The median company spends $12 per employee on AI. The top 1% Spend $7,400. In July 2026, the median US business on Ramp's platform spent $11.95 per employee per month on AI. The top 1% of spenders: $7,400. That is a 619x gap inside the same economy, measured on the same card platform, in the same month. Within four days, that single number had been spun into three contradictory headlines by three credible parties. Following how that happened tells you more about where AI adoption actually stands than any one of the headlines does. The ladder. AI spend per employee per month, July 2026 The honest scale note is the figure: on a linear chart, the median company's AI spend would not be visible. Ramp's July data has more in it than the ladder. Anthropic is now billed by 43.5% of US businesses on the platform, OpenAI by 39.7% with growth slowing, xAI by 4% and climbing fastest. And Ramp's own headline for the report was not about the gap at all. It was "Cracks in the AI thesis": their economist reads the data as businesses finding "a new upper bound for how much businesses are willing to spend on AI." One dataset, three narratives. Watch the telephone game, dated to the day. August 12: Ramp publishes the index under a spending-ceiling headline. August 14: a16z re-charts the same numbers as a "wild adoption gap," and partner Olivia Moore frames it as the story of the year: "The median company is spending $12 / employee / month on AI. The top 1% are spending $7,500 / employee / month. Not sure we've ever seen an adoption gap quite like this." August 15: Aaron Levie reads the same chart as proof that "AI spend is nowhere close to hitting any walls," with the top 10%'s behavior today previewing the median's behavior in three years. Greg Brockman replies to the 600x chart: "sounds accurate." August 18: Merge launches "Merge for Workforce" straight into the anxiety, claiming a 75% AI spend cut "in one click" via device-level model routing policies: "We just cut AI spend by 75% across an entire workforce in one click... Without Merge, your employees are left choosing one of two options: 1. Spend 100x the cost they should be from using frontier models for everything 2. Deliver subpar work" Shensi Ding, cofounder and CEO of Merge · August 18, 2026 Read the fine print on that one: Merge's own blog headline says "cut token spend in half," not 75%, and no methodology or named customer backs either number. The product itself is real and sensibly boring: an MDM-deployed client that inventories the AI tools employees actually installed, then enforces which models and connectors each team's harnesses start with. Routing policy as IT infrastructure. Coinbase built this in-house in June; now it is a vendor category. What the gap actually is. Ceiling, gap, or no walls: which is it? Held together with what this playbook has tracked since June, the readings collapse into one. Token prices fell ~99% in three years while bills went up, because usage is the variable that matters. The $12 median is not a company using AI efficiently; it is a company whose employees have a ChatGPT seat and nothing wired into the work. The $7,400 top percentile is companies where token spend became a payroll-shaped line item, sometimes badly managed, but structurally committed. The 619x gap is not an adoption curve lagging. It is two different activities sharing the word "AI." The strategic read for a founder or CEO sits in the middle rung: $650 per employee per month, the top decile, is roughly the cost of one junior hire spread across a 20-person team. That is the tier where deployed agents, routed models, and measured savings live. The median tier is a rounding error because nothing real is running. The top tier is a war budget. The decile tier is a decision. $12 is a seat. $650 is a strategy. Which rung are you on? Book a free Diagnostic: 30 to 45 minutes, no deck, no pitch. native first benchmark your AI spend per employee against the July data and find whether you are underspending on deployment or overspending on defaults. 1 Ramp AI Index, August 12, 2026 (Ara Kharazian, lead economist), July 2026 data: median firm $11.95 per employee per month on AI; top 10% $650; top 1% $7,400. Ramp's own headline framing was "Cracks in the AI thesis", reading the data as a new upper bound on willingness to spend. Sample skews toward technology companies. ramp.com 2 Same dataset, July 2026 adoption: Anthropic billed by 43.5% of US businesses on the platform, OpenAI 39.7% with slowing growth, xAI 4% and growing fastest. ramp.com 3 a16z "Charts of the Week", August 14, 2026 (Moses Sternstein), re-charted the Ramp data: "the top 1% of AI spenders are spending more than 600x as much as the median company". Greg Brockman replied "sounds accurate". a16z.news 4 Olivia Moore, a16z, August 14, 2026. x.com 5 Aaron Levie, August 15-16, 2026, reading the same chart as evidence that "AI spend is nowhere close to hitting any walls", with the top 10%'s behaviour today previewing the median's in three years. 6 Merge for Workforce launched August 18, 2026. The launch post claims a 75% cut in AI spend "in one click"; the company's own blog headline says "cut token spend in half". No methodology or named customer is published for either figure. The product deploys via MDM and enforces model and connector policy on the harnesses employees already run. merge.dev · x.com

