment opportunity changes with it.
The defining companies won't be obvious from a demo or a benchmark. They emerge through years of customer learning and operational validation, inside environments where failure carries real consequences. We write the first check at exactly that moment, while a company's approach to deployment, customer learning, and product design is still being shaped.
Our conviction comes from seeing these businesses through the eyes of the customer responsible for the operation, not only the technology. We evaluate what we back the way the buyer does, and we fund the problems our operator network is already asking to have solved rather than betting on trends from the outside in.
In markets where trust determines the winner, that perspective is our edge.
What We Believe
Every enduring enterprise software company has expanded the role software plays inside an organization. We believe the next generation will expand software into operational expertise itself.
The defining companies won't win by collecting more industrial data since most industrial organizations already possess decades of it. They'll win by transforming operational expertise into software that improves through deployment, preserves knowledge that would otherwise retire, and becomes trusted with increasingly consequential work.
Spotlight writes the first checks into founders building that future across manufacturing, logistics, construction, energy, and the broader physical economy. We back the software and infrastructure that improves how work is done inside the world's most operationally complex industries.