Overview of 20VC: Why OpenAI and Anthropic Won't Win the App Layer | Why Teams Will Get Bigger Not Smaller in a World of AI | Why AI Removes Incumbents Advantage of Bundling | China vs America: Who Wins the AI War with Arvind Jain, Co-Founder @ Glean
Harry Stebbings interviews Arvind Jain, co-founder of Glean and former co-founder of Rubrik, about the future of enterprise AI, frontier models, open source, bundling, pricing, and how AI will reshape teams. The central thesis: model companies are not the enemy for most AI businesses — they’re an enabler — while the real battleground is context, workflow integration, cost, and enterprise control.
Key Themes and Main Takeaways
Frontier model companies are an asset, not necessarily a threat
- Jain argues that for most AI companies, OpenAI, Anthropic, Google, and open source model providers should be viewed as a massive capability layer, not direct competition.
- The frontier labs make it possible for application companies like Glean to deliver products they could not have built alone.
- He believes the real competition is in owning the workflow and enterprise context, not just the model.
Enterprises are increasingly worried about operational dependence
- Enterprises are not just worried about data privacy; they’re worried about becoming dependent on external model providers for critical work.
- Jain’s point: once AI agents handle core business operations, the enterprise risks losing control over the “learning” embedded in that work.
- He frames this as a shift from software dependence to operational dependence.
Open source is becoming more important, mainly because of cost
- Jain says the move toward open source is driven primarily by economics, not just ideology or sovereignty.
- Many enterprises want flexibility, control, and lower inference costs.
- He expects most enterprise workloads to run on open source models within three years.
Chinese models are technically compelling, but geopolitically sensitive
- He acknowledges that Chinese open source models are becoming highly competitive.
- The biggest blocker is not performance, but trust, paranoia, and geopolitical risk.
- In his view, many enterprises are willing to use Chinese models if they are self-hosted and under their control, but concern remains around backdoors, reputational risk, and regulation.
Bundling still matters — but consumption pricing weakens it
- Jain says Microsoft’s bundling strategy is one of Glean’s biggest competitors.
- Bundled products can be hard to beat, especially in enterprise procurement.
- But he argues that AI’s shift toward consumption-based pricing reduces the power of bundling, because customers pay based on usage rather than just seat count.
AI ROI is real, but uneven
- He sees clear value in some areas:
- Customer support
- Some internal knowledge workflows
- Basic question answering and summarization
- But he says many organizations still struggle to see clear ROI in areas like coding, where productivity increases don’t always translate into faster product shipping.
- His view: a lot of AI adoption is improving throughput, but not always final business output.
Glean’s Philosophy on AI
Context is the real unlock
- Jain says most enterprise AI deployments fail because they’re too shallow: companies connect models to systems through basic MCP integrations and expect magic.
- His argument is that AI needs the right context, structure, and workflow design to work well.
- Without that, AI becomes:
- slow
- expensive
- noisy
- less useful than it should be
Glean uses AI heavily internally
- Glean’s engineering team writes almost all initial code with AI now.
- However, they still enforce human code reviews.
- Jain notes that AI has made code generation much faster, but also raised concerns about long-term maintainability and quality.
Team Structure, Hiring, and the Future of Work
He does not believe teams will shrink
- This was one of the strongest disagreements in the interview.
- Jain argues that in an AI-driven world, companies will need to do 10x more work to win, so teams will likely get bigger, not smaller.
- His logic:
- If one company shrinks while a competitor keeps scale, the competitor can use the same AI tools and simply produce more.
- Competitive pressure will favor companies that can deploy more talent alongside AI.
Roles will become more composite
- He expects more hybrid roles that combine functions:
- engineer + product manager + designer
- sales + solution engineering + business development
- In his view, specialization will decline in some areas in favor of more generalist, AI-enabled operators.
Some roles will disappear or be absorbed
He called out several jobs as likely to shrink or change dramatically:
- Data analysts doing mostly dashboard work
- Sourcers in recruiting
- Some HR support roles
- More administrative analyst-style work in business functions
AI Pricing, Competition, and Market Structure
The model layer is already commoditizing
- Jain believes the model market is getting intensely competitive.
- He thinks open source and frontier model competition will create significant pricing pressure.
- He expects inference to become much cheaper over time, even if prices have temporarily risen.
Current AI is still expensive
- He gave a striking example: Glean once spent about $1 million per month on a triage agent.
- His broader point is that AI’s current economics are often overstated, and many deployments are not yet cheap enough to replace human labor cleanly.
He believes AI should get much cheaper
- Jain’s long-term view is that AI must follow the historical pattern of technology: lower cost, higher accessibility, broader adoption.
- He does not believe current pricing is the end state.
Startups, Capital, and Founder Mindset
Too much capital can distort startups
- Jain thinks there is too much capital available in the startup ecosystem.
- This can cause founders to:
- overpay for talent
- build unsustainable structures
- lose discipline
- He still believes discipline matters: companies must create real value, charge properly, and earn ROI.
He sees the market as a land grab
- Glean is moving aggressively because he believes this is a land grab moment.
- Enterprises want AI products now, and it will be much harder to win those accounts later.
His own founder style is cautious and paranoid
- Jain says he is motivated more by fear of losing than by a desire to win.
- He describes the founder/CEO role as:
- stressful
- emotionally demanding
- never fully satisfying
- His philosophy: good CEOs should remain a little unhappy, because there is always more to improve.
China vs. America and the Sovereign AI Question
Sovereign models are desirable, but the urgency has shifted
- Jain says the desire for sovereign models remains strong, but it was stronger a year ago.
- He thinks many countries realized they may not be able to build full foundation-model stacks themselves.
- Still, he sees a real need for the US to develop strong open source alternatives and not fall behind China.
The US should promote its own open ecosystem
- He argues the US cannot afford to look non-innovative in AI.
- He expects support from major players, including NVIDIA, to help foster open source development in the US.
Notable Quick-Fire Takeaways
- Best legacy company at adopting AI: Google
- Advice to computer science students: keep studying it; don’t panic
- Biggest startup ecosystem problem: too much capital
- What makes being CEO hard: it’s not glamorous; it’s relentless
- What changes with success: not much, if you’re disciplined
Bottom Line
Arvind Jain’s core message is that enterprise AI winners will not be defined by who owns the best model alone. The winners will be those who:
- understand enterprise context,
- integrate deeply into workflows,
- control costs,
- and build products that make AI usable in real business operations.
He is bullish on AI, bullish on Glean, and unusually skeptical that AI will automatically shrink teams or make bundling obsolete overnight. His view is that the AI era will reward more ambition, more execution, and more operational scale — not less.