Note This Week in Extensibility: six vendors standardize agent-plugin bundles, Mozilla stays neutral on WebMCP, Atlassian opens Rovo to Forge apps. Week of July 31 - August 7, 2026: six agent vendors publish a shared plugin-packaging standard, Mozilla files a neutral position on WebMCP that leaves the browser engines split, and Atlassian opens its Rovo agent framework to Forge apps. The theme this week was a split between packaging and trust. Six rival agent vendors agreed on how to bundle plugins while deliberately leaving security out of scope, Chrome began enforcing store rules that police what an extension may do, and the browser engines filed divergent positions on the API that would let a page hand tools to a user's agent. The packaging layer is converging faster than the trust layer underneath it. Packaging: six vendors standardize agent-plugin bundles. Amazon, Cursor's maker Anysphere, GitHub, Microsoft, OpenAI, and Vercel published Agent Plugins 1.0.0, a vendor-neutral format for packaging agent extensions. The specification, announced August 6 in a joint post from Vercel, defines a bundle of a plugin.json manifest, a skills/ folder, and an mcp.json file so one plugin installs across ChatGPT, Codex, Cursor, GitHub Copilot, Kiro, and VS Code rather than being repackaged per client. It is labeled a Working Draft, and it carries no permission model, sandboxing, code signing, or secrets mechanism, with each of those deferred to future work; every client keeps its own trust and marketplace decisions. Why it matters: the same customization, a set of Skills and MCP tools, becomes portable across the major agent clients instead of tied to one vendor's format, so a user is not re-installing per platform. The vendors standardized the bundle while leaving the permission and signing model to each client, so the portability is real and the safety guarantees are not yet part of the standard. Standards: Mozilla files a neutral position on WebMCP. Mozilla recorded a neutral position on WebMCP, leaving the three browser engines split. WebMCP lets a website expose in-page tools that a user's own agent can call. Mozilla's August 5 position is neutral, while Apple's WebKit team has filed an oppose position citing privacy, security, venue, and API-design concerns, and Google and Microsoft are authoring the proposal. WebMCP remains a Web Machine Learning Community Group draft shipping only as a Chrome origin trial, so there is no cross-vendor agreement and no stable implementation. Why it matters: the API that would let any site offer tools to a user's agent now has explicit, divergent positions from every engine, which points to a contested track rather than one converging toward shared support. It does not change that the only running implementation is a single browser's origin trial. Marketplaces: Chrome Web Store begins enforcing its updated policies. Chrome Web Store policy enforcement began August 1 across its extension ecosystem. Four policies published July 1 took effect: a Limited Use rule narrowing data collection to what is strictly necessary for an extension's single disclosed purpose, a duty to notify users of post-install changes to data handling, a Regulated Goods ban on extensions enabling real-money prediction-market transactions, and a Malicious Products clause that bans extensions built to circumvent an AI service's safety guardrails or usage restrictions. Non-compliant extensions face enforcement after the date. Why it matters: a marketplace of this reach setting a hard rule against extensions that defeat AI guardrails draws a governance line at the store level, where the platform, not each AI service, decides that circumvention tooling is not distributable. Platforms: Atlassian opens Rovo to Forge apps and extends its Connect deadline. Atlassian shipped an Early Access Rovo MCP Module that lets Forge apps expose their actions as tools inside Rovo Studio. The August 3 changelog entry lets a Forge app publish its actions as callable tools for makers building custom agents in Rovo, extending the app platform into Atlassian's agent framework rather than shipping a single agent. In the same window Atlassian extended Connect end-of-support to January 31, 2027, moving the deadline out of the year-end crunch, and began direct outreach to vendors still running Connect or hybrid apps so they migrate to Forge. Why it matters: it gives Marketplace developers a supported path to surface their apps as tools inside Atlassian's agents, while the Connect deadline sets the date by which the whole ecosystem must be on Forge to reach that surface at all. Also worth knowing. Merge added a governed catalog of MCP connectors to its Agent Handler. A July 31 changelog lets a company embedding Merge browse and enable hundreds of generic MCP connectors alongside Merge-built ones on the same authentication, access-control, and governance layer, and adds an AI Guardrails experience with PII detection and a live rule tester. Wasmtime patched two sandbox-integrity advisories across four release lines. The Bytecode Alliance shipped v47.0.3 and matching v46, v36, and v24 releases on July 31 to fix two low-severity advisories where engine type indices could be confused and where traps during bulk operations could corrupt VM state. Backporting to the v24 long-term line reflects Wasmtime's use as production plugin-isolation infrastructure. CopilotKit released a Channels SDK for running AG-UI agents in chat surfaces. The August 4 SDK runs one AG-UI agent across Slack, Teams, and other chat platforms with generative UI, human-in-the-loop approvals, and cross-channel memory. It is a developer library for shipping agents rather than a customization surface handed to end users. On the radar. * August 13-14: MCP Dev Summit in Seoul, co-located with Open Source Summit Korea. * September 1: Atlassian Forge Object Store currentVersion field deprecation takes effect. * September 30: Atlassian sunsets the confluence:fullPage and jira:fullPage modules in favor of a unified global:fullPage. * January 31, 2027: Atlassian Connect end-of-support; Marketplace and custom apps must be on Forge. This Week in Extensibility is curated by Todd Schiller. Research, drafting, and fact checking are AI-assisted.