Harry opens with a candid confession: he recorded this episode while fasting and got unusually combative with his guest, which he believes produced one of the best discussions the show has had in recent memory. He teases the listener with the question of whether they prefer Happy Harry or Hangry Harry. The three sponsor segments cover Asana's pitch as an operating system for human-agent teams, MongoDB's case as the data platform of choice for AI agents (citing ElevenLabs running 40 million agents and 75% of the Fortune 100 using MongoDB), and AlphaSense's Super Analyst product as an always-on market intelligence tool. These reads set an AI-native commercial tone that mirrors the episode's central themes.

Harry cites Palantir CEO Alex Karp's claim that the largest enterprises in the world are more skeptical than ever of frontier model providers. Arvind unpacks two distinct fears: the risk of core IP and data leaving enterprise control, and a deeper operational dependency he finds genuinely alarming. He explains that when AI agents repeatedly perform business tasks, they accumulate institutional learning — the undocumented, optimised processes that make organisations competitive. If enterprises don't own those agents, that knowledge accrues to OpenAI or Anthropic. This isn't just technology dependence; it's operational dependence on a scale the industry hasn't seen before. ^{[1] — Arvind Jain "As AI agents repeatedly perform business tasks, they accumulate institutional knowledge. If enterprises don't own and control those agents,…" 07:10 Play} Harry presses on whether enterprises are moving towards open source to address this, and Arvind confirms the shift is accelerating — driven primarily by cost, not data privacy fears.

Harry forces Arvind to address the competition question head-on: Anthropic has already launched vertical product packs for Figma, legal, and health — will enterprise be next? Arvind argues those packs are shallower than they appear and are expanding the market rather than cannibalising it: non-designers using Claude Design are not displacing Figma users. He acknowledges that Claude's primary use case — question answering — is exactly Glean's core, and that MCP connectivity means enterprises already ask why they need Glean at all. His answer is that context is hard to build properly, and first-mover brand is valuable if not sufficient. He frames all frontier model progress as good news for Glean because it improves the underlying models his platform uses. ^{[1] — Arvind Jain "Frontier models are becoming a commodity. Arvind Jain says 90%+ of enterprise use cases can already be handled by open-source models, and G…" 14:03 Play} The segment closes with a crucial pivot: 90% of enterprise workloads can now run on open-source models, setting up the commoditization debate.

Harry reframes the competitive landscape: forget OpenAI and Anthropic — Microsoft's playbook of building a 70%-as-good product and bundling it into enterprise agreements has worked for decades. Arvind concedes Microsoft is one of Glean's most significant competitors and that Glean frequently loses deals simply because the prospect already has Copilot in their Microsoft agreement. Harry challenges the consumption-pricing-breaks-bundling argument with a practical objection: large enterprises like VW, GE, and Ford approve one vendor and won't manage 15. Arvind partially concedes, noting that pricing — specifically the difficulty of competing with free — is the more common killer than vendor management complexity. The exchange ends without a clean resolution, which feels honest.

Harry declares that 2026 H2 and 2027 will be the years when enterprises stop writing AI checks on faith and demand ROI accountability. Arvind doesn't disagree but provides texture: customer support is the clearest win because productivity is directly measurable (cases resolved per day per agent). Coding is where the biggest AI spend has gone, but the paradox is real — coding speed increased dramatically, but product shipping speed hasn't followed, because coding is only one component of shipping. At Glean itself, Arvind admits it's hard to isolate AI's contribution from team growth and tenured experience. He then delivers the core insight: AI ROI is a throughput problem. ^{[1] — Arvind Jain "Most enterprise AI deployments fail not because models are bad but because agents spend most of their time and tokens searching for the rig…" 24:50 Play} Enterprises deploy AI in a brute-force way, letting agents assemble context from scratch for every task, burning tokens and time on setup rather than actual work. The fix is investing in context infrastructure upfront.

