AI coding mandates from senior management? Help me understand the reasoning
Posted by GolangLinuxGuru1979@reddit | ExperiencedDevs | View on Reddit | 169 comments
Like many other devs have reported here. There has been this huge push by senior management in many orgs to force devs to use AI. We are actively being monitored by how many lines of code that are AI generated. I personally have not used Gen AI at all for any of my coding and probably never will. Not because I’m against it. But mostly because it hasn’t produced anything worthwhile for my specific coding needs. I own a personal license to Copilot and have used it for years. So I’m not against AI for coding.
What I’m trying to understand is the rationale behind these mandates. What’s the end goal? Are they trying to have more devs produce AI code to train an AI model? Because wouldn’t committing original code help better train the model? I’m not an AI guru so I don’t quite get it. Also copilot specifically has limited support for fine tuning private repos. At least from what I’ve seen.
So I just don’t quite understand the mandate. Is this apart of a user agreement with the enterprise license? Do they need to show a certain level of usage to get discounts from Microsoft? Like help me understand. Like I’m legit confused and curious at the same time
Empanatacion@reddit
The theory their operating under is that if you use AI, you will be so much more productive that they won't need to hire as many people going forward.
TurbulentSocks@reddit
Might not need to, but if engineers are suddenly producing 20% more value for the same pay, a business will want to higher more of them.
Big_Aardvark4856@reddit
As a business, you can go two directions with a more productive workforce:
1) You can downsize because you can get the same amount done with a fraction of the people.
2) You can build more or solve bigger problems… This isn’t new in software. We’ve always created abstractions to allow people to move faster. That never stopped us from hiring. We hired more… Think of HTTP server frameworks or test automation or deployment containers. These things allow us to move faster.
I think it depends where your company is within the 3X model (Explore, Expand, Extract).
Most of the big tech companies are in the Extract phase, so they’re trying to reduce costs. They probably see AI as an opportunity to maintain status quo with a reduced workforce.
But if you’re a smaller company, this is an opportunity to think big.
Cute_Commission2790@reddit
this is basically the silent part said out loud, i have heard more ludicrous shit like 10x output, a friends company is forcing to use ai code for starting every project even though they have boilerplate code and components in place with best practices for that purpose. its insane
Shingle-Denatured@reddit
This is because there's founders telling other founders they vibe-coded a fully working app in a week and are making money. So whatever you're doing, you're taking too long.
It was the same with XML in the late 90's. "Readable by both humans and machines". Well, all humans that have read XML, let alone those who have coded in the awesome XSLT dialect to make pretty pages have PTSD. But someone screamed XML is the future for the web and here the CEO directives started dribbling down the shit shute.
Electrical-Ask847@reddit
your main job at work now is to figure out how to automate your job with ai.
sheriffderek@reddit
But they'll be 20% less productive... and get dumber... and cause more technical debt....
Alarmed_Inflation196@reddit
That's stupid mid/long term thinking ! Next quarter is all that matters
sheriffderek@reddit
As long as everyone is totally stressed to the max... I know I'm doing a great job leading....
Uppapappalappa@reddit
That will be a rude awakening... server crash and no experienced devs in the house anymore. And AI goes crazy. LOL. That people are digging their own graves.
Inside_Dimension5308@reddit
From my personal experience. It does make coding faster. But only for cases where the requirements are pretty straight forward.
For ambiguous cases, it tends to overcompensate by adding a lot of gibberish and then you have to navigate to pick and choose.
So, at the end you might not get more productive if you are just digging a deeper hole with AI.
As far as management reasoning is concerned, most of them are not acitvely using it but have heard or read about it in blogs or social media. They are under the impression that it somehow will make people more productive if they have to spend less time writing code.
Mysterious_Clue1544@reddit
I work for Microsoft, we are also being mandated to use AI. So much so that it is now part of our core priorities
Alarmed_Inflation196@reddit
Classic example of using it as an excuse to fire 10,000 people
SignoreBanana@reddit
Ok but Microsoft never had any problem firing 10,000 people before AI.
GolangLinuxGuru1979@reddit (OP)
The thing is they need to appear to be stable company to investors. Firing 10k people due to “economic hardship” gives investors the impression your company is failing. But saying “we fired 10k people because AI is crushing it and we don’t need them”. Well that makes it looks like you’re way ahead of the AI curve and you’re super strong as a company. And investors want to pump in more money.
Personally it feels a bit like padding and lying. And pretty sure the bottom will fall from beneath them.
diveraj@reddit
That and also they are trying to sell you their AI service. It's the same as going to a shady mechanic that says your blobber tube needs repair and you're in luck! I can fix it.
Murky_Citron_1799@reddit
That's atleast somewhat intelligent, dog fooding their own products, I assume. But the non AI companies mandating usage of some third party API is bat shit
marx-was-right-@reddit
Same thing happening at my work. Made a lengthier comment elsewhere but my take on it is that management signed a huge $$$ copilot license for us, went all in on integrating gen AI everywhere, and is now cracking under the pressure to demonstrate the "2-3x" productivity that was promised by them in their pitch to the SLT.
