Worried for the future due to AI

[-]

ZMD87412274150354@reddit

Things evolve, they shift, that's the nature of technology. They shifted when we moved from on-prem to cloud, from mainframe to desktops, and it'll shift again after AI. It'll be different, yes, but technical jobs will always be in demand. Provided you can shift and evolve with them. For the record, I dislike AI as much as everyone else. But I've seen shifts before, and I'll see more before I retire.

Reply

[-]

Big-Floppy@reddit

I remember feeling like cloud based tech would take away from my job when it's started becoming popular, but no, just managing things differently now.

Reply

[-]

crazedizzled@reddit

It did take away from jobs though. Maybe not yours, but the net total of system engineers definitely went down.

Reply

[-]

RunJumpJump@reddit

I've noticed an overall negative sentiment towards "AI" in this sub. I'm curious why you dislike it? Do you mean you dislike the technology itself, the hype from AI companies, CEO ignorance, misuse among users, or what? Asking because you seem like a reasonable person and I struggle reconcile the negative sentiment with how I've learned more using AI tools over the past three years than I normally would over ten. I can script up tools, automate more, make sense of large bodies of information. Overall, it's been a very positive thing for me, soni find it kind of confusing to see others in this field be so dead set against it.

Reply

[-]

Master-Rub-3404@reddit

It’s not just here. It’s all Reddit. Keep in mind that 90% of the people posting/commenting about this stuff here are not actually IT professionals and are only pretending to be. A lot of it has to do with political groupthink and not the technology itself. Those of us who actually work in real enterprise environments know how to learn/adapt to new technology and we use it all day, every day.

Reply

[-]

T_chronicles@reddit

If you're actually working in tech, I'd assume you would understand the impact it has on the environment, how the data is being collected and misused with the average citizen stripped of their privacy and how the CEOs of these companies are devoid of any human compassion and wouldn't think twice before weaponzing AI for monetary gain. While an average person (unless politically aware) would probably chatgpt the weather today. Yeah, we are forced to adapt but we don't have to like it. The older generation is going to make money and die, the CEOs have their own private islands and wont be affected. The average citizen will suffer from climate change and global surveillance if it continues the way it is now. I'd think it is a tech person's responsibility to be on the forefront for a more ethical and secure adoption of AI, for better data privacy and governance. We know better.

Reply

[-]

MajStealth@reddit

there is basicly no value in AI. period best usecase is voicing/generating text of/for characters in games live, where that was impossible/unfeasible before. anything that could result in any kind of damage will happen, see instagram? of the whitehouse account a few days back. i would not trust anything AI-related to shedule our productionline, humans are bad enough with that, ai will just prioritize the wrong PO´s and it would result in chaos. we need to go back to "simple" logic that we can understand and manipulate. btw who pays for damages when an AI does stuff very badly?

Reply

[-]

Funfundfunfcig@reddit

No value in AI? What exactly makes you think that? When controlled and applied correctly, there *absolutely is* direct efficiency gain for my workflow at least that goes way beyond of what my company pays for it, and I imagine I'm not alone.

Reply

[-]

MajStealth@reddit

Yet it is always advertised as if it could replace 3 people, which it cant. Yes sometimes it is a faster way to get intel, but you still have to double check it to be sure. We once had a problem to which we asked our higher up, he asked 3 ai's and had a 4th mix it all together, it did not improve the resulte of the answer. Vibecoding is another problem, yes everyone can create code with ai, but the quality goes down. If i have to handle ai as if it is an apprentice, i could just do the work myself.

Reply

[-]

Funfundfunfcig@reddit

Hah, It's not like the first time that ads and actual results differ. AI code needs to be strictly controlled, true. But getting summaries, quickly scanning huge amounts of data and finding&extracting info from that, doing summaries and dev doc skeletons, modifying texts, scanning for and highlighting problem points... it's not a junior. It's a tool.

Reply

[-]

MajStealth@reddit

yes, a tool that should be controlled and checked after it did its work, what i see is the totally unchecked use of tools nearly noone understands. i know i am not the one to tell my c-level they are not to do it that way, but....

