Building a Sector View on Vertical AI Through Pattern Research
Developed an independent investment thesis on vertical AI by researching companies across industries and testing a hypothesis about workflow ownership.
Swastik N.
Private Equity Intelligence - Junior Analyst at Gain
From their time as
Private Equity Intelligence - Junior Analyst
Gain • 2025
Overview
Swastik developed a view on vertical AI that diverged from the dominant foundation model narrative, and he built it through research rather than reading.
The Story
Swastik developed a view on vertical AI that diverged from the dominant foundation model narrative, and he built it through research rather than reading.
His starting point was a question: if foundation models are becoming commoditized, where does the durable value actually sit? He selected companies across legal AI, coding assistance, customer support, healthcare AI, and enterprise productivity tools. For each, he asked a single question: what are customers actually paying for? Were they paying because the company had the best model, or because it solved a specific workflow problem?
He read product documentation, company blogs, customer case studies, and where available, technical documentation. He then compared companies doing similar things. If multiple products used the same underlying models but competed on workflow, integrations, and customer experience, the competitive advantage was not the model. That observation became his working hypothesis: workflow ownership matters more than model ownership.
Rather than treating that as a conclusion, he kept testing it. Every new AI company he read about became a data point. The pattern held across industries, customer types, and use cases.
His conclusion: the long-term value in AI would shift toward companies embedding AI into specific workflows rather than building another LLM. Foundation models are becoming infrastructure. The moat is in the workflow, the domain data, and the switching cost that comes from deep integration.