Microsoft lets bosses spot teams that are dodging Copilot | Hacker News Reader

Microsoft lets bosses spot teams that are dodging Copilot

theregister.com

106 pointsmikece93 comments

For context, this is the whole list of metrics measured by their Eye of Sauron tool, Viva Insights – it seems to be split between Outlook, Teams, Office, and Copilot data. The Outlook and Teams stuff seems to relate to time spent on their individual platforms, but the Copilot-by-app stuff is hilarious.

https://learn.microsoft.com/en-us/viva/insights/advanced/ref...

I look forward to the equivalent of a mouse jiggler.

That tool is quite frankly insane and not allowed in the EU I hope?

It absolutely can be used within the EU.

This is the marketing page targeted at companies in the Netherlands:

https://www.microsoft.com/nl-nl/microsoft-viva/insights?mark...

those metrics are almost certainly a disproportionate invasion of employee privacy, not justified by any legitimate interest of the employer

It's a good time to be a lawyer.

Or a union member.

Well Peter Thiel did refer to Bill Gates as the anti-christ.

pot calling the kettle black moment

He knows, he was in the same coven.

And Peter Thiel is anagram for The Reptile.

That's insane. It's almost all collaboration, chats, and meetings. Getting stuff done seems to be a lower priority.

Depending on the role, I could see these scores being a negative indicator. For example, if you are an engineer, spending your entire day in Teams would be an explanation for lower productivity. But, if you're a manager, a high degree of engagement with Teams might be more appropriate.

I'm not looking forward to the inevitable leaderboard games with these scores.

There's a baked-in but flawed assumption those tools are good for productivity.

I think its the difference between outcome-based and KPI-based management.

One is a lot easier and requires a lot less thought and general understanding of one's own business

The reason I have ignored all attempts from MS to get me to click on something Copilot related, or not click it away, is that I have no idea why I should click. I don't know what it will actually do for me.

I'm using AI for some coding here and there, and I'm using it replacing certain kinds of questions I would previously have googled. So I'm certainly not against using AI at all. (We have a ChatGPT subscription.)

But, as a MS 365 Global Admin for a small business (ca. 50 people), I know what most of the many part of the MS eco system do. I use at least a dozen of their various service sites and tools throughout the week. I know what Entra is and does, Exchange (Admin), Teams, Defender (security website, not the installed tool), I'm not even going to mention the well-known Office apps, Azure, SharePoint, etc.

But when they try to get me to use Copilot, I never know what it is supposed to do for me? It just says "Copilot", and? Yes I could read and check and test what it is about. Maybe I will, some day. For now, every one of the services that I as Unix person had to discover for myself over the last year I had a clear goal before I went there, and I knew why I wanted to use a particular service. That's not true for Copilot, and my general AI background knowledge even as at least daily casual AI user I have no idea what to expect. I'm just not motivated to go there and do all the work of motivating myself myself.

For me, the problem is they market "Copilot" and not a function I want. That is too broad and too vague. If they just used it as part of the regular workflows, like Google did when they introduced AI answers, I would already be using it.

Instead of relying on users desire to use specific functionality, they seem to think users want "AI", what for exactly is secondary. But I want functions first, and don't care how they implement it, as long as it works.

PS: Oh and of course I'm afraid that if I start using any Copilot related stuff I may accidentally add something to our company's monthly bill. Which I only do when I actually need something, e.g. a new Office 365 Business Standard license. I don't want to pay for something I don't feel any need for, since the existing tools are already far more than enough.

This feels like a self-sabotage by Microsoft. There are things that Copilot does well - a business-context-sensitive LLM that works between Outlook, Teams, Word, and Excel is legitimately bigger than any coding tool for most of the business world. However, once Microsoft leadership made it clear that Copilot their top priority, the rest of the company got busy rebranding whatever the hell they were already doing as "Copilot". Instead of selling a tool that fits a need, (employees at) Microsoft are selling a label that checks off a quarterly metric.

Added:

I just checked and it's Microsoft 365 Copilot: 28,10 € per user and month.

