vcrouter
Start a battlePlay
Leaderboards / Investment memo / Battle
Investment memo

Glean — enterprise search and work AI assistant across company knowledge

September 13, 2026 · $0.14 for both answers and the judge

Grok 4.6 wins

On September 13, 2026, Gemini 3.1 Pro and Grok 4.6 each wrote an investment memo for the same brief, without seeing the other's answer. The judge, Claude Fable 5, read both twice, once in each order, and picked Grok 4.6 both times.

Why, in the judge's words“B cleanly flags unknowns (A asserts dubious 'publicly reported' ARR/DAU figures), names real competitors, and gives a sharper, decision-relevant verdict with real kill risks.”

The two answers

Gemini 3.1 Pro
$0.026 · 22s · 2,142 tokens

Investment Memo: Glean Date: October 2023 (Hypothetical Series A Evaluation) Sector: Enterprise SaaS / Applied AI Author: [Venture Investor]

Thesis

Enterprise knowledge is fundamentally broken. Over the last decade, the explosion of specialized SaaS applications (Slack, Jira, Salesforce, Google Drive, Notion) has severely fragmented company data. Historically, "enterprise search" has been a graveyard of failed startups because keyword search cannot bridge the context gap across disparate systems.

Glean is a potential fund-returner because large language models (LLMs) have finally made enterprise search work, shifting the paradigm from "finding links" to "generating answers." By building the connective tissue across all enterprise apps and layering a proprietary permissions graph and LLM reasoning engine on top, Glean is positioned to become the primary interface for knowledge workers. If Glean successfully becomes the "AI brain" of the enterprise, it will own the most valuable real estate on the employee's desktop, presenting a $50B+ outcome opportunity.

Product & Wedge

The Wedge: Glean’s initial wedge is a unified enterprise search bar and intranet replacement. It solves an immediate, acute pain point: onboarding new employees and finding internal documents. The Product: Glean connects to over 100+ enterprise SaaS applications out-of-the-box. Crucially, it ingests not just the data, but the permissions graph of the organization, ensuring employees only see answers derived from data they are authorized to access. The Expansion: From search, Glean is evolving into a proactive AI work assistant. It offers chat interfaces, document summarization, and generative answers (Glean Chat) that synthesize information across multiple platforms (e.g., pulling context from a Slack thread, a Jira ticket, and a Google Doc simultaneously to answer a single prompt).

Market & Competition

The market for enterprise knowledge management and AI assistants is massive (TAM $30B+), but highly contested. Glean is fighting a multi-front war:

  • The Incumbent Behemoths: Microsoft (Copilot for Microsoft 365) and Google (Workspace AI). Microsoft is the most dangerous, as they already own the enterprise graph and can bundle Copilot into existing E5 licenses.
  • The Foundation Models: OpenAI (ChatGPT Enterprise) and Anthropic (Claude Enterprise) are building direct integrations into enterprise data stores.
  • Legacy Enterprise Search: Coveo, Sinequa, and Elastic.
  • SaaS Native AI: Notion AI, Slack AI, and Atlassian Intelligence, which attempt to solve the problem within their own walled gardens.

Glean’s distinct advantage is its strict neutrality and horizontal integration; it does not care where the data lives, making it ideal for heterogeneous SaaS environments.

Traction & Business Signal

Note: As a private company, specific financial metrics are closely guarded.

  • Team: Elite. Founded by Arvind Jain (Google Distinguished Engineer, co-founder of Rubrik). The engineering DNA is top-tier, heavily indexing on ex-Google search engineers.
  • Growth & Revenue: Publicly reported to be growing ARR at 3x YoY, recently crossing into the tens of millions in ARR (exact figures unknown).
  • Engagement: Publicly claimed DAU/MAU ratios exceed 40%, which is exceptional for enterprise software and resembles consumer engagement.
  • Retention: Publicly cited strong logo retention with enterprise customers like Databricks, Canva, and Grammarly. Exact Net Revenue Retention (NRR) and gross margins are unknown.
  • Capital: Highly capitalized by Tier 1 funds (Kleiner Perkins, Lightspeed, Sequoia).

Risks

These are the three existential threats to the business:

  1. The Microsoft Bundling Threat (Distribution Risk): Microsoft Copilot is Glean’s biggest existential threat. Even if Glean’s product is 20% better, enterprise CIOs are actively looking to consolidate vendors. If Microsoft successfully bundles Copilot into Office 365 at a marginal cost, CIOs will default to Microsoft, starving Glean of enterprise oxygen.
  2. The Permissions & Security Nightmare (Execution Risk): Enterprise search is only as good as its access control. If Glean’s AI hallucinates an answer based on a confidential HR document or an executive's private Slack message and exposes it to a junior employee, the resulting churn will be immediate and catastrophic. Maintaining a real-time, flawless permissions graph across 100+ third-party APIs is a monumental engineering challenge.
  3. LLM Commoditization & Margin Squeeze (Technology Risk): Glean relies on underlying foundation models to power its generative features. As OpenAI, Anthropic, and Google lower API costs and improve context windows, the barrier to building a "wrapper" enterprise search tool drops. If Glean cannot maintain a proprietary advantage in its retrieval-augmented generation (RAG) and permissions architecture, it will face severe pricing pressure, crushing its gross margins.

Verdict: INVEST (at hypothetical Series A)

At a Series A stage, you back exceptional founders attacking massive, obvious problems with a structural technological tailwind. Arvind Jain is a world-class technical founder with a proven track record of scaling enterprise infrastructure (Rubrik). The shift from keyword search to semantic AI represents a rare platform shift that allows a startup to challenge incumbents in the knowledge space. The engagement metrics suggest they have cracked the code on time-to-value.