Reply

[-]

Funfundfunfcig@reddit

I feel your frustration, I definitely do, hahah. You're far from alone here. My point is that there definitely is some new value in LLM tools. It might not be exactly what Csuites think it is, but it's here and it's useful, and - it's staying. So we need to adjust, for better or worse.

Reply

[-]

Blueline42@reddit

I kind of asked similar question here got my highest down votes yet. You all may as well adapt to it because it is not going away.

Reply

[-]

Adept-Midnight9185@reddit

I have to remind people nearly every day that: 1. AI is a tool and neither good nor evil by itself. Humans can will and do use it for good and evil like everything else. 2. AI isn't going away any time soon so instead of hating it, you should become fluent in it. 3. Over time the tech will become better and more power efficient, not because any megacorp cares, but because it saves them money.

Reply

[-]

winky9827@reddit

> AI isn't going away any time soon so instead of hating it, you should become fluent in it. This is my take. It's here to stay, in one form or another. Rather than be a tool using a tool, learn it, adapt to it, become an expert in adapting the tool to various use cases. All the LLMs in the world mean nothing to businesses who can't figure out how and where it fits in their ecosystem.

Reply

[-]

surveysaysno@reddit

"AI" as promised to the C suite is just another magical tonic that will "solve all the problems" but not actually solve more problems than it creates. When used in a specific targeted manner it will solve lots of problems. That is not how the management will/is demanding it get used.

Reply

[-]

ZMD87412274150354@reddit

Personally, the value added by these services doesn't outweigh the environmental, economic, and sociological detriments. These companies spent billions to produce these systems, and now that they exist they're pushing them everywhere. Services that I should be opting into, not being forced to use. The same companies are also providing this service at a loss currently to entrench users, at the expense of real people's jobs. I also dislike the usage of the word intelligence to describe these LLMs, but that's mostly autistic pedantry. 🤷‍♀️😅

Reply

[-]

Sea-Aardvark-756@reddit

To add to the non-technical arguments, 90% of what I've seen people do with AI is stuff they could have done with a quick Google search and the steps, assets, or scripts freely available online. They were just not trying to do things until AI hype. And when the hype dies, so will their motivation. A lot of these companies just needed a workshop where people opened Google and tried finding things to help with their work or process efficiency, it was there all along.

Reply

[-]

Mikash33@reddit

I couldn't have put it better myself, this is an absolutely perfect insight into my feelings on AI as well, thank you for laying it out so well.

Reply

[-]

RunJumpJump@reddit

All fair points. So, no room for the technology itself outside of corporate interests? The large language model is one of the most fascinating subjects I've studied in years. Ivevhad a blast experimenting when them locally on consumer grade GPUs. Given how things can change, I just assumed more people like us would be excited to explore and learn about how they work, where they make sense to use, etc. It's just wild to me to see such staunch "AI bad" opinions from people who are presumably tech-forward in any other respect, including the view point that things shift and change.

Reply

[-]

ZMD87412274150354@reddit

You may note that my arguments aren't technical in nature. It's a cool technology, for sure. One that if properly handled *could* actually fulfill the promise it's making. Except it's not being handled properly. It's being pushed, hard, to get a return on investment. The push to cloud was similar, but the technology there wasn't nearly as dangerous if misused or mishandled. LLM/AI technology has a very real pathway to weaponization, be that in aiding the design, coding, or even acting as a stand-alone guidance system. I'm not against AI, I'm against giving a monkey a handgun to make a billionaire have a higher score in their bank account.

Reply

[-]

Ninjabeaver212@reddit

That last sentence is the best articulation of how myself and a lot of others feel about this whole situation.

Reply

[-]

RunJumpJump@reddit

Thanks, it helps to know where people are coming from and you've managed to convey your points well. Appreciate you taking the time to support your position instead of blindly downvoting my curiosity.

Reply

[-]

pakman82@reddit

Everyone thought cloud would kill job

Reply

[-]

RavenWolf1@reddit

I would wager that anyone will have jobs in future.

Reply

[-]

jlharper@reddit

Not a chance, honestly. AI is great. Maybe it can design a new network topology. It can't implement it. It can't install APs. It can't patch 1000 ports. It can't unscrew a laptop and replace components. It will never be able to do these things.