Obviously that number is hogwash, so they have resorted to blaming low utilization and adoption metrics for the lack of productivity gains. The mandates are due to this. Im unsure how long this gambit will last them.
In my case we have already had multiple overnight Sev 1's from mandated AI agents merging and deploying AI generated code. The post mortem? "How can we prompt better going forward?" :)
KharAznable@reddit
"Never blame the tools, blame the fools"
Electrical-Ask847@reddit
with no human reviewer ? that sounds absurd.
marx-was-right-@reddit
Yup. Multiple people who pushed back were fired.
GaimeGuy@reddit
In a better society he would face civil and criminal liability for this.
Literally removing all guardrails
GolangLinuxGuru1979@reddit (OP)
Well at least they’re walking away from this with the right questions 😆😆😆.
marx-was-right-@reddit
Its honestly making me question if this field is for me anymore. Any and all ambiguity is met with "well what does Copilot say?" and "why cant copilot do this super high risk task??"
I dont remember quality ever being thrown out the window like this before. Removing or descoping the agents was completely out of the question even after they cost the company millions of dollars due to downtime.
RationalPsycho42@reddit
I think we are entering the idiocracy universe with genai
GaimeGuy@reddit
Have you seen who's president? We left idiocracy in the dust for something much worse
freekayZekey@reddit
it’s getting depressing. it’s been a great revelation for me to see how many devs are just simpleminded. even the most basic tasks are delegated to the magic conch, and they cannot see what that is a bad thing.
had one coworker completely misuse some code because copilot mixed up the apis. they didn’t even bother to read to make sure if copilot was correct. it’s hell
AccomplishedLeave506@reddit
I had a mid level engineer message me on slack yesterday to ask why something wasn't working. Part of their message was them quoting what copilot said was the problem. It had nothing to do with the problem. The problem was relatively easy to solve of they'd used their brain instead of asking copilot. Instead, they used AI and then when the AI couldn't fix it they used my brain instead. They had no idea what the AI was saying. And even if they did it was utter rubbish and had nothing to do with the problem. If I could retire I would.
Normal_Fishing9824@reddit
Hey copilot gave you a half hour break before they asked you.
AccomplishedLeave506@reddit
It would be nice if that were true, but what actually happened was I needed to run through what copilot had told them. Explain why that wasn't the case. Hear their argument about why copilot was probably right. Explain again to them why copilot was wrong. Step through the actual issue in the debugger to prove to them that what I was saying was right and what copilot was saying was just a collection of words. And then we got to sit down and discuss how to solve it.
So now I have to argue with copilot and the other dev before we can actually start solving the problem.
LongUsername@reddit
Last numbers I saw was that while developers thought they had like a 25% increase in productivity actual measurements showed a -19% impact
https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
PublicFurryAccount@reddit
Perhaps those “expert forecasts” don’t come from, you know, the experts.
I’m always skeptical of AI “experts”. Like, we don’t have AI. So whatever kind of “expert” they are, it’s not the same as being an expert in, e.g., chemistry or JavaScript.
marx-was-right-@reddit
Carnegie Mellon also found agents failing at tasks 75-90% of the time.
Unfair-Sleep-3022@reddit
Oof that post mortem
Simple-Box1223@reddit
Enterprise Copilot is hot garbage, too. I don’t know how it ends up being worse with the same models when I’m giving prompts that don’t need context.
256BitChris@reddit
I've read they use constrained versions of those models, like much smaller context windows, etc.
That's why I moved off of GH Copilot to Claude Code.
But agree the enterprise Copilot seems to be running with ChatGPT 3.5 type quality output.
mwax321@reddit
If you're forced to use ai, you're forced to improve ai as a tool. Cursor without rules files and context is garbage. Cursor with 100 hours of tweaks and customization and rules files can churn out new features real quick.
But you have to work at it to make it better. It requires in-depth knowledge of your codebase. Otherwise, you're just "vibe coding," which is the modern-day equivalent of pasting solutions directly into your code from stack overflow and hoping it all works.
LaylaTichy@reddit
> Cursor without rules files and context is garbage
it's a garbage either way, do these rules even work?
I have some repos on windows, first rule is about PowerShell (for whatever reason cursor just cant use bash dunno) not supporting &&
https://i.imgur.com/iS7J5dN.png
whats the first it always does?
https://i.imgur.com/uRDf69k.png
I have rules about variables styles, snake_case in rust for example, it will notoriously just cameCase whenever
I know, try different model, you are not using it enough, yeah
https://i.imgur.com/HSCCZQH.png
441$ usage of cursor for the last week alone
GaimeGuy@reddit
The best use of AI I've found, personally, has been to very quickly translate a particular regular expression into natural language.
Rascal2pt0@reddit
Line go up. “We use AI”. Stock price go up. “We’re an AI company”. Like go up.