Harry opens a direct provocation: he has sat with the biggest CEOs in the world and every single one is shrinking teams and saying AI makes that possible. Arvind pushes back with unusual force — he plans to grow Glean from 1,000 to 5,000 people, and he has a competitive logic for it. His argument: if two companies have identical AI tools and one shrinks while the other keeps headcount, the larger company can produce a 10x better product or 10x more output. Harry counters that more people create bureaucratic drag and that smaller teams with the best AI and the best engineers will ship faster. Arvind partially agrees this has always been true but insists it's not an AI argument specifically. The two genuinely disagree, and neither fully convinces the other — which makes for some of the most energetic back-and-forth the show has produced. ^{[1] — Arvind Jain "While every major CEO is cutting headcount, Arvind Jain is hiring aggressively — from 1,000 to 5,000 people. His argument: if two competito…" 27:26 Play} The segment pivots to token economics when Arvind drops the bombshell that Glean's own engineering triage agent cost $1 million per month — more than the human team it replaced. ^{[2] — Arvind Jain "Glean built an AI triage agent to handle 95% of engineering production alerts, replacing a 15-person on-call team. It worked — but cost $1 …" 31:00 Play}

Arvind's answer to the 'changed your mind' question is a moment of rare vulnerability from a founder of his stature. His natural style has been disciplined and conservative — make sure customers get value, don't assume unlimited future capital will cover gaps in fundamentals. But his own team is telling him that conservatism risks losing the land grab. He cites Uber as the canonical example that a bad business model can turn good at scale. ^{[1] — Arvind Jain "We are absolutely in a land grab, like, you know, no question. Like every single company in the world wants a product like ours today. Eith…" 40:10 Play} Harry presses whether Glean is really in a land-grab moment, and Arvind is unequivocal: every company in the world wants enterprise AI today, and getting in now versus waiting makes it 10 times harder to compete in the future. It's a confession that even the most fundamentals-oriented founder recognises when the rules temporarily change.

Harry asks Arvind to make specific predictions about the job market. On the emerging side, Arvind is confident about composite roles: the most valuable employees will combine what currently takes 3-4 specialists, whether that's a single person handling engineering, product management, and design, or a go-to-market generalist who can sell, demo, and handle post-sales implementation. Harry playfully catches an internal contradiction: composite roles imply smaller teams, which contradicts Arvind's earlier headcount growth thesis. Arvind resolves the tension by noting that with smaller teams you have to demand 10x the work from each person to deliver the same output — which is effectively the same point Harry was making. On the disappearing side, Arvind calls out data analyst roles that are pure execution (building dashboards on request rather than generating business insight), and recruiter sourcers, predicting both will be absorbed into fuller, more senior roles.

Harry raises a question from the conference floor he's sitting in: we're in Europe, and sovereignty over AI models is becoming a real policy question. Arvind notes that the desire for sovereign national models peaked a year ago when everyone thought they could build one, but most countries have backed off as they've realised the investment requirements. ^{[1] — Harry Stebbings "On OpenRouter, the top six models by usage are Chinese. The first US model — Anthropic — ranks seventh. Arvind Jain and Harry Stebbings deb…" 43:50 Play} Harry then drops a striking data point: on OpenRouter, the top six models by usage are Chinese, with Anthropic as the first US model at rank seven. Arvind acknowledges this but argues the US won't accept this trajectory and that there is significant motivated capital — including NVIDIA — investing in US open-source model development. Harry pushes further, suggesting Sam Altman may be lobbying Trump for regulatory barriers against Chinese open-source models. Arvind pushes back on that scenario but doesn't entirely dismiss it, noting the US system needs to build competitively rather than win through regulation. The section ends with both agreeing that a multimodal competitive world — with US and non-US models — is important for everyone.

The rapid-fire segment moves quickly through a range of topics. On studying computer science: don't panic, it's still the right path. On legacy AI adoption: Google rates highest because it has embraced AI both internally and in products, though calling Google a legacy company is a stretch. On investors: Arvind won't pick a favourite publicly but confirms that investor reputation directly maps to company reputation for talent recruiting. On the startup ecosystem: Arvind's biggest concern is that overabundance of capital is creating failure paths — founders paying $500K engineer salaries at seed stage build structures that are structurally incompatible with long-term competitiveness. On exit options: Arvind disagrees with Harry's pessimism, saying exits are actually easier to achieve today than in any of the past 25 years. On being a founder: it's not glamorous, it's not sexy, and you have to be mission-driven to survive. Having financial security from prior success helps you make more rational, patient decisions — but it doesn't change the fundamental requirement of outworking everyone around you.