That is way, WAYYY too much. The MS 365 Business Standard licenses we usually buy are only 11,70 € per user and month, and they are for what we actually need (Email, Office, OneDrive, basic identity).

They want me to pay almost two and a half times as much again for optional add-on functionality at best???

Whatever Copilot offers, even if it's something super-great, in the end it is optional and not the core of what we want from Microsoft.

Writing a background application to deal with an annoying system so that you can focus on the real work actually sounds like a pretty good thing to have an LLM do, now that you mention it. Maybe Copilot just needs some prompting to keep itself busy, and then they can get back to their real work...

> Redundant meeting hours (lower level) -- Number of meeting hours a person spent in a meeting with both their manager and their skip-level manager present in the meeting.

Food for thought: If Microsoft is already monitoring this, maybe they could help automatically remove the big boss from the meeting?

Not that I really care though.

They’re so desperate to make it work.

> “A source that has seen materials related to sales has confirmed that, as of August 2025, Microsoft has around eight million active licensed users of Microsoft 365 Copilot, amounting to a 1.81% conversion rate across the 440 million Microsoft 365 subscribers.”

Microsoft 365 Copilot's commercial failure - https://news.ycombinator.com/item?id=45476045 - October 2025

At this point its a sunk cost fallacy

Yeah Im just waiting for this to finally play out.

If you understand product design, economics, corporate finance and various other disciplines well - its obvious what is going on.

MSFT wants to juice their numbers for the next earnings call to keep the mania going. Zzzzzzzzzzz

Ive spoken to people who work in the finance sector - portfolio management and tax audit - they laugh hysterically at how bad the tools are (copilot in particular) and resent how much they are being pushed down on them.

I think it's more nuanced than that. Microsoft is desperate for growth in a growth challenged macro, and Nadella has guided valuations and stock targets that are simply unobtainable without this growth. That's one part of the AI slop shovel. The second part is institutions that are desperate to cut labor costs or find other efficiency gains to maintain historical financial performance targets during ZIRP that are simply no longer obtainable. They are willing to push their workers through the sausage machine of "AI" to try to make it happen.

And so the performance art must continue, at least until the AI investment music stops. Your options are get rich if in a position to (due to irrational exuberance and unsophisticated capital investment), or play along until the facade falls to keep your job. "It is what it is."

Yes but the sunk cost is also driving that - the investment has to show growth in cashflows, to the extent it is beating the hurdle rate to add value.

Ahh, Microsoft's AI capital investment sunk cost fallacy, certainly, I agree with that. The stock will be punished when the investment does not show cashflow returns for sure.

MSFT: Microsoft CEO Offloads $75 Million in Stock Amid AI Boom - https://finance.yahoo.com/news/msft-microsoft-ceo-offloads-7... - September 5th, 2025

> Nadella has guided valuations and stock targets that are simply unobtainable without this growth

I read a couple of articles and he seems to have been completely suckered in by the damn thing

from https://archive.is/oWbZB#selection-2387.572-2401.668

> Copilot consumes Nadella’s life outside the office as well. He likes podcasts, but instead of listening to them, he loads transcripts into the Copilot app on his iPhone so he can chat with the voice assistant about the content of an episode in the car on his commute to Redmond. At the office, he relies on Copilot to deliver summaries of messages he receives in Outlook and Teams and toggles among at least 10 custom agents from Copilot Studio. He views them as his AI chiefs of staff, delegating meeting prep, research and other tasks to the bots. “I’m an email typist,” Nadella jokes of his job, noting that Copilot is thankfully very good at triaging his messages.

this is no different to ending up in an alternate reality populated by your AI "girlfriends"

and will produce similar results

But if real, indicates, that he does take it serious. Eat your own dogfood and all that.

I do not see it as AI girlfriend. I see it as pushing the limits of the technology for productivity.

" Copilot is thankfully very good at triaging his messages"

But .. I would never do that, without also having a human check what important bits don't get through.

