Work rarely happens in neat lines. It moves across apps, tasks, teams, and the shifting realities of the day. We’ve redesigned the Copilot app and how Copilot shows up across Microsoft 365 apps to better move with it: cleaner, faster, and in the flow of your work.
From the start, our goal wasn’t to simply layer AI onto familiar tools, but to use it in ways that help you move more directly from intention to outcome. Shaped by your feedback, the new designs shift Copilot toward a more connected, adaptive system by turning a once static text box—the prompt line—into a task-aware workspace.
Now, within the Copilot app, the prompt line gives you more space to express your needs, while below it, Copilot surfaces tools and controls to assist with the task at hand. We’ve also created a single, flexible entry point for Copilot across Microsoft 365 apps that suggests relevant actions to help you in your work.
AI experiences are defined by the quality of what they make possible, and Microsoft is evolving its design approach to support this. In the AI era, the most important user experience for human-centered design to shape isn’t the interface—it’s the output. The tone, structure, readability, usefulness, and trustworthiness of Copilot’s responses impact whether your work moves from a rough idea to real progress, whether the task is small and immediate or complex and strategic.
For the next wave of Copilot design, we stepped back, simplified, and reworked key parts of the experience to meet your needs with more craft, intention, and speed.
The Copilot app: Design and performance that move together
In studying how Copilot is used in real working rhythms, it became clear where the experience could be faster, clearer, and more helpful over time. Our new designs introduce a more intentional interface and a faster, more responsive experience. A beautiful UI that doesn’t keep pace with the person using it breaks the moment. An answer that arrives quickly but lacks coherence, shifts the burden from waiting to reworking. Speed, structure, and output quality must be holistically designed for truly performant experiences.
To craft intelligence that feels present but not imposing, we applied the long-standing design principle of progressive disclosure: starting with a clean, focused interface, then revealing more capabilities as you need them. At the interface level, a left navigation pane that expands and contracts reveals a clearer space for agents, conversations, and history. At the same time, a shared pinning system and more room for session recall make returning to work in progress easier.
The prompt surface can expand to fill the experience, making room for deeper work: pasting content, retaining structure, and using inline formatting before sending. Rather than presenting every path at once, this design organizes what matters first and reveals more capability in context, making the experience easier to navigate, understand, and trust over time.
Progressive disclosure also shapes output. Copilot begins with a clear, readable response, then adds structure and next-step support as you refine what you need: formatting when it improves clarity, suggested prompts when they deepen the work, follow-up actions when they move it forward.
That progression is powered by Work IQ, an intelligence layer you can see when active and directly control. Drawing on your emails, files, chats, and meetings, it adapts to the depth your work requires: quick responses when they fit the task, and deeper reasoning—including the ability to choose between AI models—when that can surface more relevant results. By grounding in your broader context, not just individual artifacts, Work IQ helps Copilot better support significant shifts, like performance review cycles or an org change.
The result is output that unfolds in layers, with more guided, detailed, and actionable support as complexity grows. That sequence matters because work rarely begins with perfect clarity. It starts as half-formed questions, rough sentences, and problems still undefined. Thoughtfully designed experiences help bridge that gap, turning looser thinking into clearer direction, and clearer direction into outputs that fuel real progress.
None of this holds, however, if the experience doesn’t keep up—the Copilot app is now faster, more responsive, and more reliable. It loads more than twice as fast, with load times reduced by over 50%,1 and response times for complex chat prompts have improved by 10%.2 We’ve also improved the quality of responses. With more structured outputs that are easier to scan and better aligned to your intent, it moves you faster from idea to outcome. The result is an experience people rely on more consistently. In early customer feedback, users describe the updated Copilot app as cleaner and more intuitive, with design choices helping them better focus on their work.3
Copilot across Microsoft 365 apps: A cohesive, agentic experience
This momentum is also evident in using Copilot across other apps. Since rolling out the new in-app experiences, Copilot usage has increased by 27% in Word, 33% in Excel, 43% in PowerPoint, and 30% in Outlook.4
The new experience shifts how, when, and where Copilot shows up across Microsoft 365 apps to better support how you naturally work and think. Software has long taught people to move between modes—think here, create there, refine somewhere else—but thinking and making are often entwined. You find the story while building the deck, interpret the signal while organizing the data. Copilot should honor that reality, and powerful new agentic modes are creating new paradigms.
Through capability-focused agents like Designer, Researcher, Word, Excel, and PowerPoint, interactions now fit the task at hand, rather than forcing a generic one. They evolve Copilot in Word, Excel, and PowerPoint from a tool that responds to prompts within a single document, into an experience that can take action, draw on broader work context, and more independently operate inside the apps where you already spend your time. That shift demanded a new design foundation, and everything from the entry point to the interaction model has evolved to support it.
We introduced a new experience for Copilot in the apps—a consistent entry point across apps that sits above your work and understands the context beneath it. Rather than scattering touchpoints across the interface, it anchors Copilot as one connected system across Microsoft 365, surfacing relevant actions that help you stay in flow.
From the new entry point, Copilot opens a side pane that works directly with your document: not just as chat, but as an editing partner that can suggest changes or make them, with clear signals so you always know what it’s doing. Copilot can also be invoked on the canvas itself, within a paragraph, cell, or slide, so the interaction begins where the work already lives. That movement between canvas and chat reflects how real work unfolds: not in separate modes, but in a continuous loop where ideas take shape as you build.
The bigger shift
These updates are more than a refreshed interface. They’re a broader shift in how we design AI for work. We’re moving from individual features to connected experiences. From adding capabilities to shaping outcomes. From asking people to adapt to technology, to shaping technology around how people actually work.
As AI accelerates what can be built, the question becomes what should be built. Design exists to answer that question. We are building experiences that are faster, yes, but also clearer, more useful, and more human. In that, craft isn’t a decorative finishing touch; it’s how intelligence communicates care.
1 Based on customer testing conducted March 10–17, 2026, comparing approximately 11.06 million users in a treatment group to 11.16 million users in a control group. Load time improvements are measured using User Perceived Load Time, defined as the maximum latency of key app load components (including the left navigation pane, prompt input box, and prompt suggestions). Results reflect observed performance under test conditions and may vary based on user environment, device, network conditions, and usage patterns.
2 Based on customer testing measuring Chat First Token Response time, defined as how long it takes for Copilot to start responding in the Microsoft 365 Copilot app after a prompt is submitted. Results show approximately 10% improvement in response time at the 95th percentile, which reflects performance for the slowest 5% of chat requests. Measurements compare system performance before and after changes under controlled test conditions. Results may vary depending on prompt complexity, user scenario, device, network conditions, and other factors.
3 Based on early qualitative research conducted by an internal Microsoft research team including 8 interviews and 79 survey responses from customers participating in a feedback program. Findings reflect directional user feedback and are not necessarily representative of all Copilot users. Experiences may vary.
4 Based on product usage data comparing Copilot usage among commercial users before and after rollout of new experiences. For Word, Excel, and PowerPoint, usage changes reflect a comparison of activity from May 8–12, 2026 after rollout versus May 1-5, 2026 before rollout. For Outlook, results reflect a comparison of daily active usage from January 27-February 24, 2026 after rollout versus December 30, 2025-January 27, 2026 before rollout. Results reflect short-term changes observed during these timeframes and may not be indicative of long-term usage trends. Results may vary by user, organization, and usage patterns.