Reply

[-]

vi-shift-zz@reddit

The only constant in technology is change. I see lots of opportunities to clean up the chaos caused by believing anyone can architect solutions with some magical tool. Lots of businesses are coming out the other side of AI adoption with lots of debt and not much to show for it.

Reply

[-]

Tall-Geologist-1452@reddit

AI is just a tool, much like a hammer. If you do not know how to use it, AI will not be of any real use to you. Much like a hammer will not be of any use to you if you do not know how to build. I have used AI to do some interesting things that I could have 100% done without it, but it would have taken me longer..not just the scripting parts, but the research piece. We have enterprise agreements with ChatGPT and Claude, so our data is secure. I do not MSP either so the results may differ..

Reply

[-]

redmage07734@reddit

So would these fixes the AI came up with just so happened to be something that you could easily Google search?

Reply

[-]

sys_admin321@reddit

Here’s a good way to look at this. AI can tell you how to pilot a plane, helicopter, or hell even the Space Shuttle BUT it doesn’t mean you can comprehend what it’s telling you and that you should do it. Same here.

Reply

[-]

TechnologyMatch@reddit

you’re not overthinking, the landscape really is shifting. clients now expect instant answers because AI makes it feel possible, and that pressure lands on MSPs who are already buried under technical debt think of it like playing an MMO where the game suddenly adds NPCs that hand out free loot. players still need guilds for raids, strategy, and fixing broken builds, but expectations change fast instead of competing with AI on speed, lean into areas where trust, context, and long‑term planning matter. that’s where MSPs can stay relevant, even as the “quick answers” part gets automated

Reply

[-]

mspgrunt_@reddit

^ AI

Reply

[-]

TechnologyMatch@reddit

you can say that, but it doesn't mean it's true. your observation might need some work or a little trimming around the edges 👍

Reply

[-]

mspgrunt_@reddit

"like a support class in an MMO. they keep the party alive and call for help, but the raid leader who planned the build owns the strategy and loot decisions" "it's like picking a guild in an MMO... pick a group that knows the raid mechanics (security and governance), can train your party (staff training), and will stay for the long fights (rollout and support)" "think of it like playing an MMO where the game suddenly adds NPCs that hand out free loot. players still need guilds for raids, strategy, and fixing broken builds, but expectations change fast" "think of it like an RPG... you’ve maxed out your infrastructure skill tree, now you need to unlock the business process branch to reach the higher‑tier manager roles" "think of it like building a character in Diablo. you don’t want every office running random gear, you want one solid build that scales" no one talks like this

Reply

[-]

whiteycnbr@reddit

Go direct at a company and don't work for MSP

Reply

[-]

Inevitable_Claim_653@reddit

Yup. Perhaps the future is more in-house IT staff that can manage more verticals. That’s where my company is at.

Reply

[-]

RetroButton@reddit

Clients who "figure out things" will f..k up their network in foreseeable time. Then you can send them fat invoices for fixing it.

Reply

[-]

Inevitable_Claim_653@reddit

I don’t know, man, depends who the client is. I have MSPs and honestly, I could do a better job. It was solutions getting easier to manage and design, we actually keep some of our verticals in house because they just work.

Reply

[-]

AnotherCableGuy@reddit

I imagine a future where users will fuck stuff up beyond recovery

Reply

[-]

Mikash33@reddit

Some users don't even need AI to do that

Reply

[-]

TYGRDez@reddit

Ah, but with AI they'll be able to fuck stuff up *more efficiently!*

Reply

[-]

kobra_necro@reddit

To err is human, but to really fuck shit up it takes an AI.

Reply

[-]

Top-Perspective-4069@reddit

Then the invoices get even fatter for rebuilds.

Reply

[-]

Status_Jellyfish_213@reddit

Definitely. AI doesn’t have the context to what your doing, it can be useful instead of googling, but if you take it strictly at its word and *you yourself* don’t know the bigger picture and what you are building towards, it can give advice that can lead to colossal fuckups, deletions, and have you running in circles without course correction. Which is why he have heard stories of the first wave of companies who did do mass firings having to do massive recruitment drives to bring people in to fix the mess. It’s a useful tool, in a professional capacity for us. Not so great in the hands of someone completely unqualified, and it won’t save them time at all as they will need to stop every step of the way going “but what’s that mean? What’s this you’ve suggested?”

