--- title: The May board productivity number Microsoft just handed every executive slug: may-board-productivity-number-microsoft-just-handed-you date: 2026-05-03 excerpt: "Microsoft's Q3 earnings and the May 1 Agent 365 GA quietly handed executives a board-credible coding-agent productivity scoreboard. Here is the number to walk in with, the number to refuse, and the question Anthropic equity gains forced onto the renewal." featured_image: "https://bbtxujdxvidaghmhxkqs.supabase.co/storage/v1/object/public/generated-images/blog-1777792747537-may-board-productivity-number-microsoft-just-handed-you.webp" featured_image_alt: "A wood-paneled corporate boardroom table at dusk with a single open laptop displaying a coding-agent adoption dashboard, the rest of the table empty, conveying a quiet executive moment of measurement before a meeting." canonical_url: https://cerevisor.com/blog/may-board-productivity-number-microsoft-just-handed-you updated_at: 2026-05-03T07:19:09.09557+00:00 --- # The May board productivity number Microsoft just handed every executive TLDR Microsoft's Q3 FY26 results and the May 1 Agent 365 launch quietly handed every executive a board-credible coding-agent productivity scoreboard: nearly 140,000 organizations on GitHub Copilot, CLI usage roughly doubling month over month, and a $15 per-user-per-month governance line item now landing in procurement queues. The new May question for executives is not "are we using AI" but "what fraction of our engineers are on a harness, and is that fraction doubling like Microsoft's customer base." That number is the one to walk in with. I spent most of last weekend reading Microsoft’s Q3 FY26 numbers, the post-earnings analyses that landed Thursday, and the Microsoft Security Blog post announcing Agent 365 hit general availability on May 1. I do not normally find earnings releases moving. This one was different. Not because the numbers were bigger than expected, but because they collectively handed every executive in the world a brand new productivity scoreboard for the May board meeting. Almost nobody seems to have noticed. The board meeting question last month was “are we using a coding agent.” That question is now retired. The new one is “what fraction of the [engineering org](/blog/harness-platform-team-coding-agents) is actually on a harness, and is that fraction doubling the way Microsoft’s customer base is doubling.” If the answer is no, the follow-up is uncomfortable. If the answer is yes, the follow-up is also uncomfortable, just in a different direction. ## What Microsoft just put on the scoreboard Microsoft reported Q3 FY26 on April 29 and the in-window analyses dropped April 30. The headline numbers everyone quoted were the $37 billion AI annual revenue run rate and the 123 percent year-over-year growth. Those are real, but they are the wrong numbers for an executive scoreboard. The numbers that matter are the adoption velocities buried below the fold. ~140,000 organizations now using GitHub Copilot, with Copilot CLI usage nearly doubling month over month That is the new benchmark. Not “Fortune 500 adoption” in the abstract. Not “AI is everywhere.” A specific, vendor-disclosed organization count, paired with a specific velocity, both of which a board can independently verify by reading any of four in-window write-ups. American Bazaar Online’s April 30 coverage made the numbers explicit. > "Nearly 140,000 orgs now use GitHub Copilot. Usage of GitHub Copilot CLI is nearly doubling month-over-month. Microsoft's AI business has crossed a $37 billion annual revenue run rate, marking a 123% increase." American Bazaar Online, April 30 2026 UC Today’s read on the same release added a second-order signal that I think actually matters more for board prep. Microsoft 365 Copilot paid seats crossed 20 million in the quarter, up from 15 million in January, and year-over-year seat adds are growing 250 percent. First-party agent monthly active usage is up 6x year-to-date. Copilot queries per user climbed nearly 20 percent quarter over quarter. And nearly 90 percent of the Fortune 500 now have active agents built with Microsoft’s low-code tools. Translation for the board: the velocity of agent adoption inside enterprise customers is not slowing. It is accelerating from a base that was already large. --- ## The number to walk in with (and the number to refuse) Here is what I would actually do if I were prepping for a May board meeting this week. Two numbers. That is all. The first is the engineering org’s coding-agent adoption rate, defined narrowly. Not “engineers with a license.” Not “engineers who tried it once.” Active monthly users in the last 30 days, divided by total engineering headcount. Microsoft has nothing like this number broken out for individual customers, but the implicit benchmark in the Copilot CLI doubling stat is that any internal CLI-class usage that is not at least sequentially growing is losing ground against the mean. The second is the month-over-month change on that adoption rate. This is the number that catches a board’s attention because it is intelligible without context. Microsoft’s CLI usage doubling MoM at the platform level is a wild number. Most internal numbers will not double. But if the rate is flat for two quarters in a row, the problem is no longer about the tool. It is about the workflow. Key Insight The May productivity number is not "hours saved." Hours saved is a self-reported estimate that no CFO has ever believed. The number is a monthly active engineer rate on a coding harness, and the month-over-month delta of that rate. Both are auditable. Both are comparable to a vendor benchmark. Neither requires anyone to invent a productivity model. Now the number to refuse. Microsoft Agent 365 hit general availability on May 1 at $15 per user per month standalone, or bundled inside the new Microsoft 365 E7 SKU at $99 per user per month. The Microsoft Security Blog announcement was clear that “Microsoft Defender will be able to block [coding agents](/blog/harness-3-signals-renewal-contract-changing) and generate detailed alerts to support security investigations,” and that support for GitHub Copilot CLI and Claude Code “will follow.” Thurrott