In the years after Amazon launched AWS, I was at VMware, a company under pressure as on-premises workloads started moving to the public cloud. The virtualization revenue that had funded an entire era of enterprise computing was sliding every year. Customers who used to buy in nine-figure blocks were starting their next budget cycle by asking how much of their workload they could move to AWS, and later to Azure. Inside the company, the mood was not panic. It was the quieter dread of a business whose quarterly numbers were still strong but whose customers had begun planning the next 10 years without VMware in those plans.
What turned it around were a few sentences on a whiteboard, sketched in a meeting by Raghu Raghuram, the executive who would later become VMware’s CEO. Essentially it said that people do not want virtual machines. They want to run apps. Apps run in data centers. So let’s virtualize the entire data center. Raghuram’s vision was that customers could spin up data centers on the fly and run whatever they wanted on them, with the efficiency, agility, and economics they were otherwise expecting from the public cloud.
That sketch became the software-defined data center. I spent the next two years leading the go-to-market that turned it into the dominant frame the industry used to talk about infrastructure for the following decade. What unlocked the campaign was empathy with the CIO, who at the time was terrified of losing budget, staff, and control as workloads migrated to public cloud. The pitch I built told them they could have the economics of public cloud inside their own firewall and keep all three: budget, staff, and control. Within a year, every major competitor was bolting “software-defined” onto products they had been shipping for years.
VMware’s real achievement wasn’t inventing virtualization. It was naming a future that CIOs could buy into, and watching the market reorganize itself around that language.
THE RACE TO CLAIM AN AI-NATIVE OR AGENTIC FUTURE
I have spent the last 18 months watching AI companies skip the naming work entirely. Collectively, AI companies have raised over a trillion dollars in private capital. They are spending it building products the market cannot tell apart.
Open the homepage of any AI startup in 2026. You will be told the company is “AI-native.” That it builds “agentic” systems. You will not be told what outcomes it delivers for you, what pain it takes away in a differentiated way. You will be told what is inside the box.
Translation: The market cannot tell these companies apart, because the language used is the same.
This is what happens when an industry uses its own internal vocabulary as its external pitch. “Agentic” is an architectural word. “Evals” is a benchmark word. Neither of them tells a buyer what problems they will be able to solve or outcomes they will be able to achieve on Tuesday that they could not do on Monday.
THE MOVE THAT WORKS
Now consider the opposite move: naming a clear vision for where a product takes you.
Starting in 2013, Gainsight evangelized the customer success manager role and gave it a shared definition, taking a title that existed only in pockets and turning it into a standard profession. The category created demand for its customer retention software, not the other way around. HubSpot pulled the same move with “inbound marketing,” naming both an outcome (customers find you) and a role (the “inbound marketer”) that became a certification, a search term, and a line on a million resumes. Salesforce had done it earlier with No Software, a slogan that named the customer’s new posture toward IT and seeded the SaaS category now worth hundreds of billions. More recently, dbt did it with the “analytics engineer,” a role absent from every org chart in 2019 and now standard at every data-mature company.
To close the loop on where this article started: Software-defined data center did the same work for VMware. The phrase didn’t describe a product. It described where the CIO wanted to stand: public-cloud economics, private-cloud control, and a job that still belonged to the IT organization. Within a year, every infrastructure vendor was using the phrase. Within five, it was the industry’s dominant frame for talking about the next decade of enterprise computing. The naming was the moat, and it held for 10 years.
Every one of these follows the same pattern. The namer didn’t describe what was inside the box. They described what the customer became after buying it. A marketer who attracts instead of interrupts. A customer success manager who owns retention instead of inheriting it. An analytics engineer who ships data products instead of dashboards. An IT leader who runs a hybrid cloud instead of a server room. That’s the move.
The race to build the best AI is crowded and well underway. The race to name what AI is for is wide open. The whiteboard is still empty. Someone is going to write on it.
Arun Lal is SVP of global marketing at Sauce Labs.