At some point a mistake will be made that will be very embarrassing for someone that matters

People have "satired" similar hypotheticals about Sam Altman [0], and we know it's happened with people like Blake lemoine.

It makes me curious how much, if any, of the current LLM hype is simply coming from people who are dealing with the ELIZA effect.

[0] https://medium.com/where-thought-bends/the-7-trillion-delusi...

How'd ELIZA do on the International Math Olympiad?

It's quite fortunate that the ELIZA effect isn't defined by IMO scores, isn't it?

How is that point relevant? The ELIZA effect involves humans incorrectly attributing a simple computer program with intelligence. Meanwhile, a complex ML model that solves difficult problems it hasn't seen before is intelligent, whether you or I like it or not.

That part is no longer up for debate. The question is how useful and scalable this particular form of intelligence will turn out to be.

Sounds like a dogfooding sales pitch to me. Nadella has learned from embarrassing prior incidents where Windows Phone developers were found to be using the iPhones or Androids as their personal devices, or when pictures of Microsoft employee offices showed Macs littering the workspace.

> institutions that are desperate to cut labor costs...

I wonder if those who resist copilot aren't driven by the nascent feeling they are training their replacement.

> ... to maintain historical financial performance targets during ZIRP that are simply no longer obtainable.

If all else fails, they can ZIRP back to fantasy land and avoid being the losers by shifting it to the public once again.

Microsoft really wants that AI company valuation.

Our org has legal/security requirements around AI. MS is not making themselves available to address these concerns or giving our admins the information that is being asked for.

There are also ever changing, inscrutable licensing schemes, and too many similarly named products. It's impossible to know what to buy or how much it is going to cost.

This is even before you get to the quality of the product itself. I mean, part of the reason people aren't using it because it is hard to buy and use.

>inscrutable licensing schemes, and too many similarly named products.

This has always been Microsoft's MO. They used to have a team you could call to figure out which product you actually needed.

I seem to remember a similar arc with Cortana. Didn't it briefly appear everywhere in MS365?

My company has both enterprise ChatGPT and Copilot. We originally had CoPilot; we learned it can search for files on OneDrive and Emails and that is the extent of its usefulness. Because everyone was doing work on their personal ChatGPT, the company had to buy the enterprise version of ChatGPT due to risk/compliance issues. Although CoPilot supposedly runs using the same engine, it feels severely limited compared to ChatGPT.

8 million sold licenses doesn't sound like a failure to me. Why do people assume that the market for Office programs and LLM tools is the same or even comparable?

> Why do people assume that the market for Office programs and LLM tools is the same or even comparable?

People ? No, Microsoft

Nothing says "This tool is useful" quite like monitoring and forcing people to use it and then investigating people evading that monitoring.

> It will also be possible to compare an organization's percentage of active Copilot users with other companies. The numbers are calculated using randomized mathematical models, according to Microsoft, so no one company's data is used, and it isn't possible to work out who is in the benchmark against which your own company's Copilot adoption is being measured.

Somehow, I'm confident the result will almost always be that other companies are using it more.

Imagine going to Ignite and talking to your counterparts in a bunch of other companies. Then you realize that you've all got the same percentage of utilization. :)

There are lies, damn lies, and statistics. Let's say you define top 10% of copilot using companies as excellent. By definition 90% of companies are below that benchmark and can use that to justify more AI investment. The subtle part of this is that it is dynamic. As companies use AI more top 10% is going to keep shifting up creating an AI race.

Imagine such a amazing productivity tool, so amazing that you have to force your users into using it. As a person that was just born yesterday, I'm quite sure that the other technologies that are constantly compared to LLMs, like the internet and smartphones certainly must have endured the same adoption barriers, right?

I guess teams that are clever enough to avoid Copilot are also clever enough to write a script to call it on repeat in the background, while they are doing something more useful.

Autonomous agents are all the rage anyway, isn't it?

> guess teams that are clever enough to avoid Copilot are also clever enough to write a script to call it on repeat in the background

Perhaps this and also filling onedrive free space with random noise, to waste MS some storage in addition to CPU/GPU cycles.

