Chart of the Day | Coatue Summary

1. Key Themes

The Agentic AI Tipping Point in Enterprise

Agentic AI (task-execution) has overtaken conversational AI (question-answering) as the dominant mode of enterprise AI usage. Per the chart, by June 2026, agentic Codex represents 64% of enterprise output tokens vs. ChatGPT's 36% — a near-complete inversion from August 2025, when agentic usage was near zero.

"Enterprise usage is shifting from asking AI questions to letting it do the work."

Model Releases as Adoption Catalysts

Each major product release — GPT-5 (Aug 2025), Codex general availability (Oct 2025), GPT-5.2 (Dec 2025), Codex for macOS (Feb 2026), Codex for Windows + GPT-5.4 (Mar 2026), GPT-5.5 (Apr 2026) — corresponds to a visible step-change acceleration in agentic token share. Model and platform launches are acting as demand unlock events, not just capability upgrades.

(Source: OpenAI, 28-day trailing average — as cited on the chart)

Token Volume as the New Enterprise AI Metric

The chart measures "share of enterprise output tokens," signaling that token consumption (not seats or logins) is becoming the core unit of enterprise AI engagement. Agentic tasks, by nature, consume far more tokens per session than single-turn Q&A, implying rapid growth in total token volume even if user counts grow modestly.

2. Contrarian Perspectives

Conversational AI (ChatGPT-style) Is Already a Minority Use Case in Enterprise The consensus narrative still treats ChatGPT-style chat interfaces as the primary enterprise AI paradigm. The data contradicts this: as of June 2026, chat represents only 36% of enterprise output tokens and is in structural decline relative to agentic usage. The shift happened faster than most market observers anticipated — from near-zero agentic share in August 2025 to majority share by early 2026, roughly 6 months.

Chart shows agentic Codex crossing the 50% threshold around February–March 2026, less than 6 months after Codex general availability.

Platform Availability (Not Just Model Intelligence) Drives Agentic Adoption The sharpest inflection in agentic token share occurred not at a model intelligence milestone but when Codex launched a native macOS app (Feb 2026) and subsequently a Windows app (Mar 2026). This suggests that reducing friction to agentic workflows — via native OS integration — matters as much as raw model capability in driving enterprise adoption.

3. Companies Identified

OpenAI

Description: Developer of ChatGPT and Codex
Why mentioned: The chart is sourced from OpenAI's own enterprise token data; their Codex product is the vehicle driving the agentic shift
Quote (chart source): "Source: OpenAI. 28-day trailing average."

Coatue Management

Description: Prominent technology-focused hedge fund and venture investor
Why mentioned: Publisher of the C:\Takes newsletter and the analysis; framing the agentic shift as a key investment theme
Quote: "Enterprise usage is shifting from asking AI questions to letting it do the work."

4. People Identified

No specific individuals are named or quoted in this edition of the newsletter.

5. Operating Insights

Prioritize Agentic Workflow Integration Over Chat Tooling Enterprises still investing primarily in chat-based AI interfaces are likely optimizing for a shrinking share of AI value creation. The data suggests operators should audit their AI stack and shift resources toward agentic tools that execute multi-step tasks autonomously — coding, research, process automation — where token consumption (and therefore AI leverage) is orders of magnitude higher.

Native Platform Presence Unlocks Enterprise Agentic Adoption The Codex macOS and Windows app launches each produced visible inflections in usage. For AI product builders targeting enterprise, this implies that web-only or API-only distribution meaningfully constrains agentic adoption — native OS integration is a go-to-market lever, not just a UX nicety.

6. Overlooked Insights

The 28-Day Trailing Average Understates Peak Agentic Share The chart uses a smoothed 28-day trailing average, which means the actual real-time agentic share at the end of June 2026 is likely higher than the 64% shown. The trend line is still steeply rising at the chart's right edge, suggesting the 64/36 split may already be outdated — a detail easy to miss but significant for investors modeling near-term trajectory.

Agentic Dominance Is Being Measured Within a Rapidly Expanding Total Because agentic tasks consume vastly more tokens per task than chat, ChatGPT's 36% share likely still represents growing absolute token volume — the overall pie is expanding explosively. This means ChatGPT is not necessarily declining in absolute terms, only in relative share, which has implications for OpenAI's total revenue growth that the percentage framing alone obscures.