confirmed the same wording on May 1 and Winbuzzer reconfirmed on May 2. That $15 per-user-per-month line item is going to land on a lot of procurement desks in the next ten weeks. It is the governance plane for the [coding agents](/blog/ai-replacing-engineers-myth-three-numbers) already being paid for. The question for boards is not “should we buy it.” The question is “do we have the productivity-per-engineer number to defend or refuse it.” If the harness is making engineers measurably faster and CLI usage is climbing, the governance line item probably pays for itself through reduced incident risk. If the harness is sitting on 30 percent adoption and flat MoM, the company is about to layer governance overhead on top of underutilized seats. That is a margin compression story, not a productivity story. I have watched enough Q3 procurement cycles to know how this usually plays out. The line item gets approved without context because nobody in the room has the adoption number. Then someone in finance notices six months later. Then there is a meeting. --- ## The Anthropic question renewal teams have not asked yet This is the part I find genuinely interesting, and a little unsettling. Fortune published a piece on April 30 that I had to reread twice before it sunk in. > "Nearly half of Alphabet's record $62.6 billion profit, about $28.7 billion, did not come from search ads. Pre-tax gains of $16.8 billion included in non-operating income from our investments in Anthropic. Amazon's $8 billion investment is now worth more than $70 billion." Fortune, April 30 2026 About half of Alphabet’s quarterly profit and more than half of Amazon’s pre-tax income for the quarter came from equity gains tied to Anthropic. Anthropic is reportedly in talks at a $900 billion valuation. One in-window analyst piece estimated that “4% of all public GitHub commits are now authored by Claude Code” and that Claude Code “reportedly overtook GitHub Copilot in professional usage.” Both of those secondary claims trace to a single analyst aggregation, so I would not present them as established fact at a board meeting, but the directional point is hard to argue with. Here is the question I would put on the agenda. If a hyperscaler is now reporting AI profitability that depends materially on the same vendor whose harness an engineering org runs on, what does that mean for renewal leverage in Q3? The first answer is that vendor consolidation has quietly created a counterparty problem nobody designed. The team buying AWS or Azure capacity, the team buying Anthropic API tokens, and the team buying Claude Code seats are likely three different teams. The fact that all three vendors are now financially intertwined is not on anyone’s risk register. It probably should be. The second answer is more practical. With Claude Code on a $2.5 billion run-rate revenue trajectory by February and Anthropic’s broader run rate past $30 billion at roughly 80 percent enterprise and developer API, renewal pricing power is not symmetric. The Q3 renewal letter will not arrive with a discount. The defensible posture is to walk in with adoption data showing actual usage, and to ask for unit-economics improvements (caching, batch, gateway routing) instead of seat discounts. --- ## What I would tell you over coffee If we were sitting across a table this week, here is the version I would say out loud. The market just shifted from “do we have AI” to “show me the adoption velocity.” That is a much harder bar to fake. The board members who used to nod through the AI slide are about to start asking for the number, because they read the same earnings coverage and they noticed Microsoft put a specific organization count on the page. Three priorities for this week. Pull monthly active coding-agent users as a percent of engineering headcount, plus the MoM change for the last three months. Get a heads-up on the Agent 365 line item from procurement before it hits the budget meeting; $15 per user per month sounds small until it gets multiplied by full Microsoft seat count. And ask one person on the team to write a one-paragraph note on counterparty risk between the hyperscaler, the model API vendor, and the harness vendor. The point is not to solve it. The point is to put it on a list before the next renewal cycle starts. That is the whole thing. The May board meeting is a chance to look like the executive who saw this coming. Not because anyone predicted Anthropic would carry half of Alphabet’s quarter, but because the right person walked in with a real number when everyone else walked in with a story. The story has gotten old. The number is the new story. #### Sources - [Microsoft Agent 365, now generally available, expands capabilities and integrations](https://www.microsoft.com/en-us/security/blog/2026/05/01/microsoft-agent-365-now-generally-available-expands-capabilities-and-integrations/) - Microsoft Security Blog, 2026-05-01 - [Microsoft Agent 365 Platform Goes Out of Preview and Adds Support for Local AI Agents](https://www.thurrott.com/a-i/335594/microsoft-agent-365-platform-goes-out-of-preview-and-adds-support-for-local-ai-agents) - Thurrott, 2026-05-01 - [Microsoft Agent 365 Hits General Availability With Local AI Agent Controls](https://winbuzzer.com/2026/05/02/microsoft-agent-365-general-availability-local-ai-agents-xcxwbn/) - Winbuzzer, 2026-05-02 - [Microsoft AI business hits 37B as Nadella bets on agentic computing](https://americanbazaaronline.com/2026/04/30/microsofts-ai-business-hits-37b-as-nadella-bets-on-agentic-computing-479945/) - American Bazaar Online, 2026-04-30 - [Half of Google and Amazon blowout AI profits came from a stake in Anthropic, not from their actual business](https://fortune.com/2026/04/30/google-amazon-ai-profits-anthropic-stake-bubble-earnings-2026/) - Fortune, 2026-04-30 - [Microsoft Earnings: AI Business Hits 37Bn Run Rate as Copilot Passes 20 Million Seats](https://www.uctoday.com/unified-communications/microsoft-earnings-2026-ai-copilot-enterprise/) - UC Today, 2026-04-30 - [Microsoft 3QFY26 Earnings: Is AI a CIO Cycle or a User Cycle?](https://nikhs.substack.com/p/microsoft-3qfy26-earnings-is-ai-a) - Lens.Kristal / Nikhs Substack, 2026-04-30