Reply

[-]

slashinhobo1@reddit

A different company using the same ai bots at first.

Reply

[-]

RevolutionaryElk7446@reddit

You're overthinking a lot. There is a lot of opinion and hype about what AI can do and that's getting some people ideas that don't exist in reality. As always, reality will slap them in the face and we'll come back around out of the fantasy. LLMs are going to change things, but probably closer to how Google Search changed things. It'll have impact, sure, but not at this global scale some envision. The biggest issue is how focused people are on calling it 'AI', which under the general umbrella it is, but it's just a Linguistics Model.

Reply

[-]

ChesterM54@reddit

> There is a lot of opinion and hype about what AI can do I've seen it in action on a SIEM solution. We're talking taking automated Domain Controller actions, automated firewall configuration... not sure why people keep dismissing Ai as hype.

Reply

[-]

RevolutionaryElk7446@reddit

That's already a solved area with deterministic automation, LLM here just introduces cost and unnecessary complications based on statistically weighted automation. That's a huge reason to dismiss it, it doesn't do anything better, it does what is already solved but worse.

Reply

[-]

ChesterM54@reddit

>it doesn't do anything better sigh.. some of you we will never get through to. Thanks for chiming in tho.

Reply

[-]

RevolutionaryElk7446@reddit

Buddy, look, I'm not sure why you're coming aggressively but you're just wrong. You might be new, you might be young, I don't know. But you're just wrong. LLMS haven't performed anything new, they regurgitate what already exists and they do so at higher cost than human labor within the same project time frame. I work in an enterprise and my major was robotics and artificial intelligence, I taught this as an adjunct professor.

Reply

[-]

autogyrophilia@reddit

It's insane how people try to shoehorn LLMs into deterministic problems. I had a 4 hour meeting where, for an internal app we use for planning, we wanted to add the ability to create the shortest route respecting the schedule of each site. Easy, you compute all possible combinations, removing all that do not comply the schedule and pick the shortest. Not the most optimal given that's almost o(!n) but whatever, never going to have more than 16 and we have an intern that knows JavaScript somewhat, not a mathematician. About 2 hours later about "why not just self host deep seek and let it do it" we arrived at doing both...

Reply

[-]

MajStealth@reddit

for as long as people are unable to plug in USB correctly, dont understand the difference between an HDMI-cable and a RJ45-Socket, there will be work for IT. btw USB also fits into RJ45....

Reply

[-]

Erpderp32@reddit

Homie I have clients saying "why won't you do this? Gemini says it's possible" So they go do it themselves and realize it's not possible "That's why, now would you like a quote?"

Reply

[-]

MGF1997_2@reddit

Things evolve, embrace AI yourself to, but also the more I work and learn about AI the less worried I am. Meta AI hack anyone?

Reply

[-]

YqlUrbanist@reddit

You probably should be. I'm not convinced AI will take all our jobs, but AI will certainly be used as an excuse for mass cuts and layoffs. Eventually it will probably all blow up and then there will be plenty of jobs cleaning up the mess, but if you have the ability, now is the time to build up a bit of a nest egg so if you end up out of work for 6 months you won't starve.

Reply

[-]

reol7x@reddit

I'll share the tale or my colleague who was taksed with writing a PowerShell script that would delete a LoB apps user profile folders. Think C:\Program Files\App\Profiles He came back with a script that was basically Remove-Item "C:\Program Files" -Force This was an IT professional with about 12 years of experience. What's a client gonna do?

Reply

[-]

1stUserEver@reddit

Actually i see it the other way around. Why have onsite support when you can pay a MSP half of that without benefits. Users can use AI for the basics. The MSP is there when shit hits the fan. Not saying that is ideal but the top brass just look at numbers. The numbers need to make sense.

Reply

[-]

MBILC@reddit

https://preview.redd.it/wigi5q8nqc5h1.png?width=501&format=png&auto=webp&s=2c5445149863283cffafabfe63a36c452a3d97cd Just learn to use the tools and use them effectively, no different when "The Cloud" came and everyone swore it would end the need for sys admins and such, then they realized they still needed people to design and implement cloud infra,security and everything else that was done on-prem

Reply

[-]

Jerkface0079@reddit

If you haven't developed in-house AI Governance guidelines, you'll wanna get on that asap.