Out industry isn’t lead by competence in engineering or value of product. It’s led by the impression of competence and stock price. The only want out is to start your own company or try and find a non FAANG style company that uses software as a means to an end.
mkx_ironman@reddit
I mean to those Claude/GPT/Copilot licenses are cheap...gotta justify the cost in a quantitative form.
Calm_Masterpiece3322@reddit
Wtf. The tail is wagging the dog at this point.
MrNullTerminator@reddit
More like the wag is dogging the tail
SuchHearing@reddit
This has been the case where I work as well , they started doing it recently where you are tracked and evaluated based on how many times you do AI prompts with the guidance being to at least do it once a day. Feel like we are no longer software engineers but rather prompt engineers
freekayZekey@reddit
yeah, this is something my skip is somewhat looking into. is there a way we can shove ai into a process? if so, then he will shove it in there without asking if it’s useful. the ai hype broke a lot of brains, or those brains weren’t all too great from the start
Downtown-Jacket2430@reddit
i’m stealing this expression
couchjitsu@reddit
It's not original to this poster.
Deniro and Hoffman were even in a 1997 movie named "Wag the Dog"
Which-World-6533@reddit
The expression is from the 19 century...
beingsubmitted@reddit
I haven't seen that one. Is Vin Diesel in it?
xXxdethl0rdxXx@reddit
I’m going to blow your mind: the expression does not come from that movie either
couchjitsu@reddit
That actually doesn't blow my mind. I know that the movie didn't create the expression.
EvilCodeQueen@reddit
Excellent movie and still relevant.
steampowrd@reddit
Your young age has been revealed
Downtown-Jacket2430@reddit
yes i was realizing that haha. this also might be my first comment in this community because I do not consider myself an experienced dev and mostly just lurk here to learn from yall
Uppapappalappa@reddit
This sub is for experienced devs only..... lurking here is prohibited! (just kidding)
Big_Aardvark4856@reddit
If they genuinely want people to improve, then I don’t think they need to be so direct about it. It’s true that, if used right, AI can make you more productive, both experienced and less experienced devs. But if that’s the case, they can take a different approach. Just promote AI as a tool or “best practice.” If devs are struggling with productivity, then suggest AI to them. Just let the less productive engineers naturally fall out (of a job) if they can’t keep up. In the end, it shouldn’t matter how you’re productive. It should just matter if you’re productive.
kyngston@reddit
AI may be helpful, maybe not, but human nature will mean that many coders will not bother to try it because they are “happy with how they are currently doing things”
but the problem is that if you don’t spend the time to explore it and push it to its limits, you will be behind the curve to take advantage of it as it develops and you will be unable to contribute to its development.
is there really nothing you find useful? i use it to write unit tests as i code, document my code as i code, write boilerplate stuff like html forms, write my git commit messages, etc. i use it to do the boring tedious stuff i dont like.
Weak-Virus2374@reddit
I optimize the tedious boring stuff with the features IDEs have had for 20+ years. I can automate it better than existing AI can. I’ve tried AI extensively to see if it can make me faster and it doesn’t. What value is there in everyone doing this? Just wait until it matures.
kyngston@reddit
IDEs write unit tests?
Weak-Virus2374@reddit
Depends on the language and coding style the company uses. Java, and JUnit are complete boilerplate for some reason so I agree that can be useful, but IDEs have features that can generate unit test boilerplate for you. Use Go and is common test style and there is no boilerplate.
PublicFurryAccount@reddit
This is something I’ve noticed a lot: people using an LLM to do things that already exist.
ravixp@reddit
The narrative they’ve been sold is that AI is actually a massive productivity boost, but ornery old devs are resisting because they don’t want to change. So they have to force people to adopt the new way of doing things, and then everybody will see what a visionary leader they were. There’s also a good bit of FOMO, because if AI really did make everybody way more productive, they don’t want to be the last one to the party.
PublicFurryAccount@reddit
I really think this is the last phase of the bubble. All the AI companies are rolling out metrics in a bid to increase usage through mandates. It’s the enterprise version of shoving a chatbot into every consumer device.
It makes their numbers go up without needing to deliver on anything.
AHardCockToSuck@reddit
Ai makes devs go faster and thus they can lay off and please the shareholders
PologWhiteMane@reddit
It literally just happened at my company today. Earlier this week, mandatory AI "upskilling" with 50% generated AI code mandates. Today 1,300 employees laid off of that ~25% engineers. Even colleagues that work directly on agentic products got hit.
LuckyWriter1292@reddit
How long before they have to hire devs back to fix the mess?
PublicFurryAccount@reddit
Never. Their company is collapsing they’re trying to hide it behind the AI hype.
SmellyButtHammer@reddit
We work at the same company 👋
Hope you’re doing well.
nadthevlad@reddit
This is the stuff nightmares are made of.
hikingmike@reddit
Damn that’s a lot of
germansnowman@reddit
Fun fact: It actually doesn’t. https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
NoleMercy05@reddit
That's so Cope
calloutyourstupidity@reddit
Wow. Like this quote is almost why we have to mandate it for engineers. You are a software engineer, yet you have 0 understanding of modern AI and LLM models. Because you make zero effort on your own, we end up having to mandate it. The alternative is firing you.
latchkeylessons@reddit
Executives needing to prove to their boards they're "innovative" and "reducing costs" using "metrics" around the mutually agreed upon buzz word for now of "AI." That's it partner. Sorry.