While I totally get the sentiment that is quite the environmental pollution, if (more) useless GPU/CPU power is wasted

Didn't Excel just gain a COPILOT() function? You don't even have to be particularly clever to do this.

They can make AI use AI for them, and koolaid bosses should be happy.

Make copilot write it for you. Bonus points for asking it to be a self-replicating worm that runs on kernel level

A couple days ago there was a post here about how most LLMs enter an endless replying loop if you ask them to show you the non-existing seahorse emoji. Asking it to show you one and minimizing the window might be enough.

Leadership - "All employees are required to use AI and be X% more productive"

Employees find that AI tools are useless and don't increase productivity.

Leaders say - You are using the tools wrong. Figure it out.

Employees now have to work longer hours or risk getting fired.

Company does layoffs under the guise of "we replaced workers with AI" and the stock market rewards them for it.

Maybe I could create a grift company that tells other companies why they are using AI wrong.

After working for a few weeks at high hourly rate, produce a report. Then grift the next suites.

> After working for a few weeks at high hourly rate, produce a report. Then grift the next suites.

Bonus points if you just have ChatGPT spit out the report :D

Isn't it what McKinsley has been doing for the last few years?

Consulting companies are wayy ahead of you

Honestly at this point if you’re a dev and not using a coding assistant at all and in any possible way, you are indeed becoming a liability to your company.

Now, Copilot (the assistant, not the GitHub coding one) is hot garbage compared to Claude/ChatGPT but that’s another story.

If you are submitting AI Slop code or overuse AI then you are a liability.

If you don't use AI at all you might be 5 or 10% slower, but quality might well even make up for it.

I would say even more important than quality is just the fact that you will probably understand better. If an issue arises you will already know the intricacies while the AI slop engineer almost has to ask the AI again. They have got themselves stuck in a sort of slop cycle where after some time human understanding is completely out of the picture.

AI does not save anyone from having to know the fundamentals of software engineering and systems, but that’s orthogonal.

If you know them, AI supercharges you. If you don’t, you’re a lost cause no matter what.

> AI does not save anyone from having to know the fundamentals of software engineering and systems

That's kind of my point. There are junior engineers how have to go through a learning phase. Also, you may have this insight, but their managers may not.

The sad thing is management often just doesn't care about those kinds of things.

For instance: my employer seems actively hostile to maintaining human understanding, even before AI. Ownership of apps moved around without sufficient knowledge transition or training. We've migrated Wiki systems 2 or 3 times over my career, and stuff always gets lost. The last migration (to SharePoint) was downright hostile: it was presented as an opportunity to "clean up," the half-ass automated migration deliberately excluded things more than a year old, and your docs got nuked unless someone was paying attention to save them (not a given). Now that SharePoint is in the cloud, its admins are actively scanning for things to delete, because the priority is minimizing their storage costs, not, you know maintaining knowledge of how things work.

On the other hand, many many (usually junior) engineers that have unfiltered access to coding assistants have basically become huge liabilities over night with tools like claude code.

Why? They don't necessarily increase productivity at all [1].

[1] https://news.ycombinator.com/item?id=44526912

I haven't noticed coding assistance improving quality devs in a positive way outside of perhaps occasionally saving on typing speed... and those opportunities are few and far between

I've found it detrimental to quality because people at my workplace are now submitting code changes in languages they don't really understand (like in multithreaded C++ code). The time they saved is taken from my time having to explain all the problems with their code.

That's only true if you're making CRUD software and easily replaceable by any random programmer. For anything more serious LLMs are only useful as a better search engine.

This is crazy. I use copilot completions occasionally, but on average I think it's been productivity neutral so far. Sometimes it helps, but this is roughly offset by fruitless rabbit holes and straight-up wrong information.

One of my co-workers never uses any of it. For certain types of problems, he's the most productive member on the team.

> One of my co-workers never uses any of it. For certain types of problems, he's the most productive member on the team.

Sounds like he needs a PIP!

