# Satya Nadella: AGI Is Nonsensical Benchmark Hacking. Here's What Actually Matters. *The Builder Weekly* ยท https://www.youtube.com/watch?v=Qm4o2dL1tQE ## Summary Satya Nadella discusses how AI agents are being used inside Microsoft, including a multi-agent orchestrator built by a networking engineer to automate fiber DevOps. He explains his vision for proactive agents tied to user intent with transparency, and defends his comment that AGI is benchmark hacking, arguing the real measure of AI is economic impact in areas like health care and education. ## Key points - A Microsoft networking engineer built a multi-agent orchestrator with low code/no code and Foundry tools to automate manual fiber DevOps. - Nadella compares AI tool empowerment to Excel as the world's most ubiquitous programming tool. - An on-device proactive agent demo on Copilot Plus PCs summarized email in Outlook without internet. - He cites Mark Weiser's vision that technology should be powerful enough to disappear, and agents should act on high-level intent while staying inspectable. - Nadella says his AGI comment was not a shadow on AI research but a critique of celebrating tech companies over technology's real impact. - Health care is 19-20% of US GDP, and he wants the multi-agent health care orchestrator used by Stanford to become ubiquitous. - A World Bank study in Nigeria gives statistical evidence that giving students Copilot changes educational outcomes. ## Notes ### Agents inside Microsoft - A Microsoft networking engineer faced manual DevOps for the AI fiber network, handled via emails and calls for fiber cuts. - She built an agent, in fact a multi-agent orchestrator, using low code/no code and the Foundry tools. - Nadella compares this to Excel as the greatest, most ubiquitous programming tool: empowering people with tools and agency over the workflow closest to them is what needs to happen. ### Proactive agents and control - A demo showed an on-device proactive agent on Copilot Plus PCs opening Outlook and summarizing email without internet. - Nadella cites Mark Weiser's quote on ubiquitous computing: technology is powerful enough to disappear. - The goal is to interpret intent, plan, and get things done with the least friction. - Computer use agents must become fast, better, and more reliable, but more importantly be chained to a high-level intent of the user, who can inspect the work. - He praises the coding agent's session log with draft commits as the right balance of proactivity and transparency across multiple invoked agents. ### AGI as benchmark hacking - Asked about his viral comment that AGI is nonsensical benchmark hacking, Nadella says it was not a shadow on great AI research. - He argues society celebrates tech companies far too much versus the impact of technology. ### Real value: health care and education - Health care is 19-20% of US GDP, and much of its cost is workflow inefficiency that providers cannot tame despite digital systems. - He would love to see the multi-agent orchestrator for health care, just launched and used by Stanford, become ubiquitous so providers deliver better care at lower cost. - In education, a World Bank study in Nigeria provides real statistical evidence that giving people Copilot or any other agent changes outcomes. - Nadella says that kind of impact is why he joined the tech industry, and he dislikes the industry celebrating itself. ## Transcript [00:00] Copilot or anywhere else. It's just a diffusion of the tools. I'll just give you you know one of the best examples at Microsoft was I was talking to an engineer in our networking our AI one, right? I mean I was talking about man the AI one is growing like crazy. The number of fiber operators we work with is like some you know insane number and by the way all the devops on it right all you know if there's a fiber cut somewhere is all manual, right? Literally it is emails and calls and so on. So she just said okay I can't deal with this. I have to really automate this because otherwise you know all of Microsoft will be just working on devops of fiber. And so she just built an agent and in fact a multi-agent orchestrator, right? [00:41] And by the way she just did this using some of the low code no code and the foundry tools, right? That ability to empower our people with tools and agency to go and look for the workflow closest to them. That I think is what needs to happen. Just like how we were able to use you know I think I I always say Excel is the greatest programming most ubiquitous programming tool on the world. And all of us learned how to program in Excel essentially and that's the thing that I think is going to happen again. And we don't have much time here but I want to quickly touch on proactive agents. So behind the scenes we actually got a demo of like on device proactive agent working on the Copilot plus PCs where [01:23] this agent essentially opened up Outlook and summarize email without internet. But that's just one example. Are there any examples that you have that you're really excited for? Yeah, so I think that if I look at the continuum, right? So I I'm really looking for in some sense I'm a big fan of this Mark Weiser quote about [music] ubiquitous computing, right? Where technology is powerful enough to disappear, right? [music] So therefore to some degree the march of natural user interface and so on is about doing all the things that I intend to be done with the least amount of I would say friction, right? So that's why I think when I want like interpret my intent and plan and get it done, right? So that's where I think some of [02:11] these proactive agent stuff comes in. And I I do I want us be in control though. Like for example, that's why I think the the core agents for example, the computer use agents have to sort of get to a point one that got be fast and better and more reliable. But the more importantly they have to be chained as part of some intent of mine, right? It's not like I'm going to go give [music] task task task task. I'm going to have a high-level intent someone's going off and doing a bunch of work and I can inspect. So that's why I think like this get up you know coding agent the session log, it's pretty cool. Like I go back after having a sign it's like I'm watching the session log with all the draft commits, right? So that I think is [02:55] the balance, right? So in an interesting way we talk about the proactive agent and a session log and that way then it's sort of more transparency of it and then when you invoke multiple of them, you are able to still track effectively what's happening. One last question for you and this is something that recently went viral about two months ago. You said you mentioned that AGI is just some nonsensical benchmark hacking [music] and the true value of AI is global economic growth. Where do you see agents having the most value in that regard first? It's a great one. I mean to me take health care. What is it? 19-20% of our GDP in the United States is in health [music] care. A lot of that cost the inefficiency. If you talk to any of our [03:44] providers [music] is all in the workflow, right? They're working so hard to provide unbelievable care except sort of you know and they're using all these digital systems but they're really not able to tame the complexity and reduce the cost. if there is going to be one place where I would love to love to see that productivity gain would be like that you know that multi-agent orchestrator for health care that we just launched used by Stanford become ubiquitous. And there are stories that where people are saying oh wow, this has been the biggest thing that you know every provider is able to provide better health care at lower cost, right? So something like that has to happen because my only comment there was not [04:27] about any sort of shadow on any great AI research. It is more I think we as a society celebrate tech companies far too much versus the impact of technology. Quite honestly, if there was a more balanced way to talk about sort of not the tech industry but the use of technology, right? That's why even the you know Copilot being an Think about the decades all of us have spent saying >> [music] >> can we have one tech intervention that makes a damn difference in education? Right? Finally, you have real statistical evidence thanks to that World Bank study in Nigeria. That you know you give people Copilot there or any other agent it doesn't matter something changes. That's the story to me, right? That's [05:15] why I joined the tech industry. I mean it was a very different time and suddenly it became like you know the place you were celebrating ourselves and I just hate it. [music] I just want to get to a place >> [music]