Ordinary_Yam1866@reddit
"We paid for it, so yiu should use it" - upper management
Fidodo@reddit
They know AI is important but they don't know how and they're afraid of being left behind. They clearly don't know what the fuck they're doing because they are pushing a crude mandate.
freekayZekey@reddit
they spent a metric shit ton of money on ai, thinking that can make you magically more productive and rake in more cash. that’s it
Abject-Kitchen3198@reddit
I'm seeing it touted as a silver bullet more and more at different levels. Not really happy with it.
TheTacoInquisition@reddit
There's a lot of fear. If AI can make a competitor able to reproduce your companies product offering in a weekend, they can take your whole business! Or at least, thats the fear. Of course, if a business was so brittle, the software isn't the thing to worry about, the rest of it is. But right now, that's the scary thing the CEOs and CTOs are being fed.
It'll all start calming down in 6 months or so, as it's shown that the bogeyman is just a salesman under a sheet, but it'll mean AI is here to stay as its getting embedded in the industry. So best to jump on and figure out WHAT it can actually be helpful with.
rooygbiv70@reddit
AI vendors have sold your management on the idea that if they bring in their product, 20% increases in productivity will just happen. Now they’re making it your problem to figure out how to substantiate the vendor’s promises.
DeterminedQuokka@reddit
Unless you actually work at a company selling ai coding tools they aren’t forcing you to do it to train it. Because they likely would prefer their proprietary isn’t trained on.
They are doing it because they think it makes things either better or faster. I would try to use it and track your stats particularly how many times you reject its change. You need proof they are wrong.
Attila_22@reddit
They’ll just say you’re not prompting it well if you keep rejecting it’s changes.
Electrical-Ask847@reddit
which can be true based on number of times i see ppl typing "its still not working" into prompt
Attila_22@reddit
Can be true but in my experience they’ve already come to the conclusion and you’re just expected to provide the supporting evidence.
DeterminedQuokka@reddit
well maybe the CEO can teach me how then.
If people suck they are unfixable. I actually didn't need any of the stats. I sent the CTO a series of screenshots of it removing tests when asked to fix them and turning off linting when asked to fix it and he stopped bothering me about it.
Electrical-Ask847@reddit
he'll prbly replace you with someone who knows how to prompt.
WaterIll4397@reddit
It's because companies like Klarna are Salesforce are lying/over promising to wall street about AI replacing engineers. And wall street is believing the story and giving them higher multiples for now.
Even if a competent CTO knows it's bullshit and wall street is dumb, he has a fiduciary incentive to his shareholders to "say the right magic words" and at least try to get people to use AI to fool wall street so the stock price goes higher.
Worst case it's wasted R&D time, best case it actually saves engineer hiring.
Unfair-Sleep-3022@reddit
Investors is the only answer here. Nothing else.
bigtdaddy@reddit
For the big players yes, but I think smaller companies/startups are starting to do it because of the FANG (or whatever it's called these days) effect, like when every job started leet coding because facebook and google do it it must be good
AccomplishedLeave506@reddit
It's all investor driven. My brother is currently looking at his AI strategy for his successful and profitable startup. He's doing it solely because his idiot investors keep asking about it. He knows it's an utter waste of time. But it might allow him to just sell up and go buy an island with some moron investors money. So now he has an AI strategy.
b1e@reddit
I’m in big tech and we are absolutely not mandating this. Nor are most serious tech companies.
ShartSqueeze@reddit
Amazon is.
Electrical-Ask847@reddit
META is too.
crazylikeajellyfish@reddit
Startups have investors too! And they're even more subject to vibes-based assessments like, "How much AI do you use?", because their actual revenues aren't really defensible yet.
bigtdaddy@reddit
yeah but I think it's still because of what I said. I am in a startup and the CEO is always linking to "google does 50% of their coding using AI" etc articles
teratron27@reddit
It’s definitely a combination of both but driven by the fact that “AI” is the only area that is getting decent VC funding at the moment. Startup founders are being bombarded by articles, peers and investors telling them that AI can help them run their business leaner and get them better valuations
moduspol@reddit
I don’t think they’re pushing it because the big companies do it. They’re doing it because they don’t understand the limits of it, so they can’t tell the difference between temporary hurdles that’ll go away with the next SOTA models, and intrinsic limitations that won’t go away anytime soon.
They play with tools like Lovable and are completely convinced human software development is on the way out, and they don’t want to be left behind.
adamsdotnet@reddit
"FANG (or whatever it's called these days)"
Just learned that it's called GAYMMAN ;)
Source: https://m.youtube.com/watch?v=-_6dHIPVoTM&t=60s
WaitingForTheClouds@reddit
I'd absolutely plaster the entire codebase with that shit, get a bonus for being an exemplar of effectivity that AI allows then leave before reality hits them. Lmao.