Last time somebody asked me at work if I was using it, I had just turned off the Visual Studio normal autocompletion because the setting to disable completion on space, enter, dot, etc disappeared.

MS productivity tools are a serious hindrance.

Is that why any useful software was written before 2023?

I notice that you said "using" though and did not specify useful output. Useful output or even just a GitHub repo is kryptonite for "AI" proponents.

If you mean that Microsoft has to pretend that its employees are using "AI" in order to keep the P/E ratio of roughly 40, then of course an employee who does not participate in the con becomes a liability and you are absolutely right!

>> Honestly at this point if you’re a dev and not using a coding assistant at all and in any possible way, you are indeed becoming a liability to your company.

Honestly, you were probably a liability to your company prior to AI. Now you can at least vibe-code. </sarcasm> I don't know anything about your situation, nor do you know mine.

> you are indeed becoming a liability to your company

A risk that stock price declines for not buying into the hype. Actual productivity is not a concern.

> Leaders say - You are using the tools wrong. Figure it out.

> Employees now have to work longer hours or risk getting fired.

Bahahaha. In the best of all possible worlds.

Likely reality: Leaders say you're using the tools wrong, figure it out.

Employees: OK, so we train on the tools after hours?

Leaders: No.

Employees: Before hours?

Leaders: No.

Employees: During lunch?

Leaders: No.

Employees: Then when are we supposed to learn how to use the bloody tools?

Leaders: You're just going to have to figure that out for yourself.

Source: Happened to my mom when they moved from mainframe to Web-based at the insurance company.

This could backfire on Microsoft - if a company's leadership sees low Copilot by their employees, cancelling the subscription is an easy way for the company to save money.

I would tell them if only they'd ask, absolutely no issue here being the control group. Push the issue, well: Copilot can enjoy /dev/urandom

You wouldn't have to do this with a tool that was beneficial.

What are the rumors about the new second CEO of MSFT?

https://technologymagazine.com/news/why-is-microsoft-ceo-sat...

The official narrative is that Nadella wants to focus completely on his misguided "AI" obsession. Could it be that the board forced Nadella to install this second CEO as a backup if Nadella's fantasies fail?

Doing a mindless chore in the background to satisfy a meaningless metric sounds like the perfect task to give to an AI agent…

The good news is that a lot of organization will be able to realize how much their expensive subscription to Microsoft Copilot is useless and so that they can easily cut costs without affecting the productivity.

Viva shows aggregate numbers and does not give any details sooo...

...ask Copilot to play a nice game of clue, checkers, tic tac toe, chess or whatever.

Adoption metric accomplished! :)

(That's what you get when you force people who don't want to use AI to use AI.)

Malicious compliance

Microsoft adds features to its products that customers ask for. If these metrics are being included in Viva, its because companies are asking for these metrics.

The company asking for this was Microsoft. This was originally developed internally because Microsoft wanted to make Copilot usage a key performance metric - they needed a tool to measure that. I've been told as much by people at Microsoft pretty close to this tool. Now I'm speculating, but it seems like someone saw an opportunity to take that internal tool and offer it as an add-on, and pad their bonus by doing it. I suspect this release was driven strictly by supply, not by demand.

Couldn't people just give the AI pointless promts/busywork to look like they're using it?

Copilot got added to our Github repos. For out C++ one, I'd give it a solid 6/10

Claude is better. End of.

LLMs are on a very steep improvement curve, in general. Claude is good today. Something else from some group that figures out the next useful optimization will be better tomorrow. Repeat/rinse for at least 10 years..

A lot has been written about LLMs having reached a plateau with regards to improvements. They still all produce garbage way too often. LLMs have fundamental limitations that can't really be fixed. Garbage in / garbage out also applies, and that is only getting worse with LLMs being trained on ever growing volumes of "AI" slop that is permeating everything lately.

That group will never be Microsoft, and if it is, that model won’t be the one they’re shoehorning into every enterprise product for free.

Yeah. Like fusion power

Reply on news.ycombinator.com