In yesterday’s earnings name, we shared that Microsoft 365 Copilot has surpassed 30 million paid seats, with internet seat provides greater than doubling quarter over quarter.1 This is a crucial milestone, nevertheless it’s additionally a sign that factors to a wider shift: Throughout each trade, AI is transferring from assistant to lively participant, turning into a vital a part of how work will get carried out, relatively than only a device individuals use. Organizations are redesigning workflows round it. And staff are directing brokers to tackle duties, initiatives, and processes, increasing their company and growing their influence.
The primary part of enterprise AI was outlined by adoption—deployments, licenses, and hours saved—actual positive aspects that can proceed to compound. The following part will probably be outlined by transformation. At Microsoft, it’s our mission to empower each particular person and each group on the planet to attain extra, and we’re excited by the rising alternative to meet that mission as people and brokers collaborate to construct what was beforehand not possible.
New knowledge from Microsoft 365 Copilot telemetry alerts, together with examples from our prospects around the globe and groups inside Microsoft, reveal this transformation in motion. It exhibits up in 3 ways: in how deeply and broadly individuals now use AI, in what AI makes potential—from sooner work to net-new capabilities—and in how organizations tailor it to particular roles. Let’s take a better look, beginning with the know-how itself.
A step change in innovation
Up to now few months, a brand new form of AI product has emerged: a mannequin wrapped in an agentic harness that may shut its personal suggestions loop—planning, executing, testing, and correcting its work to ship a completed end result. This method enabled Copilot Cowork, which grew to become usually accessible in June. Cowork can tackle multi-step work: you outline the duty, and it runs it finish to finish and returns a accomplished end result. Its multi-model design allows you to run the fashions a process wants, so functionality scales with the work as extra fashions turn into accessible. And testing exhibits that Cowork was 30% to 40% cheaper2 in comparison with a single mannequin choice. We additionally launched our first autopilot agent, Microsoft Scout, which stays lively within the background, with its personal identification and permissions, to maintain work transferring even when your consideration is elsewhere. The entire outputs are grounded in Work IQ to know the context that shapes work inside your group, together with individuals’s roles, initiatives, priorities, and institutional data—so they’re uniquely related to your online business.
We’ve seen a step change not simply within the merchandise we ship to prospects, but in addition in the way in which our engineering groups construct them. Whereas many staff have contributed to Cowork, the core group stays deliberately small—rising from three engineers in the beginning to simply 9 when the product grew to become usually accessible. And but the speed-to-market was outstanding, with Cowork transferring from inception to preview to being utilized by half the Fortune 500 all throughout the house of six months. The breakthrough got here from rebuilding the software program growth course of round brokers, permitting a tiny group to attain ranges of velocity, scale, and automation that may have been not possible with conventional engineering workflows.
Microsoft Scout additionally began with a small group of individuals working alongside autonomous brokers with shared context, evaluations, and clear specs, and once more moved from an preliminary idea to a working product in a matter of weeks relatively than months. “Up to now, I spent a number of time serving to groups scale: aligning massive teams, coordinating work, and ensuring the whole lot moved ahead collectively,” says Omar Shahine, Company Vice President of Microsoft Scout. “Now it’s extra about creating the circumstances for professional groups to maneuver rapidly, experiment safely, and constantly seize what they study so each challenge begins additional forward than the final.” Frontier groups like these—nimble, targeted, and outlined by human-agent collaboration—present a powerful sign for a way data work will rework within the months and years to come back.
Depth and breadth of use
As AI merchandise and the method to constructing them mature, Copilot utilization throughout organizations can be deepening and broadening. Within the previous yr alone, the variety of conversations per person has practically doubled,3 whereas the variety of customers partaking with a number of Copilot options grew by triple digits.4 And the truth that common weekly engagement with Copilot is now on par with Outlook and Groups5 speaks to the velocity at which AI is turning into part of on a regular basis work.
Once we take a look at what individuals ask Copilot Cowork to do most in a given week, we’re seeing an increase in requests that span a number of steps, like analyzing a set of information and drafting the e-mail that explains it or researching an issue and writing the code to unravel it. Once we account for these multi-step workflows, analysis-related work accounts for 49% of all duties, up from 29% when evaluation is carried out as a standalone process.6 In different phrases, the info suggests AI is transferring past serving to individuals full duties to serving to them full total workflows.
Zooming out, early adopters like manufacturing, banking, and software program/know-how stay among the many industries with the best variety of lively brokers within the Microsoft 365 ecosystem. However what’s stunning is that industries like automotive and healthcare are rising as among the fastest-growing agent adopters,7 signaling that agent worth is spreading effectively past the early leaders.