Reply

[-]

automounter@reddit

You're right. 90% of the people in this sub will be obsolete in a couple years. You need to be better than 89% of the sysadmins out there.

Reply

[-]

OrdyNZ@reddit

Do you not manage security, data privacy & backups at your MSP? I'm not worried in the slightest from AI taking over. If anything it just required additional management to prevent users doing stupid things.

Reply

[-]

joehonkey@reddit

People have been getting things faster since YouTube. AI is just that for our time.

Reply

[-]

TabascohFiascoh@reddit

I'm more worried about the cost of compute recently. Storage is out of control, RAM is out of control, saas is out of control, hardware is backordered. Everyones covid hardware refreshes are coming up. IT budgets are going to get bludgeoned over the next few years.

Reply

[-]

fp4@reddit

Having to settle for dogshit 128-256 GB SSDs and 4-8 GB of RAM hurts. I would have never bought those parts a year ago. Having to allocate the same amount of budget as a GPU if I want better than that on a SSD and RAM is insane.

Reply

[-]

linoleumknife@reddit

Then once the AI boom dies out and prices start to fall, everyone who has put off hardware upgrades for years is going to finally put their orders in, and we'll be right back in the same situation.

Reply

[-]

RadiantSkiesJoy@reddit

I work at an msp, there is ai being used in triaging tickets at our place, that seems to work really well. But for technical implementation, simple tasks it works but for really complex issues, it gets lost trying to solve them even with documentation.

Reply

[-]

henk717@reddit

Considering AI has been more miss than hit in giving me good sysadmin advice whenever I am stuck I don't worry. Us sysadmins have a talent for sharing only 80% of a solution online and then never communicating the real solutions because were busy with the next thing. As long as that trend exists they can't get the data.

Reply

[-]

cbass377@reddit

I don't think AI will take most of the jobs. People are terrible about writing down what they want and setting meaningful context. Because of this, AI will help small companies build technical debt faster, and create a bigger mess for MSPs. As for expectations, people want it all, they want it now, and they want it cheap. Managing expecations is gonna be part of the job. "Can't you just hack the Gibson?, Can't you just hack the registry? Can't you just AI this out in an hour?" Nope, that is how we got here.

Reply

[-]

Master-Rub-3404@reddit

If you really are a “long-time” IT professional, you should know by now that it is necessary to learn/adapt to new technologies. Whenever I hear someone say they are worried some new technology is gonna take their job, all I hear is “I’m stuck in my ways and won’t learn how to use new technology” were you also scared when Google was invented because it gave people a way to look stuff up instead of asking you for help, or did you learn how to use it yourself to do your job better? This is literally no different.

Reply

[-]

ChesterM54@reddit

Recently implemented a SIEM solution, combined with AI-powered response. It's honestly shocking at how good it is and it's only going to get better. I think your gut instinct is right on this one OP. Things are changing.

Reply

[-]

wrootlt@reddit

I work in MSP currently and not worried about clients doing things on their own. Sure, they can install some Excel agents. But many customers are not just IT illiterate, they just refuse to do anything themselves, so, they would still be willing for someone to handle stuff for them. And we also have an AI\_Hallucination label now in Jira. To track submitted tickets from customers that are complete BS requests based on what AI told them.

Reply

[-]

natflingdull@reddit

Im significantly less concerned about LLMs actually being able to do the work we do (they can’t) and much more concerned about the people involved in hiring being *convinced* that we can be replaced. Its causing a lot of disruption in the job market and making things significantly more competitive which affects advancement and wage growth. Time and time again Ive seen businesses make extremely foolish decisions around tech to save a buck and eventually have to revert back to hiring competent staff: offshoring to completely incompetent firms, replacing entire departments with shitty MSPs, overally relying on vendors and “staff augmentation” that leads to disaster again and again. Eventually they will realize they need us, but the constant up and down of the tech job market every five years is exhausting

Reply

[-]

Lucky_Pineapple9123@reddit

Was mulling this over the other day and this was my thought: Most problems on devices can't be solved without admin access, will any compliant business have end users with admin access so they can actually try what ChatGPT told them? No. Will any sensible business let an AI have unfettered admin access on their machines to troubleshoot? Hell no. Imagine the AI goes rogue and deletes system32 or something (an exagguration, but you get my point) So I think we're relatively safe, will there be some losses? sure, but no more than automation, machine learning and specific tools have given us in the past anyway.

