Interview, Fireside Chat
Why OpenAI and Anthropic Won't Win the App Layer | Glean Founder
- The market faces a "land grab" dynamic where early adoption is critical, with a prediction that entering later will be 10 times harder for companies.
- 90% of enterprise use cases are expected to be fully handled by open source models within three years, leading to the commoditization of the model layer and a potential "race to the bottom" on pricing.
- The majority of enterprise workloads will shift to open source models, driven by cost rather than data leakage fears, as current frontier model pricing is viewed as unsustainable and absurdly expensive.
- Sovereign model development is anticipated to rise due to geopolitical shifts, particularly the inability to rely on US access if banned, though execution challenges and a lack of resources in the US compared to China remain barriers.
- US companies are expected to build their own open source models to maintain technological innovation credibility, despite the high upfront investment required.
- Technology spending by the world's largest companies is projected to increase from 8-12% to 16-20% in the future, fueled by the need to generate 10 times better products to maintain revenue amid rising productivity.
- Per-person productivity will surge, leading to composite roles where generalists replace specialists, forcing smaller teams to deliver work that previously required larger groups and potentially reducing headcount even as tool usage rises.
- Application-level companies are expected to emerge as model providers build ecosystems, while the model business itself may become less lucrative due to fierce competition and open source pricing pressure.
- Consumer models will likely move to consumption-based pricing, eliminating bundling advantages and requiring significantly more effort to generate revenue from customers.
- Large enterprises will make it difficult for new vendors to enter due to vendor management constraints, favoring "compliance as an enterprise bundle" over best-of-breed approaches, though some segments may still support best-of-breed solutions.
- Arvind Jain anticipates Glean's workforce will grow to 5,000 or 10,000 people within five years, with nearly 100% of its code currently written by AI.
- Fear of being eaten by model providers will drive demand for control over context and institutional learnings, prompting companies to seek open source solutions to avoid operational dependence.
- AI spend is currently concentrated on coding, but actual shipping speeds for products have not necessarily increased despite higher coding velocity.
- The overabundance of capital in the startup ecosystem is creating failure paths and unsustainable structures, potentially forcing a shift in capital spending discipline away from conservative strategies.
- Analyst roles that lack business thinking skills are expected to disappear, and HR roles such as recruiting sourcers will be absorbed into full-cycle recruiting positions.