The clearest signal of this momentum could also be how rapidly organizations are scaling Copilot throughout their workforces. The variety of our prospects with greater than 50,000 seats has elevated greater than 7x yr over yr. In the meantime, the variety of enterprise prospects deploying Copilot to the vast majority of their info staff grew practically 75% quarter over quarter, a sign of how central Copilot has turn into to their operations. And the time it takes to achieve excessive utilization, that means month-to-month lively utilization above 80% throughout a prospects’ person base, has fallen from months to simply days over the previous yr.8 Taken collectively, these shifts point out that organizations are more and more embedding AI and brokers into the circulate of on a regular basis work.
What AI makes potential
At its most elementary, AI-enabled velocity creates efficiencies. However the organizations which are getting this proper are transferring past effectivity for effectivity’s sake and beginning to see how AI modifications the work itself—and may even increase what a corporation is able to doing.
For instance, Microsoft’s personal Cloud Provide Chain group took a “lean earlier than brokers” method. It first mapped and simplified six end-to-end workflows and constructed a shared knowledge basis, then deployed greater than 70 purpose-built brokers throughout planning, sourcing, achievement, and logistics. The end result was a 75% discount in cycle time throughout chosen workflows, with the lesson baked in: simplify the work itself earlier than you automate relatively than layering brokers onto a damaged course of.
Cycle time is one measure of progress. The standard of the selections beneath it’s one other. At clever energy administration firm Eaton, root trigger evaluation had lengthy relied on time-intensive investigations. To lighten the load, the corporate has begun to make use of an AI-powered high quality agent that’s serving to consultants overview stories on manufacturing high quality points to rapidly floor potential root causes, tendencies, and provider insights. Utilized throughout roughly 5,000 stories to date, it has accelerated evaluation, strengthened data-driven choice making, and enabled earlier danger identification and prevention actions whereas maintaining Eaton’s professional group on the middle of each choice.
Different organizations are discovering what’s potential by empowering these closest to the issue to construct the options. Premera Blue Cross got down to let staff create their very own AI brokers to take repetitive, guide work off their plates and sort out rising backlogs and guide workflows. The well being plan, which serves members throughout Washington and Alaska, began with placing Microsoft 365 Copilot and Copilot Studio into the instruments their individuals already used day-after-day, backed by coaching and clear guardrails so staff may experiment with out placing member belief in danger. The group has since constructed greater than 900 brokers and counting, together with a contract exhibit agent whose clever doc processing minimize a guide workflow from 30 to 45 minutes right down to about three—releasing up staff to deal with members who want entry to care or solutions to crucial questions.
Additional alongside the spectrum, AI can unlock the flexibility to do work that beforehand couldn’t be carried out in any respect. At Levi Strauss & Co., the place AI is being woven throughout the enterprise from shopper perception to human assets (HR), an agent analyzed 1,100 finance commonplace working procedures and cataloged 18,000 particular person duties in a day, creating an in depth view of labor that wouldn’t have surfaced in any other case. “The agent sized a chance I’d assumed was not possible to quantify,” stated Lisa Sterling, Vice President of Finance at Levi’s.
And at world consulting group EY, an Autonomous Sourcing Agent (ASA) automates low-cost, low-risk transactions throughout the procurement course of, from requisition and negotiation to buy order creation. It could negotiate with one or many suppliers directly throughout value, supply timelines, and guarantee phrases whereas maintaining people within the loop for validation and escalation. Since its October 2025 pilot, the ASA has supported greater than 200 transactions and is anticipated to deal with 1,500 over the subsequent yr. The result’s a sooner, extra environment friendly procurement course of that helps groups get extra worth from suppliers whereas making certain vital judgement calls stay with individuals.
Specialization at scale
The distinction between the organizations seeing modifications like these and people nonetheless ready isn’t how broadly they’ve deployed AI—it’s how effectively they’ve tailor-made it to the work. But the commonest deployment sample continues to be a general-purpose one: give everybody the identical device and allow them to determine it out. The organizations creating essentially the most distinct worth are doing one thing totally different—beginning with how particular roles really work, then constructing AI round these patterns.
One instance: S&P World is displaying how AI could make deep, specialised experience accessible to extra individuals by bringing trusted intelligence immediately into the circulate of labor. As demand for AI-ready knowledge grew, the corporate used Microsoft Copilot Agent Builder and Microsoft Graph connectors to create a Copilot agent in Microsoft Groups, giving prospects a less complicated technique to entry and work together with its commodities analysis and market insights. By delivering 95% sooner knowledge extraction and 98% sooner comparative evaluation, the answer helps prospects flip huge quantities of data into actionable intelligence extra rapidly, empowering sooner, extra knowledgeable choices throughout world vitality and commodities markets.
