It was a great week for innovation across the model ecosystem. We're bringing these new models into Copilot, enabling it to take on increasingly complex work, from quick questions to delegated tasks and, increasingly, complete long-running jobs through Autopilots. Here's one fun example: using an Opal-powered Autopilot running on... a secure Windows 365 Cloud PC, Copilot goes through a month of trail cam footage, finds every animal sighting, creates a highlight reel of the best clips, labels each one with the camera, date, and species, catalogs every sighting in a spreadsheet, builds a PowerPoint summarizing the findings, and shares everything in Teams for colleagues to review. This is the next frontier for Copilot: software that doesn't just assist with work, but can own and complete entire tasks that unfold over hours or days.show more

Learn more about Opal, the technology powering one of these new Autopilot experiences (in Frontier rings now and coming to Copilot soon):

Helpful NINJACAT is here again. Always supporting you Microsoft 🐾

i want @microsoft to lose billions on AI copilot crap

The trail cam example is a good pick because it's tedious enough that nobody actually wants to do it manually. Curious how well the species labeling holds up on blurry nighttime shots though.

Copilot is overdue for a major upgrade. Microsoft should build it more like Grok Bot, with one tool completing a full task across several apps from start to finish.

does it ever come back and ask what counts as an animal, or does it just guess and move on

Satya is anyone using this?

Congratulations! Copilot is getting very very impressive. Big shout outs to the team! 👏

let’s talk about the Microsoft 🤮UI and 🤮UX for how Coplit has been butchered into the Microsoft ecosystem of Enterprise at work. adding a Copilot button within 50+ Microsoft products is a failure of 🤮UX that only Microsoft could make! Grade: D- borderline F-Fail. WE WANT 💥ONE APPLICATION 💥THAT CAN READ AND WRITE TO ALL MICROSOFT PRODUCTS INCLUDING SHAREPOINT. not 50 applications, just 1 application layer. do you get it stupid? now fuc! off for wasting so so much of my time over the years in this swamp of Microsoft applications! Comparing Microsoft ‘AI’ vs Cursor or Claude shows Microsoft has literally no idea about productivity! zero, none, ziltch. this Copilot implementation is evidence Microsoft has no vision and is doomed to cater to legacy workers and even worse legacy engineers and Product Managers! the worst part is honest criticism like this will be pushed deep below the layers of Bureaucrazy across the many many teams for each product (Word, Excel, Teams etc). it takes a Leader to stand up to the internal noise and say enough is enough. The people want and need ONE APPLICATION (not 50 zillion)

O

Dude just get basics right pls 365 copilot basic is fucking up context ignoring the last message and referring to even older messages,it's crazy how bad it is at keeping to prompts.

This is the real Copilot shift: from interface to labor. Once software owns the workflow, the question becomes which jobs still need a human.

Cool use case but I feel like you can do so much more. AGI is here

An agent going through a month of trail cam footage and building the whole report on its own is wild. Feels like "completes tasks over days" is the real next step, not just faster chat replies

This is a big jump.

The AI features in windows are such bloatware and don't work well. Just make an OS, give us the option to purchase an AI enabled version. Instead of forcing things on people that they don't want while they continue to use other better AI models. Compare the sales and get back to me, I bet the bare windows will get more sales.

the playbook never changes: pick a task nobody wanted to do, show ai doing it slower than a human would, call it “the next frontier.” investors eat it up, actual users ask why their copilot still can’t summarize an email thread reliably

managing a month of trail cam footage with an autopilot is wild lol. genuinely curious how it handles the edge cases like blurry frames or weird lighting.

The shift from assisting to owning tasks is the bigger story. AxiomOS fits naturally into a world where agents execute across tools, data, and workflows.

真正的变化是任务开始按“小时到天”计算了。如果 Copilot 长任务跑到一半被中断，它会从 checkpoint 接着跑，还是重新规划？

Nice demo, but Opal on a Windows 365 Cloud PC has been around since Ignite 2025, so this reads more like a maturing feature than a new frontier. Curious if it keeps one name longer than its predecessors did (Scout, Claw, Cowork, Tasks). A month of trail cam footage is actually a solid patience test for an agent, not just an intelligence one.

Microsoft needs to get on the AI race Satya Microsoft IS A SOFTWARE COMPANY

I know you don't care about consumers anymore. But nobody (consumer) wise wants this.

Best use cases to jump start Copilot inside a company: 1. Look for work that happens in the browser, spans several systems, slows down because of volume, follows relatively stable rules, and produces something a person can review. That includes: • collecting audit evidence across internal systems • reconciling invoices, contracts, and purchase records • monitoring competitors and updating a living brief • triaging support or IT queues and escalating exceptions • assembling recurring financial or operating reports • reviewing large media libraries and cataloging their contents • preparing vendor due diligence packets • tracking a research question as new evidence appears • maintaining recruiting pipelines across sourcing and applicant systems • turning customer feedback into categorized product evidence The safest early jobs are the ones where progress is visible, exceptions can be escalated cleanly, and the final artifact can be reviewed before anything consequential happens. 2. The best first Autopilot jobs concentrate human judgment. The agent carries the repeatable path, preserves the evidence, and surfaces only the points where the rules no longer determine the next action. A job is ready for delegation when its uncertainty can be localized into reviewable exceptions. That means you do not need perfect autonomy before long running agents become useful. You need a workflow where the routine path is stable, the ambiguous parts are recognizable, and the handoff back to a person is clear. That is how long running agents create leverage before every edge case has to be solved.

सर्वे भवन्तु सुखिनः, सर्वे सन्तु निरामयाः AI’s greatest promise is not smarter machines, but smarter governance transparent, consistent and accountable decisions that expand fairness and opportunity for every citizen and for nature. Technology should reduce inequality, not privil

Almost all the apps from copilot ecosystem are pure garbage. To anyone who is reading this - just use codex, it is the best general purpose app out there

The jump from basic queries to actual autonomous agents is huge. Interested to see how these long-running jobs handle edge cases in real-world workflows.

Curious how Autopilot handles error recovery in long-running jobs like this — e.g., if it misclassifies a species mid-way through a month of footage, does it self-correct on review, or does that require human intervention at the end? That's usually where 'autonomous' workflows break down in practice. #BharatronTech

That sounds like a huge step forward! It's amazing to see how much these models are improving.

And none of that innovation came from your company let they sink in!

The demo's real message isn't the trail cams, it's that the task touched a Cloud PC, Excel, PowerPoint, and Teams in one pass. Autopilot isn't selling autonomy, it's selling the reason to keep every one of those seats licensed.

Handmade Origin Artwork of sir, Satya Nadella. Look the artwork carefully. Can you guess.? Stay tuned '⭐'

The magic won’t be an AI that makes a deck. It will be one that remembers the half finished work from last Tuesday and quietly gets it done before you have to ask again.