AI Commoditizes Public Market Models
Engineering leader at Tegus on building a data platform for expert interviews
AI is turning hand built public market models from a premium product into a faster, cheaper feature. Canalyst’s core job was to take filings and earnings calls, map them into a clean Excel model, and update that model quickly after each quarter. That used to require a large analyst team. Once LLMs can read a call transcript, pull the numbers, and update a template with acceptable accuracy, the moat shifts away from model assembly and toward proprietary content, workflow integration, and trust.
-
Canalyst sold speed and structure. Its product library includes more than 4,000 pre built models, near real time earnings updates, and APIs and Python tools. In plain terms, customers were paying for someone else to do the spreadsheet work first, then hand them a model they could start editing immediately.
-
The weak spot was labor intensity. Tegus described Canalyst’s process as analysts listening to calls live, copying numbers into models, and racing to publish updates within hours. That is exactly the kind of repetitive extraction and formatting work AI compresses first, even if human review still matters for the last mile.
-
That is why Tegus bundled transcripts, filings, and models into one research seat, and why AlphaSense later bought Tegus for proprietary content rather than just search. Public information is increasingly easy to synthesize. The scarce asset is unique information, plus a product that drops it directly into the investor’s workflow.
The market is heading toward model generation on demand. The winning platforms will not charge mainly for maintaining a giant human modeling bench. They will charge for trusted outputs, fresh proprietary inputs like expert transcripts, and distribution into chat, Excel, APIs, and the rest of the investment workflow.