Microsoft’s gross sales group mapped distinct vendor personas to devoted AI brokers—Researcher, Analyst Agent, Gross sales Agent, Deal Agent, and MSXi Copilot (a gross sales intelligence and reporting agent)—every designed to suit particular moments in how sellers spend their time. With brokers taking up extra administrative and coordination duties, vendor time devoted to customer-facing actions doubled, from 25% to 50%. The pilot group noticed 9.4% extra income per head and 20% sooner deal closure, and in a survey reported a 3x enhance in adoption of precedence use instances.9
Kantar took a ground-up method. On the advertising knowledge and analytics agency, a “individuals agent” for HR queries (actually a system of about 10 brokers with 60 subjects and associated flows working collectively) is being constructed and expanded via collaboration by individuals that truly deal with and help these requests throughout its Folks group. It already resolves a rising share with out human intervention: an employment verification letter that when took three days is now returned in 120 seconds. Kantar stories the share of HR queries dealt with this fashion has risen from zero in February to about 40%, in opposition to a year-end goal of 95%.
Microsoft’s Folks Operations group runs greater than 1,000,000 worker interactions a yr throughout greater than 100 methods, doing work that carries actual weight for individuals. They’re utilizing Frontier Tuning to scale HR experience throughout their processes and worker moments relatively than bettering one process at a time, capturing the experience of practitioners as duties, expertise, and rubrics that outline high quality requirements. The result’s an AI-powered system that may tackle operational work with the context, requirements, and institutional data crucial for fulfillment, reworking their practitioner know-how into organizational functionality. The group can see how the agent performs on actual work and tune the levers—instruments, context, expertise, and the mannequin itself—to make sure constant efficiency. The agent developed with Frontier Tuning is now stay in Folks Operations, delivering 5x enchancment in profitable process completion in comparison with a baseline general-purpose mannequin.10 As routine actions are more and more dealt with by AI, Microsoft HR’s Folks Operations consultants can deal with the complicated worker interactions, nuanced choices, and strategic work the place human perception issues most.
What’s subsequent
As AI know-how advances and organizations in all places start to see actual worth emerge, the measures that matter are altering as effectively. We’ll see elevated deal with depth of use, measurable change within the work itself, the expansion and governance of brokers, the redesign of roles and workflows, and the emergence of capabilities that didn’t beforehand exist. The organizations constructing sturdy benefit are those the place know-how, working fashions, and human company are evolving collectively. That’s the Frontier Agency taking form.
Footnotes
1Microsoft This fall Earnings name.
2This evaluation was carried out by inside Microsoft groups. The evaluation in contrast 125 check runs throughout a complete of 12 mild, medium, and heavy work knowledge prompts in Copilot Cowork and Claude Cowork with their Microsoft 365 connector, each utilizing mannequin Opus 4.8. For Copilot Cowork, the prices have been calculated utilizing variable charges primarily based on related fashions, context, instruments, and runtime from our inside logs. For Claude Cowork, the prices have been calculated utilizing publicly accessible API charges primarily based on token utilization and Microsoft 365 connector utilization. We used these prices to get a complete value per run and in contrast the costs throughout mild, medium, and heavy immediate units between Copilot Cowork and Claude Cowork with their Microsoft 365 connector. Value financial savings outcomes primarily based on assessments carried out in June 2026. Precise prices and potential financial savings might fluctuate relying on utilization, configuration, time, and different elements.
3Microsoft This fall Earnings name.
4Based mostly on anonymized, aggregated Microsoft 365 Copilot productiveness alerts, depend of monthly-active M365 Copilot customers who use 2 or extra distinctive Copilot-based options over a 28-day interval. Contains all international locations and enterprise prospects in June 2025 and June 2026.
5Microsoft This fall Earnings name.
6Based mostly on anonymized, aggregated Microsoft 365 Copilot productiveness alerts, Cowork telemetry from July 19, 2026 to July 26, 2026, analyzing the commonest duties prospects delegated to Copilot Cowork in the course of the reporting interval.
7Based mostly on anonymized, aggregated Microsoft 365 Copilot productiveness alerts, Month-to-month Lively Brokers measures the depend of distinctive lively brokers noticed via telemetry throughout Declarative Brokers, Customized Engine Brokers, and SharePoint Brokers. The metric is calculated over a rolling 28-day window. An agent is counted as “lively” if, inside that 28-day interval, it meets a minimum of considered one of two standards: (1) it has a minimum of at some point of user-initiated utilization, or (2) it completes a minimum of one autonomous run. Every qualifying agent is counted as soon as no matter what number of instances it was used, making certain a de-duplicated measure of the lively agent base. Progress is assessed over a 12-month interval spanning June 2025 via June 2026, evaluating the lively agent depend in the beginning and finish of the interval.
8Microsoft This fall Earnings name.
9Inside Microsoft gross sales group knowledge primarily based on 687 sellers of Microsoft 365 Copilot from January to June 2024, as in contrast with sellers with low utilization of Copilot.
10Success was outlined as a “excellent run”: the duty achieved a 100/100 rating in opposition to human-defined analysis standards. Duties have been evaluated throughout a number of executions to account for variability in completion paths, process sequencing, and power choice, with solely constantly excellent outcomes counted as profitable.