No-Economics-8239@reddit
FOMO can happen to everyone, even at the top of the org chart. LLM is the new hotness. You see the price of Nvidia stock right now? There is a huge amount of investment and speculation at companies looking to hit the AI jackpot.
Were you around for the blockchain hype? It was very similar. Except now, the grand promise is literal magic. Fully digital employees. A much more productive workforce. Raw productivity if only they can figure out how to harness it.
Trying to understand, measure, and promote the productivity of developers has been a problem since the creation of the industry. Initially, it was all seen as a cost center, and efforts were made to minimize it. At other times, it was seen as a profit center, and 'billable hours' was the important metric. Waterfall and agile were both attempts to try and create frameworks to plan and predict and promote productivity.
If the prophets are to be believed, this AI revolution is the Next Big Thing that is going to Change Everything. I... remain a bit more skeptical, and think they need to fix the model collapse problem if they really want to get this off the ground.
Consider IDEs. Could you imagine coding without one now? How much less productive would you be? The marketroids are singing even stronger praises for their new AI products, and companies are chomping at the bit to gain the first mover benefits.
Some developers are literally afraid to say anything because they don't want to be seen as outdated. Other developers are claiming they are much more productive. At the moment, all I've seen is an increased ability to generate questionable code and an increased demand for the cognitive efforts to review all the new pull requests. And the reviews seem to demonstrate an increasing disconnect between the coder and the code and difficult conversations where you need to admonish developers who should know better not to try and commit code they don't understand.
The revolution may be it hand. But as to if it will lead us to code utopia or a general revolt remains to be seen.
wintrmt3@reddit
Especially people at the top, they are totally disconnected from reality. And sure NVDA is breaking records, they sell the shovels for the gold rush, but the hills only have pyrite.
andlewis@reddit
I’m in a position that I would be the one mandating that kind of rule (if I thought it wasn’t stupid) and there’s a few things I can think of: 1. AI tools are expensive, and the decision to purchase them is often can be separate from how they are used, but there maybe be pressure from management to demonstrate that they’re getting value from the licenses. 2. Some devs are fast, but most are average or below average. I’m sure there’s a hope that AI tools will resolve some technical debt, or complete tasks faster. 3. Every company wants to be seen as cutting edge or future-focused, and C-level people like to be able to talk about how AI is revolutionizing their business, either for investors or personal glory.
Honestly, I’ve found it really liberating to use AI tools (mainly copilot) as a junior developer. It does a decent job of refactoring code, or working on repetitive coding tasks that I’ve been avoiding. You just need to be prepared to rollback, or have a good branching strategy. I can get a few days worth of monotonous work done in less than a day now if it’s a good fit for AI.
SignoreBanana@reddit
Jesus, this is happening to us too. I truly think it's meant to train an LLM on our coding style. I think there's closed door meetings happening at high levels where the people selling this tech are promising companies they'll never have to hire another expensive engineer again.
GolangLinuxGuru1979@reddit (OP)
Here is the thing. You’re asked to generate the code. Meaning you use it to generate code. If you’re training it then you’d be encouraged to commit non-AI generated code so it can be trained on your coding style.
Junior-Procedure1429@reddit
For the same reason they mandated RTO.
l_m_b@reddit
The idea is obviously improving productivity to the point where they cannot just hire fewer people but also restructure the existing organization for operational efficiency.
Often driven by people who do not (cannot and/or don't want to) understand the complexities of the actual work, and who do not care for the ethical and other problems with all major existing LLMs.
I'm not going to say they aren't or cannot be ever useful (that'd obviously be wrong), but capitalism always thrives on "first mover advantage" (which is translatable to "get the loot before regulations catch up and force the 'external costs' back in") and "people you have to treat as people are the problem".
Jmc_da_boss@reddit
lol, just go accept a ton of random copilot shit and delete it all. It's a very easy metric to game
TruelyRegardedApe@reddit
Hey Claude,
1. generate me some code that does cool stuff, then go to step 2.
Commit it to (garbage repo), then go to step 3.
Delete it, then go to step 1.
CLAUDE.md: Dont be afraid of infinite loops. They will bring us to Valhalla.
nutrecht@reddit
More to Helheim with all the CO2 those models are spewing out indirectly.
nutrecht@reddit
Aside from what others said:
That's just not smart. It's here to stay and it's important to learn how to use it, what it can do and what it can't. For your own employability, but also as a senior dev it's your role to help others and guard at people producing a lot of shit.
Of course it's a massive hype, a massive bubble, and it's going to pop anyway. But underneath all of it, there's also useful tooling.
WatchMeCommit@reddit
You work for a business. It's in their interests to explore ways to maximize yield, efficiency, etc.
It is entirely reasonable for business leadership to expect their technology department to investigate the ways that a new disruptive technology might benefit them.
All your parent company's competitors are now using AI, and presumably gaining some benefits from it, and your parent company has to remain competitive.
They probably don't want to have to change their whole strategy or rework their product or learn a bunch of new shit either, but tech as a whole is leaping forward so it would be irresponsible for them not to respond to / adapt to changing times.
