Hebbia CEO George Sivulka argues technology only becomes transformative once products capture it, casting AI as the seventh such revolution and his company as the vehicle.
Sivulka says Hebbia's spreadsheet-style agents cite sources before generating text, trading some creativity for near-perfect accuracy that users must actively drive.
Hebbia users self-report saving 8 to 20 hours a week, and processes once taking 120 to 200 hours now take 90 to 150 hours on average.
Hebbia's founder says AI needs its Excel, not just chat
Hebbia is betting that AI becomes transformative only when a product captures it, as Excel did for computing. Chat is a first iteration, and Sivulka sees the next one as an interface where humans orchestrate many agents on complex knowledge work. Because the software retrieves citations before writing, it trades some creativity for accuracy, and users drive it rather than trust it blindly.
Takeaways
Products, not technology, capture a revolution's value. Sivulka counts AI as the seventh major technological revolution. Fire, the wheel and computing only transformed life once products captured them, such as Excel and the graphical interface for computing. Chat is a first iteration, like a single threaded calculator.
Hebbia turns spreadsheets into multi-agent workspaces. Every cell can be an agent running over data. Grids, workbooks and pipelines let humans and agents share multistep work such as investment research, drug discovery and finding geological sites. The platform is model agnostic across OpenAI, Anthropic and Google.
Citations first makes answers near-accurate but less creative. Hebbia finds citations before generating text, so each answer is corroborated by a highlight. Sivulka concedes the cost. Output is almost 100 percent accurate but less suggestive, so users must drive the software rather than trust it blindly.
Users report large, partly unmeasurable time savings. Users self-report saving 8 to 20 hours a week. Processes that took 120 to 200 hours now take 90 to 150 on average. Customers keep upselling, though Sivulka says the biggest value, better investment decisions, cannot be quantified.
AI will create jobs and reshape company structure. Sivulka expects most companies to move from hybrid to hybrid AI, with some fully agentic. Technology has historically created jobs, and he expects roles like prompt engineer and agent orchestrator. He rejects universal basic income because the human drive to build persists.
Technology is no moat and B2B SaaS is not dead. Sivulka calls the claim that B2B SaaS is dead ridiculous. Taste, good ideas and people are the hard part of building a company. He compares this to music, where cheap production has not changed the number of songs in the top 100.
Hebbia is an applied reasoning lab that rides model progress. Hebbia focuses on information retrieval over massive private, unstructured data rather than capabilities research, so it improves as underlying models improve. Sivulka expects inference to move increasingly to the edge, where NVIDIA is less dominant than in training.
Tap a name to go there · tap a line to see why they are linked