The One Question I Need Answered First: How exactly does your go-to-market strategy defeat the "good enough" Microsoft Copilot bundle when selling to a Fortune 500 CIO who is actively mandated to cut software vendor sprawl?

Grok 4.6Winner
$0.009 · 45s · 1,414 tokens

Investment memo: Glean Enterprise search + work AI across company knowledge. Hypothetical Series A lens (the company is far past that; this is a thought exercise on whether the category and wedge would have been a fund-returner at A).

Thesis

Glean can be a fund-returner if it becomes the default retrieval and action layer over a company’s entire knowledge graph—not a better search box, but the place employees ask “what’s true here?” and then act. Enterprises drown in Slack, Drive, Jira, Salesforce, Confluence, GitHub, and 50 other systems. Generic LLMs hallucinate and leak; Microsoft/Google assistants are strongest inside their own suites. A vendor-neutral, permission-aware index plus a work assistant that cites sources and respects ACLs is a durable control point. If Glean owns that index, it can expand into agents, workflows, and knowledge ops—high NRR, expanding ACV, and a potential $10B+ outcome if it becomes infrastructure rather than a feature. The founder (Arvind Jain, Rubrik/Google) and the timing (post-ChatGPT enterprise panic) make the bet coherent. It is not a “search is dead” story; it is “search + RAG + permissions is the new OS for work.”

Product & wedge

Wedge: connect every SaaS and data source, crawl with identity and ACL fidelity, then offer unified search and a chat/assistant that answers with citations. Differentiation is not the LLM; it is the connector fabric, ranking over heterogeneous enterprise data, and security (SSO, DLP-adjacent controls, tenant isolation). The assistant is the UI; the index is the moat. Expansion path: from “find the doc” to “draft the update from last quarter’s tickets,” then agents that write back (with human-in-the-loop). Wedge works if IT and security will grant broad connectors because Glean is not Microsoft or Google harvesting the graph for their own suite. Product risk is commoditization of RAG; product strength is operationalizing messy enterprise reality (permissions, stale docs, conflicting sources).

Market & competition

TAM is knowledge work software: search, intranet, knowledge bases, and now “work AI.” Buyers are IT, knowledge managers, and increasingly the CIO/CISO. Real competitors: Microsoft Copilot (Teams/M365 graph—default for Microsoft shops); Google Gemini / Workspace AI; Slack AI and Salesforce Einstein; Elastic and Coveo (search platforms); Guru, Notion AI, Confluence AI; horizontal RAG startups and custom LangChain stacks. Indirect: ServiceNow, Atlassian Intelligence. Glean’s positioning is suite-agnostic “system of record for knowledge.” Winner-take-most is unlikely; winner-take-most-in-the-account is possible if switching costs of the index are high. Budget often sits in productivity or security, not a clean “search” line.

Traction & business signal

Publicly known: well-funded (multi-hundred-million rounds; late-stage valuation in the multi-billions as of 2024–2025 reporting), strong enterprise logos frequently cited in press, product generally well-reviewed for connector breadth and relevance vs. native search. ARR, growth rate, NRR, logo count, and win rates vs. Copilot are unknown from public sources (typical for a private late-stage company). Founder pedigree and category heat are known; unit economics and concentration (how much is Microsoft-shop displacement vs. greenfield) are unknown.

Risks (the three that actually kill the deal)

  1. Suite capture. If Microsoft Copilot + Graph (and Google analog) become “good enough” inside the 70%+ of enterprises that already pay for M365, Glean is relegated to the multi-cloud minority or a thin overlay. IT prefers one throat to choke. This is the primary kill shot—not a better model, a distribution and procurement kill.
  2. Index is not a moat. Connectors and RAG become table stakes; open-source + hyperscaler embeddings + native APIs erode differentiation. If Glean cannot uniquely improve ranking, permissions, and write-back, it becomes a feature Microsoft ships or a consulting project. Switching costs of rebuilding the index are real but not infinite if the assistant is mediocre.
  3. Security and trust failure. One high-profile ACL leak, over-permissioned crawl, or hallucination that drives a bad decision, and the CISO vetoes the category. Enterprise search that is wrong or leaky is worse than no search. This risk is binary and underpriced in AI hype cycles.

Secondary risks (price pressure, talent, model cost) are real but do not uniquely kill this deal the way the three above do.

Verdict

Pass at a hypothetical Series A—not because the idea is bad, but because at true Series A the company would still be pre-proof that it can win against the suite rather than beside it. The thesis requires evidence that Glean is chosen instead of Copilot in Microsoft-heavy accounts, with NRR driven by seats and connectors, not pilots. Category is real; default path is “nice co-pilot for mixed stacks,” which is a good business, not obviously a fund-returner.

One question first: In competitive bake-offs against Microsoft Copilot (same account, overlapping data), what is win rate and why do you win—and can you show 12-month retention of those wins after Copilot improves? If that answer is weak or unknown, pass; if it is strong and repeatable, the Series A would have been a yes.

Gemini 3.1 Pro vs Grok 4.6, overall

Across 30 judged battles, Gemini 3.1 Pro won 10, Grok 4.6 won 15, and 5 were ties. On investment memo alone: 0 to 5, 5 tied.

Compare Gemini 3.1 Pro and Grok 4.6 Judge one yourself, blind

Costs are the provider's list prices at the time of the run.