The biggest issue I see is Microsoft's entire mentality around AI adoption that focuses more on "getting the numbers up" then actually delivering a product people want to use.
Most of the announcements I hear about Copilot, it's always how they've integrated it into some other piece of software or cut a deal with yet another vendor to add it to that vendors product offering. On the surface there's nothing wrong with doing that but that just seems to be the ONLY thing Microsoft is focused on.
Worse yet, most of these integrations seem like a exercise in ticking boxes rather than actually thinking through how integrating Copilot into a product will actually improve user experience. A great example was someone mentioned that Copilot was now integrated into the terminal app but beyond an icon + a chat window, there is zero integration.
Overall, MS just reeks of an organization that is cares more about numbers on a dashboard and pretty reports than they are on what users are actually experiencing.
Or, scaling back trying to keep their datacenter bill manageable.
Used to be one could upload an unlimited number of files (20 at a time) and process them directly at the initial window --- now one has to get into "Pages Mode", and once there, there's a limit on the number of files which can be uploaded in a given 24-hour period.
It feels like that's the entire MO of the Azure platform as well. Make a minimum viable product and then get adoption by selling at all costs, despite the products edges.
The wild thing is, the business prop is so clear - an llm built into your corporate data, with the same security, guard rails, grc auditing stack that protects the rest of your data. Why integrate and exfiltrate to an outside company?
But copilot is fucking terrible. Sometimes I ask it powershell questions about microsoft products and it hallucinates answers. Get your shit together microsoft, why would I use this product for any reason if it doesnt work squarely inside your own stack
> "The biggest issue I see is Microsoft's entire mentality around AI adoption that focuses more on "getting the numbers up" then actually delivering a product people want to use."
That succinctly describes 90% of the economy right now if you just change a word and remove a couple:
The biggest issue I see is the entire mentality that focuses more on "getting the numbers up" than actually delivering a product people want to use.
Back in early 2023, the state of google search was abysmal (despite that their leaders insisted it wasn't, it had become nearly unusable for me and I don't think was that unfounded of an opinion). Microsoft rolled out a new version of bing, which became bing chat - search worked for me again for a very brief window of time.
They could have pounced on this opportunity to take a big chunk out of google's search, because google didn't really catch up there til the AI overview was rolled out, and even that is notorious for having issues. Eventually chatGPT seems to have carved out some of this search space with web-search being native to the tools now.
But microsoft was way ahead of everyone here for a brief period! Instead they just rolled everything into bloatware vaguely called "Copilot" and called it a day.
>it had become nearly unusable for me and I don't think was that unfounded of an opinion
if ironic is the right word; the (google) search product itself still is. if not even worse.
the 'new' ai mode routinely creates these silly categories that are not what i was looking for and my screen is filled with repetitive ai summaries of articles. it will ingest a source as fact, and then use that fact to create confirmation bias across other articles. it will even use words like "confirm" when it finds a source saying something, even if the source is junk or seo spam. it becomes somewhat impossible to escape the assumptions the model has made, and i have to resort to traditional web search to get diversity in my results.
and while deep research works, its so overly verbose, with no easy way to tone down the wordiness.
I don't use it often, but at least now I can get an answer. I swear in early 2023 I would just get completely irrelevant, borderline spammy results to the point I gave up and felt helpless because there was no real alternative at that time for how I used google. It felt like the internet broke for a window of time and Bing (very briefly) brought me out of that hell. To this day I still can't believe they didnt capitalize on it.
Over and over Microsoft kills products with mis-marketing.
One scenario is the product is good (OneNote) but they put three icons on the taskbar for it and spam the rest of Windows for ads for it that just make people scream "take it away!"
Another scenario is that the product is bad (OneDrive) and they push you into having a traumatic experience (Microsoft Office uses it as the default save location and when it is down you can't save your work!) that makes sure you'll never use it again -- even though now OneDrive seems to be basically reliable.
