Foundation Models Eat Ad Hoc Analysis
Julius
The real risk is that basic spreadsheet cleanup, charting, and one off analysis are becoming bundled into the model layer, which pushes Julius to win on workflow rather than raw intelligence. ChatGPT now natively analyzes uploaded Excel, CSV, PDF, and JSON files, generates charts and tables, and exposes the code it ran. Claude now offers a built in analysis tool and code execution for uploaded files. Google is adding similar code driven analysis inside BigQuery and Looker. That makes the general purpose tools good enough for many technical users doing ad hoc work.
-
What foundation model products still do poorly is the messy middle of business analysis. They work best when a user already knows which file to upload and what question to ask. Julius is stronger when the job includes repeated file handling, cleaning odd columns, building usable visuals, and guiding a non technical person from raw data to a shareable answer.
-
This is the same pattern seen across other AI application categories. As base models absorb generic tasks, the surviving apps are the ones that package the model into a specific workflow with saved context, integrations, permissions, and outputs that match how a team actually works. That is the path that enterprise analytics products like Snowflake and Looker are also taking around model providers.
-
The pricing pressure lands first on hobbyists and power users, because ChatGPT and Claude already include analysis in broader subscriptions and Google is embedding it into existing cloud analytics products. Julius has more room where customers care about governed datasets, repeatability, and a tighter analysis interface than a general chatbot offers.
The market is heading toward a split. General models will absorb everyday analytical tasks, while specialized products keep moving up the stack into opinionated workflows, team collaboration, and enterprise controls. Julius expands its market if it becomes the easiest place to do recurring business analysis, not just the smartest place to ask a data question once.