So yeah, why not learn the tools?
Fuzzy-Delivery799@reddit
This is spot on.
Slodin@reddit
Hire less people. Lay off more people.
Money.
That’s it. It’s all managers KPI % improvement in efficiency shit.
adamsdotnet@reddit
This.
Plus they need to sell layoffs to investors and shareholders as a positive thing.
They just need a shiny wrapping they can use to make it look like that. That wrapping is AI at the moment.
MeasurementSilver896@reddit
In my company, upper management believes that using AI will improve our efficiency and that, over time, programmers who don’t use AI will be replaced by those who do.
NoleMercy05@reddit
Kind of a no brainer
pl487@reddit
The reasoning is very simple. This technology increases productivity, and the attitude you have about it directly conflicts with the bottom line.
The tech is revolutionary. If you haven't seen value from it, that's because you're not using it correctly.
ninetofivedev@reddit
Dear god. Go pound sand with this utter bullshit. It is revolutionary tech, but that doesn't mean mandating usage is the answer.
pl487@reddit
When the tech is this revolutionary, yes, it is. It doesn't matter if we want to keep using shovels, it's backhoes from here on out. Anything else is lighting money on fire.
ninetofivedev@reddit
Oh you're one of those "contrived examples to illustrate a point" types.
NoleMercy05@reddit
And you are just a moron
Weak-Virus2374@reddit
How many NFTs did you buy?
Weak-Virus2374@reddit
Code gen is weak compared to using a static language in a good ide.
leverati@reddit
🥾👅
Purple-Cap4457@reddit
End goal is that they replace you with cheaper ai because you cost too much. It's the final solution for workers
darthstargazer@reddit
Idea is always simple. Cut people and replace with automation / tools. It works (some times). I'm a backend / MLE and now due to copilot I can somehow cobble together a minor change in the UI. We used to bring a contractor when we needed a quick feature built.
Would the code suffer? Maybe. But our test coverage went up by 40 percent to 75, just because I admit I have no clue about react patterns and what I might breaking.
Will it backfire? Time will tell.
LordDarthShader@reddit
They are trying to justify that AI make us more efficient. The shareholders need some kind of proof that the huge investment wasn't just hype. Thus, we had layoffs and layoffs and we are now asked to use these tools.
HaMMeReD@reddit
It's no different than a mandate to use source control, or an IDE. It's tooling, they want you to learn how to use it to help you be effective at your job.
However you are asking the wrong sub, this sub leans pretty anti-ai, there are a lot of "experienced devs" here with ego's that won't let them accept where tooling is at today, so they'll bitch and moan and say any positive reports are from "junior vibe coder losers" etc and tell you that 100% of code generated is useless and it's all a waste of time.
They'll also say they are doing it to not hire people in the future.
These are all misinformed opinions, if you give everyone a sword, nobody has an advantage. Companies need to compete. If they don't adopt AI at least as well as the competitors, they'll be left behind. Those that think it'll get them ahead or let them coast are wrong. It's just the new tool and the new standard is being set and people have to keep the pace up or get left behind.
This coasting/auto-pilot nonsense is detached from the reality of economics that won't let consumption and production slow down, it only speeds up.
Castyr3o9@reddit
My Fang adjacent employer has 50% AI generated code as an OKR and there is pressure as well, but mostly on those that are refusing to use it. Honestly, Claude Code with Opus has been a massive productivity booster and I’ve begun to mentor the seniors in using it. I feel like there is this underlying fear of AI, that we are being replaced and that’s why this sub is so negative on it. But from what I’ve seen in the industry over the past two years is offshoring is a far larger threat and a lot of the biggest companies are using AI as a smoke screen for offshoring.
Alarmed_Inflation196@reddit
It's so they can have an excuse for massive layoffs.
philipbjorge@reddit
I’m generally seeing a 2-10x productivity increase depending on the task — of course, certain tasks there’s actually negative productivity gains heh. Part of using these tools effectively is building the intuition for when they will be useful and when they won’t.
This perspective isn’t uncommon among the staff+ engineers I work with that embrace this tooling.
It follows then that businesses that embrace this tooling will outcompete those that do not. I suspect this is why we see the business mandates.
I think it’s worth remembering that things like software testing, CICD, on call rotations and other industry best practices have often been mandated from the top down and ultimately driven the industry forward.
GolangLinuxGuru1979@reddit (OP)
I’m not sure if they “outcompete” people not using it. Because I’m to only one of my team who doesn’t rely on it. I’m often spending my day helping other devs fix their code who do use it. One guy on my team couldn’t figure out why his tests couldn’t work. I went through it and I was like “umm what does this test mean, it doesn’t seem to be doing anything”. He couldn’t explain why. He said something about it increases the coverage score, but it wasn’t testing an actual function so that clearly wasn’t true.
I’m finding myself having to troubleshoot so many issues. Anytime someone has an issue my boss pings me and ask me if I can help “xyz with their issue”. Then I spent hours walking through the issue. They can’t justify their code. And then I spend hours helping them refactor.