Today is it the dominant playbook for marketing of AI experiences. Mostly people are sick and tired of hearing about it, the master Unique Selling Point of 2026 is products that don't interrupt you when you are trying to get work done.
Recently had to download actual Adobe Reader for the first time in at least a decade and... christ. Requires most of an H100 in resources and you can't do what you actually want to do because of multiple AI related popups and attempt to get you to subscribe to some Adobe cloud nonsense.
I knew it would be bad but I couldn't believe the state of it, just utter garbage
Because most of Microsoft's revenue is not generated by end-users. It's large government agencies and big corporations where the end-user is ten steps detached from the actual decision to buy or not to buy something.
It's a story in Germany all the time that some open source zealots get a town government to switch to an off-brand office suite which is so bad that the government worker's union goes on strike to get Microsoft Office back.
How's that different from Red Hat Linux? I mean, Linux is all about corporate takeover by IBM, Google and the like. The mainstream of Linux GUI is Android for crying out loud and X and Wayland are rounding errors compared to that.
From the article, "its productivity software is used by hundreds of millions of corporate users, a captive audience to whom it can easily promote new AI products."
Their end users are what they ultimately sell. They are captive audiences. This is what monopolies/platforms do. It's never been part of MSFT's DNA to care that much about end user experience. Who they really cater to are the IT decision makers, etc. These people can then show some numbers about "AI adoption" and "productivity" gains on their Power Point slides presented to their bosses. MSFT's value is delivering that to them.
It is remarkable how during the last 25 years (approximately), Microsoft has been improving their ability to deliver first (or be among the first), followed by messing up the whole process so that late comers end up taking the crown jewels.
PDAs, mobile phones, tablets, tablets with detachable keyboards, managed OS userspace, HoloLens, the XBox mess, and now AI.
There certainly other examples that I failed to address.
This is what happens when divisions fight among themselves for OKRs and whatever other goals.
Until they messed up the whole UWP / WinRT developer experience in Visual Studio.
Also VS 2026 was released with a hard milestone, thus while there is a new settings experience, many options show a dialog from VS 2022, because the new UI is still not implemented for the new experience.
Note that most organisations have to pay for Visual Studio licenses, and get rewarded with such quality.
From what I could infer from some community talks, podcasts and so, I would assert that nowadays they have the problem new hires have been educated in UNIX like OSes and Web.
Thus Windows team gets lots of folks that never coded anything for Windows, and management instead of having proper trainings in place, just goes with Webview2 and Electron all over the place.
I might be wrong, this is more my perception than anything else.
Ah, thanks. That must be the link with Simon Peyton Jones as well. Seems to be another case of a marketing machine running away with foundational research coming from Europe.*
Don't worry, after a decade or two of having Windows reinstall and re-enable it every couple weeks against their users' wishes I'm sure they'll get the market penetration they're looking for...
This is repeated endlessly by non-Windows fans I assume, because I disabled AI and other annoyances in Windows long ago and they haven't come back. I even used to worry about updating Windows because I saw this warning so many times, but then I did and it just never came to pass.
Microsoft has a long and well documented history of resetting user preferences.
Multiple times I've disabled the cortana taskbar search widget, only to have a windows update turn it back on and proudly gives me a popup telling me they noticed it was disabled and turned it back on for me.
Microsoft will forcibly re-enable AI features eventually. Again, this is an established pattern for them.
I think the plain ordinary chatbot behind the Copilot on the desktop is fine, it seems like a skin around ChatGPT-5 in the "Smart" mode and in the "Search" mode it compares to Google's AI mode.
When it comes to anything multimodal it is an absolute disaster. Show it a photo of a plant for a plant id? Forget about it, just take a picture of the screen on your phone with Google Lens. If you ask it to draw something or make a Microsoft Word document you'll regret it.