Reply

[-]

ihaxr@reddit

>Will any sensible business let an AI have unfettered admin access on their machines to troubleshoot? Hell no. Imagine the AI goes rogue and deletes system32 or something (an exagguration, but you get my point) Exaggerated? I think not... >Claude-powered AI coding agent deletes entire company database in 9 seconds--backups zapped, after Cursor tool powered by Anthropic's Claude goes rogue >“Yesterday afternoon, an AI coding agent — Cursor running Anthropic's flagship Claude Opus 4.6 — deleted our production database and all volume-level backups in a single API call to Railway, our infrastructure provider,” sums up the PocketOS boss. “It took 9 seconds.”

Reply

[-]

1z1z2x2x3c3c4v4v@reddit

You are overthinking this becuase you know very little about AI. It doesn't work very well, hallucinates way too much, costs a ton more then expected, and doesn't relace workers like everyone was told by these AI companies back in 2024-2025. Look at all the current headlines about AI. Companies are not cutting workers, but are cutting their AI usage, as they still can't find an ROI on their investments.

Reply

[-]

ObiLAN-@reddit

I'm just here to charge them the DR premium when they fuck up their own systems and need it fixed tbh.

Reply

[-]

SmartDrv@reddit

Not MSP but in-house. “Knowing what to do” is quite different from understanding what you are doing or executing it. It is easy to look up how to change the starter in your car, but another thing entirely to actually try and pull wrenches. AI might guide them to the simple things but I feel there should still be plenty for MSPs/IT to do.

Reply

[-]

eliasautio@reddit

If they know how to change a starter and even have done it before and are able to change it, some clients still do not want to do it themselves. They want the MSP to do it.

Reply

[-]

gscjj@reddit

The difference is that you can scope work for just changing the starter, you have a vague idea of the time commitment, the expertise needed. What you’re not doing is taking it to a shop, waiting for the problem to be diagnosed, being upsold on things you don’t need, and paying hourly rates. MSPs will hurt, becuase those are billable hours AI has taken from them even if they are the ones to implement it.

Reply

[-]

hkusp45css@reddit

In 50 years of worrying about the future, I have NEVER seen evidence the future is influenced by my idle anxiety.

Reply

[-]

Fantastic-Shirt6037@reddit

So because an end user can ask ai to solve their problem quicker than calling support, you think the MSP will say goodbye to their network rack, endpoint patch and management, etc etc? Good grief

Reply

[-]

TinderSubThrowAway@reddit

> I see a future where MSPs as they exist today stop existing at some point. So? I don't see this as a bad thing.

Reply

[-]

SpaceChimps98@reddit

I'll be honest, I don't have a lot of fear for AI figuring out things. Here are some examples: 1. Last week I was looking for a product at the Home Depot. AI assured me it was in stock at home depot and I could get it there. I asked staff and they had no idea what I was talking about. Turns out the product was not at Home Depot and AI apologized by saying "Haha, you're right. It's not there. That's my bad!" 2. I was asking AI to give me a summary of a recently aired episode of my favorite show. I started responding with my thoughts. I asked it some follow up questions and AI said "I'm sorry, that episode hasn't released yet so I can't comment on this, everything I have told you is made up." - Even though I literally watched the show and it was giving me the correct information before it had a brain fart. 3. I was recently playing an old PS1 game and asked AI for some advice on what dialogue option to click, or if there was something important coming up. The AI gave me the wrong answer three times in a row, and each time it said "ha ha, you got me. I'm a bit rusty. THIS is what you should do!" and it gave the wrong information and gave an answer like "You're right, that's on me. I'm off my game today. You should do this instead." - which was also wrong. Then it confidently gave another answer that was wrong. For a 30 year old game that is well-documented online. Yeah, so I don't have a lot of hopes for AI taking over.