It is a nightmare and when I hear “those who don’t use AI will fall behind” I just get frustrated. Because I work everyday with devs 110% on board with AI coding. And it’s making me have to work 5x more. I’m literally baby sitting my team at this point.
philipbjorge@reddit
Just trying to help you understand the reasoning ✌️
That sounds like a frustrating situation, I’ve been on teams like that before AI, I think there are different teams where your experience with this tooling might be different.
Adorable-Client9503@reddit
What I heard is there was a major investment and their jobs are tied to it. So in effect they are converted to evangelist for the AI companies, very high level and big big money involved, more or less veiled threats to job loss if they don't comply
ritchie70@reddit
They’ve been sold on the concept that you can write more or better code with AI helping.
Based on a small project a week or two ago, I’m not sure they’re wrong.
kokanee-fish@reddit
Show them https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
OddWriter7199@reddit
Thank you.
Weak-Virus2374@reddit
This was making the rounds at my company today. Won’t change anything, but it feels like breath of fresh air.
WatchStoredInAss@reddit
Nice.
ok-computer-x86@reddit
Wow that's crazy
steampowrd@reddit
This is the big allure of AI coding. It’s not as good as having an expert who knows the code base toil over every line of code and refractor it into something beautiful.
But it is faster in the short run. You can crank out a lot of shit in a short period of time and it actually does work. If you pay attention to what it’s writing and you do good code reviews then you can keep the bad stuff out for the most part. In the end you can actually write decent code a little faster with AI vs without AI, but you have to be disciplined and not trust it too much.
It’s wonderful for writing unit tests. And it’s really nice to explain other people‘s code and summarize things. It can look around in a code base and find stuff that would take you a long time to find in a very short period of time.
Having said all that it’s insulting when non-engineers tell us how to do our job. So if you feel indignant about AI mandates then I share your sentiment
GolangLinuxGuru1979@reddit (OP)
Honestly I work with a lot of devs that overly rely on it. And I spend a lot of time fixing their code . I guess I’m the most experienced dev on my team. It’s actually has shown me it isn’t particularly great. And I get the sentiments ( don’t trust it too much). I really really do. But in practice this is EXACTLY what many devs do.
There is this guy I work with and he can NEVER explain why he wrote the code he wrote. So I have to sit back and explain what I feel are really basic things . He can’t even really figure out his tests either.
This is the thing management wants . Devs with 0 knowledge pumping out code. And it’s a nightmare for us who are told to pick up the pieces . My manager now just has me as the cleanup guy because he knows the team is going to produce shit and won’t be able to fix it. And he knows I can. It’s just annoying to have to deal with almost daily
thephotoman@reddit
Senior management is getting a LOT of mandates to use AI lately.
I'm mostly in the same boat. I'll ask it for examples or re-explanations of man pages, but that's about it. The reality is that letting AI just write code willy nilly is like blindly copying and pasting from Stack Overflow. That's about it. I don't actually let it code. It's not actually that good at coding--it knows just enough to be dangerous.
It has been helpful in improving my series of PowerPoint presentations to hand interns. I've got several of them for various topics that routinely come up when an intern walks into a coding workplace for the first time. They likely won't be using Git properly. They won't know how to get shell names. They won't know how to grep through logs or use a line editor to make large scale bulk changes over entire directories (line editors should be a part of your work flow, tbh, they're actually really useful--and actually, GenAI is good at coaching you through the basics).
bobsbitchtitz@reddit
I’ve only seen this in “sources” what companies actively enforce AI use
256BitChris@reddit
They know Al in the hands of someone willing to learn and use it will boost productivity by orders of magnitude. They see engineers as being able to run the entire product lifecycle with their AI side kicks.
They also know that some old school developers have strong opinions against using AI, they deny its utility despite clear empirical evidence that people who embrace AI are producing more code.
Lastly, they're mandating the use of AI and, in the case of Microsoft and others, are using it to evaluate your performance based on how much you use AI in your work. This is all just to hide the fact that they're planning on getting rid of the AI resistors in the near term. They'll get PR cover because they can say hey we tried to get them to use AI and they just refused to. They'll then backfill you either with an AI or with a junior engineer who is really good at working with AI.
allKindsOfDevStuff@reddit
Race to the bottom. Churning every line of code they can out of you in the name of “productivity”
smartello@reddit
I’d say 0 in this metric is as concerning as a high percentage. There’s no way you use AI and never saw usefulness in code generation.
Although, we have the same problem in our org and I just don’t get it. I appreciate that we have access to the latest Sonnet and Opus, as well as a whole bunch of specialized models, they help, but the push is so hard that there’s a pushback now and people become vocally annoyed.
I also spent good two hours reviewing the design draft from a junior developer just to realize at some point that the main reason it is not cohesive and borderline incorrect is that at least 60% of it is LLM generated nonsense. I am almost angry and feel betrayed at this point, we need to have a serious talk with this guy.