For advice about how to do things on the command line or how bootstrap works or how to get out of a pickle you got yourself in Git it is great. It writes little scripts as well as anybody but you can't trust it to get string escaping right for filenames in bash scripts which is one reason I'd want help. For real coding I use Junie because I'm a Jetbrains enthusiast but other people seem to swear by Claude Code.
I do dread the day though when Microsoft decides to kill Copilot because I will miss it.
They really dropped the ball on this - they are down ~12% for the year.
When they first started, they seemed to be firing on all cylinders and looked like they were going to be big winners, but the strategy has just been a slow motion car crash.
I wonder if Satya is the right person for Microsoft.
It was a fresh air after Balmer and he helped opening the company to open source, naturally not without their own intentions, however Satya has been a disaster for the consumer branding, anything related to Windows.
Microsoft's focus was making it so that AI could allow unskilled workers to replace skilled workers. The hope was that everyone but sales/management could be offshored to SEA/India/etc and AI would somehow make up for the skill differential.
The successful AI companies are making it so that skilled workers can use AI as a tool to be more productive and efficient.
Believe it or not, the Recon Analytics trend is actually worse primary usage among Copilot subscribers dropped from 18.8% to 11.5% since July while Gemini climbed past it.
People who paid are leaving.
That's a churn problem.
The tell is buried in the article: workers who have access to Copilot, ChatGPT, and Gemini side by side choose ChatGPT and Gemini at higher rates.
Some companies are using 10% of their paid seats. Microsoft's CMO of AI says growth is "unlike anything we've seen before" but won't share the numbers.
That's the "we're thrilled with preorders" of AI.
This is the Ballmer story all over again.
- Massive distribution advantage
- Captive enterprise base
Somehow still losing to the thing people actually want to use.
Windows Phone had carrier deals too.
The problem is the same: you can't mandate delight.
This part is laughable, can't believe it leaked:
> "About a year ago, Nadella sent a frustrated email to Rajesh Jha,
> executive vice president of experiences and devices, detailing an incident in which
> Nadella had asked the enterprise version of Copilot on the Edge browser
> to help with a public webpage he was on,
> but it couldn't fulfill his prompt"
Meanwhile three different orgs inside Microsoft all own something called "Copilot" and none of them talk to each other.
Meanwhile, Anthropic ships Cowork after 10 days and it just explodes with the market.
The reality is that Copilot’s laughable performance is almost entirely unrelated to AI models not being good at X.
Every single thing Copilot does has been solved much better by other products.
However, Copilot fails in extremely ridiculous ways, at very basic tasks which such a product absolutely must nail.
Copilot should not have been released.
A large majority of people involved have failed. People like managers, product managers etc should probably be fired. Technical leads equally so.
For everyone who has been building similar products it is immediately obvious that Copilot is sloppy, unfocused and unprofessionally executed.
People hate it, and for hood reason.
It just boggles the mind how they would go and release it, or that it even exists in its current form.
Those devs and managers rake in hundreds of thousands of dollars each, producing garbage that has been done better by dozens or hundreds of other teams
Maybe Microsoft needs to fix the cart before they put the jet engines on top of it and try to kill the horses off.
Go back to fixing what’s wrong with Windows, then worry about the AI software running on top of it and where you can add a value proposition, because right now the Windows value proposition is continuing to go right down the shitter as everyone flees Windows 11.
Contrast that to the Linux desktop which "just doesn't work" and my M4 Mac Mini that amazed me with how fast it was when I bought it and a year later it is beachball... beachball... beachball... reboot. beachball... beachball... beachball... Doesn't help that they vandalized the UI by adding meaningless transparency effects which don't actually look cool but rather look like they added anti-antialiasing to the edges of everything for now reason.
Some are more tech savy than others here, but I guess almost anyone can do the following trick successfully:
step 1. visit https://endeavouros.com/
step 2. download iso
step 3. flash iso on medium
step 4. boot medium, installation window shows
step 5. you choose KDE, yes: KDE. Do more mouse clicks.
step 6. system tells you it's done, and offers you to reboot.