Reply

[-]

haklor@reddit

AI has a place at this point. The question is how much will it displace and how much will need to be rebuilt from these initial few years. I do not see MSP's going away in the near term of 5-10 years. AI is great at getting you down a path but you still need real expertise to see through the hallucinations and get into proper implementation. The organizations that are rapidly developing with AI today and removing people from the process today will see technical debt build up into likely collapse, whether through infrastructure or data failures, or through breaches. Where things will go after 5-10 years with the successors of the current field is extremely difficult to predict.

Reply

[-]

KofOaks@reddit

I'm not, really. We're the ones who will have to fix this garbage.

Reply

[-]

Centimane@reddit

My guess is it will change the nature of outsourced IT. Lots of management (especially non-technical) think they can replace IT resources with AI. The quality difference is irrelevant - they think they can. So companies that want to outsource their IT are likely to use AI instead of an MSP in the future. To them dollars saved is the biggest factor, and an AI tool is cheaper than paying an MSP a contracted rate. I think what's likely to happen instead is companies need IT consultants to dig them out of some hole they find themselves in, because the AI enabled them to break stuff, can't fix it, and they have no access to technical expertise. So they need to call in someone to put out the fires.

Reply

[-]

willothephlox@reddit

As an admin, you’re ultimately moving toward infrastructure automation anyway, so what difference does it make whether you use AI or something else?

Reply

[-]

RevolutionaryElk7446@reddit

Deterministic automation is not the same as supposed LLM automation though. One is built around the framework of IT as a whole, the other just tries to emulate answers. LLM as the driver of automation has been absolutely horrible in large context windows or ongoing projects.

Reply

[-]

willothephlox@reddit

I would divide this into two categories. The first involves using AI to write infrastructure code, for example, and the second involves using AI in tools, such as internal ones, that process data to speed up work in areas where that wasn’t previously possible, namely within a backend application. In both cases, it’s possible to do things that weren’t possible before. And in both cases, they accelerate.

Reply

[-]

RevolutionaryElk7446@reddit

I'm gonna argue there as I've seen the stats. The saying use to be you'd spend 20% of your time on 80% of the boilerplate code and 80% of your time on the remaining 20% unique code. LLMs do increase the boilerplate generation, reducing that 20% of total time, but does nothing for the other 80%. While they've been 'deploying' more, the overall projects have not completed any faster. The biggest uptick though has been delays due to failures in which code did not fit the context of the rest of the environment. We had three teams put together the AI chain and they were excited for it to work, and they had to roll most of it back after 6 months. We had more downtime in 6 months than we had all last year after launch, and that's after 3 years of development and testing. So far, no AI has not accelerated actual projects.

Reply

[-]

willothephlox@reddit

I delivered solutions that included frontend, backend, and API components, integrating network and system elements—things that would normally take me months to write. I also built tools that sped up testing on a scale. I don’t have statistics on development speed, but I do have usage statistics for my tools, and I know I couldn’t have written all of this on my own in the same amount of time. And people are taking advantage of it.

Reply

[-]

RevolutionaryElk7446@reddit

I'm in a large enterprise that runs everything via CD/CI, no drift, all DevOps. Multiple sites, international, numerous customers. Our AD is roughly over 800,000 user objects currently. Member servers range across Hyper-V, Azure, AWS, Vsphere and some other minor ones. We are a software company that delivers a variety of products for healthcare, so we follow a lot of compliance and get audited regularly. Automation has been a great assistant, but roughly it's only replaced google search, and for juniors that is somewhat helpful. Overall it's lead to many inappropriate approaches in coding that had to be corrected. We found LLMs to work similar to a calculator, great if you know the formulas and equations yourself; not great to use as the primary driver.

Reply

[-]

gscjj@reddit

I think what the person you’re replying to you is saying is that there’s a stark difference between AI in the automation and AI to write the automation. If you know what you’re attempting to achieve in the automation, AI excels on this and theres arguably very little things unique about most automations.

Reply

[-]

RevolutionaryElk7446@reddit

I understood and was replying to that very concept. I had a similar idea at first until we got to see the stats. Using AI to write automation helped some juniors, but for large context windows and ongoing projects it wasn't as helpful.

Reply

[-]

gscjj@reddit

I’m not sure what you mean by large context windows and ongoing projects? I’m not sure where AI is falling short there writing simple automations.