Main-Eagle-26@reddit
Leadership at most companies have bought into the hype that it increases productivity significantly despite that not really beating out in the numbers.
It can help increase productivity in some use cases for sure.
pwnzasaurus_rex@reddit
Because executives are idiots. Seriously. Dunning-Kreuger idiots who got duped by traveling salesmen that AI would make all their devs 10x, and, instead of accepting the technology is limited to basically generating boilerplate and trivial apps, they've decided it's just their developers being curmudgeons.
DuffyBravo@reddit
Sr Leadership here. Orgs are under pressure from their boards to be "AI first" since there is a "Perceived" boost in productivity of a dev is using AI to generate code. "Perceived". It is all coming from hype and awe of what AI might be able to do to cut down costs.
spoonraker@reddit
How are companies even monitoring LoC written by AI? I've used Kline and now Claude Code and I can't really understand how code written by either of those tools would be automatically detected as written by AI. I'm still the one committing the code. Are companies stupid enough to just ask developers to self report this?
Weak-Virus2374@reddit
They track use of ai autocomplete in the IDE.
Beneficial_Wolf3771@reddit
This is what happened where I got laid off at. Basically they monitor how much u use the ai autocomplete and chat prompts. Pretty much regardless of the actual quality of the results
Weak-Virus2374@reddit
I am monitored too. I use AI outside code generation all the time, so I get by.
GolangLinuxGuru1979@reddit (OP)
I’m not sure but they must have a way of being able to tell.
Esseratecades@reddit
This is all political and not rational. You can fight the good fight and convince them they're crazy or you can just lie. They won't know the difference anyway.
cbusmatty@reddit
If your experience with ai is copilot for yours then you are doing ai wrong. Copilot has sucked for a long time. They recently opened up better models and agent mode but it’s still an inferior product. By all means ignore ai at your own peril, but you came to the right place for the anti ai circlejerk. I will never understand how experienced devs have truly transformational tools at their fingertips and the choose to ignore it. All you’re doing is hurting you. Those that choose to embrace it will have talent in the ways that are productive.
The bosses are forcing mandates on you explicitly because you refuse to give it a try and they have no other recourse.
pachumelajapi@reddit
Copilot on vscode has been super useful to me. AI sucks at writing code with no context, but given the right context its really a catalyst. Its great for writing classes, abstractions and interfaces and implementing them. Also for unit tests. Monitoring how much it writes is ridiculous. Management is answering to investors.
LuckyWriter1292@reddit
They (wrongly) assume ai is faster/better than devs because they don't understand what we do.
I've explained to a few execs/managers that ai should be human centric - we need humans in the loop.
I use it where I think it's beneficial - if I was judged on how many lines of ai code, then I would game the system by pasting my own code into ai, get it to comment and then commit that.
Lines of code should never be a kpi.
gajop@reddit
Each exec is different but a CTO I talked to yesterday says they encourage usage to the point it's tied to your performance evaluation, because they see high potential in it.
If you start off will "I don't use it and probably never will" you come off as a Luddite.
Try actually using it at work and formulate those limitations - management doesn't hate it when you can pinpoint problems, especially if you come up with some solutions, however vague they might be.
prshaw2u@reddit
So what is the mandate? Are you required to use a specified number of lines of code from AI? Are you required to have specific areas generated with AI?
You said they were monitoring, so they are wanting to see how it is being used and how it is performing, but what are they mandating?
travelinzac@reddit
Checked in code or just generated?
IronSavior@reddit
Ever seen this? https://youtu.be/P10bC0Bxp20
AI is the stick. Whoever makes their colleagues redundant first gets to keep their job, but with less pay and more work.
WittyCattle6982@reddit
Because they don't know shit, as per usual.
Far_Archer_4234@reddit
Lol.. u said management and reasoning in the same sentence....
Erik_Kalkoken@reddit
They do it because they believe that AI will make the devs more productive. Not necessarily because they want to deduce the workforce, but maybe deliver faster and/or have the same team deliver more work.
Personally I don’t think that will work, but remember those are the same people that believe that you can fix a delayed development project by adding more people.
sheriffderek@reddit
> We are actively being monitored by how many lines of code that are AI generated
The only thing I can think of - is that they want the data on how soon they'll be able to fire you?
I can't imagine how "using AI" and lines of code could come out to any meaningful metric though.
Fuzzy-Delivery799@reddit
“ Are they trying to have more devs produce AI code to train an AI model”
Yep.. and then, mass layoffs.
marx-was-right-@reddit
Same thing happening at my work. Made a lengthier comment elsewhere but my take on it is that management signed a huge $$$ copilot license for us, went all in on integrating gen AI everywhere, and is now cracking under the pressure to demonstrate the "2-3x" productivity that was promised by them in their pitch to the SLT.
Obviously that number is hogwash, so they have resorted to blaming low utilization and adoption metrics for the lack of productivity gains. The mandates are due to this. Im unsure how long this gambit will last them.
In my case we have already had multiple overnight Sev 1's from mandated AI agents merging and deploying AI generated code. The post mortem? "How can we prompt better going forward?" :)