Satya Nadella
It’s why today we announced 30+ new Copilot skills across Dynamics 365 Sales, Service and Customer Insights, as Jeff explains here, bringing these CRM capabilities directly into the flow of work. We’ve also introduced Microsoft Copilot Managed Runtime, which provides hosting infra which lets code run safely inside your company’s environment, governed by IT. And it’s just the start for us. With these skills, you can use Copilot across Chat, Cowork, Autopilot, and even Code to complete even more end to end workflows across your business. But it’s not just about getting work done with Copilot.
Because these skills are deeply integrated with your systems of record, we can use tokens where intelligence actually adds value, while relying on deterministic software for execution when that is faster, cheaper and more reliable.
Speaking on the changing architecture of enterprise tools, Nadella explained in a post on X (formerly Twitter) that workers have spent decades having to “work around software”, wasting hours toggling between siloed applications, rebuilding lost context and translating tasks into rigid database commands. Artificial intelligence reverses that dynamic, organising software around actual tasks rather than forcing humans to adapt to fragmented tools, according to Nadella. Nadella emphasised that deploying an intuitive user interface – the “head” – is only half the equation. Enterprise agents require governed access to the “headless” layer where business logic and operational records reside inside CRM and ERP suites. We are entering a new era of truly “personal” software, where you can easily tailor the apps you use to the way you work, but stay connected to the systems of record your business already runs on, preventing fragmentation of your most important systems.
Microsoft CEO Satya Nadella has suggested that software is not going anywhere and once again, argued autonomous agents and generative tools are fundamentally reorganising how people interact with corporate technology, essentially helping people organise software around the work, rather than them working around the software. Nadella pointed to early trends in software engineering on GitHub as proof that agentic tools do not diminish traditional platforms. Despite autonomous coding agents completing large volumes of work, GitHub has experienced accelerating repository creation, along with higher pull-request and commit activity. Microsoft’s overarching ambition is to turn Copilot into a universal operating system for work that spans every underlying model, computing form factor, and daily assignment. So much of what we do at work is still spent working around software. We switch between apps, rebuild context each time, and translate what we want to do into the steps each system expects. AI is changing that, helping us organize software around the work, rather than us working around the software. We’ve already seen an early version of this shift in software development, one of the first areas to adopt agents at scale. As agentic development has grown, GitHub has seen accelerating repo creation and PR and commit activity. The lesson is that more agents don’t make systems of record less important. They make trusted places to maintain information, coordinate changes, and manage state even more important. We see a similar pattern in sales, customer service, finance, and ops. It’s why our ambition is to make Copilot a new OS for work that spans every model, every form factor, and every task. Here’s how he explained this: When you put it all together, you get something much more powerful than yet another AI feature inside SaaS. You start to get what I think of as an infinite SaaS factory. Think about simply going to Code in Copilot, describing what your business needs, and building a customization or an entirely new Saas module. With these plugins and skills, you can extend our Dataverse schema, build on existing business logic, and create new experiences and workflows connected to the systems your business already runs on. We can reinvent systems of record for an agentic world to handle high-volume agent access, bring together context and intelligence across data sources, and work natively with assistants like Copilot. For some customers, that will mean extending existing systems. For others, it will mean replacing them. With Code and these underlying skills, you are doing all this with essentially zero friction, meaning it’s easy to quickly customize what you build and adapt from one project to another as your business changes. And importantly this is not about using an LLM for everything. I believe all this can reset business SaaS. The opportunity is to reinvent systems of record for an agentic world, not just build new experiences on top of them, while giving every business an entirely new way to orchestrate work and build what is missing. That is the direction we are taking at Microsoft: Copilot as a new OS for work, with a governed infinite SaaS factory built in. You use AI every day. Now get your AI Quotient. Take the AIQ test.
And the opportunity is bigger than simply extending the systems we have today.
But a new front end like Copilot (the “head”) is only half the story. We also need a governed foundation that makes the underlying business logic and context historically locked inside individual CRM, ERP, and other apps available to agents (the “headless” layer). After all, a model’s intelligence only becomes useful to a business when it has the right context and the right tools to act. You need a system that can understand someone’s intent, bring in the relevant organizational and business context, and invoke the right workflows. For us, that means bringing together Copilot as the multi-model harness and agentic layer, the business logic in our apps, and the systems of record they run on, combined with connectors to other data sources. Microsoft IQ connects those layers together. This gives Copilot and agents a rich understanding of the business, connecting organizational knowledge with business data, processes and tools so they can reason over the work and act within the rules of the enterprise. It is not just about exposing existing APIs to an LLM, but architecting business context itself for AI.