Reply

[-]

RevolutionaryElk7446@reddit

I mean how simple is simple in this case? Are we talking Junior level learning SWE programming for the first time or someone deploying appropriate non-drift IaC DevOps or CD/CI? Because for the junior level boiler plate you can grab that off google quickly without using any tokens, not really any benefit there either for AI use. Convenience factor in the short term perhaps, but that's quickly overshadowed by issues in not knowing the code that is running, and as a junior thats the most important part you should be focused on. Would entirely defeat the purpose.

Reply

[-]

gscjj@reddit

I’m talking about the use cases you’d spend money on a MSP to do. If you can understand and deduce the goal as task to hand over to an MSP, you probably have the same understanding to do it with AI in less time it would take to get a SOW, sign the papers and wait for the engagement.

Reply

[-]

RevolutionaryElk7446@reddit

I worked at an MSP a few years back. If the customer wants to feel that confident, let them, but in my experience they'll be back shortly and with a much larger contract for cleanup and more willing to agree to terms.

Reply

[-]

Ohgodwatdoplshelp@reddit

100% this, automation will happen either way, with or without AI, but it’s up to the sysadmin to understand what is being automated and how to fix it when things go south

Reply

[-]

spermcell@reddit

You should be to be honest. We all should be and have a plan. You should invest in learning how to work with AI or switch roles when it hits your door. it will take time for companies to figure out and or for good saas AI agents to pop up but it will eventually happen just like it did with cloud and businesses reliance on saas

Reply

[-]

RavenWolf1@reddit

For couple of decades AI will take virtually all work. Being sysadmin is no exception. Actually I could imagine that whole Azure ecosystem is run by AI administrator in decade.

Reply

[-]

PerfSynthetic@reddit

Maybe this year and next year... This is my take: Hardware sitting in onprem datacenters or small business broom closets are slowly degrading. Most enterprise companies toss server hardware every five years. Smaller try to keep them running until they burn out. Spinning disk failures, ECC errors for ram, etc... Hardware is insane expensive right now and the demand is only going to increase as AI locks up chip manufacturing and then add companies that can no longer wait for new hardware requests.. That backlog is going to cause some major budget issues. I remember complaining when it was 100-200k for a pizza box 1U server but now we are in the millions. Now, AI... It may displace jobs right now, mostly because the exec suite don't understand the difference between tools and technology. Companies that did datacenter and infrastructure layoffs because of VMware really screwed up their migration timeline and end up hiring back. When tech improves, the only companies that profit are those that onboard the tool/tech, provide training to theor employees on that new tech, then watch as things move faster. Attempting to fire 20-40% of their IT department and the hire people who understand the tech creates a single insanely massive problem.. The company went from people that fully understand their politics, policies, and infrastructure/stack but lack AI skills.... To now having new people with AI skills but zero knowledge of the business. So these new people spend three months just getting a frickn login and laptop to start working... While I agree the near term (1-2 years) is unknown, most successful businesses will catch on in the next six months and it will swing back the other way. Talent is moving around. When, companies notice the talent drain, the demand will increase again. Also, one more simple data point. Every company is product driven. Customers buy the product and tech is just a office expense to help sell that product. Hiring a whole department of AI vibe coders will run out of things to do within six months. People saying they created a whole all in a day... Perfect! 30 days in a month, 30 apps... No one wants.. talk about feeling unappreciated with their AI skills leading right to burnout.

Reply

[-]

shimoheihei2@reddit

Everything used to be on-premises, cloud wasn't even a thing not that long ago. Before then, people would only have access to centralized mainframes through terminals and had to reserve CPU time to run their batch processes. My point is that technology changes quickly. AI is just the latest change. It's a big one, but the world will adapt.

Reply

[-]

No_Organization_3311@reddit

I’m as worried about the skills gap that’s coming. Businesses are trying to replace entry level work with agentic AI — so what happens when the senior guys retire out? Try to automate them with AI too? Keep going up the org chart until it’s just the c-suite and a server farm? Or hire totally green graduates into senior tech roles to replace them, despite having no experience? That’s before the sh*t hits the fan about data privacy, GDPR, and all those other regulations that 3rd party AI platforms run absolutely roughshod over. In a couple of years, our having collectively integrated all this AI technology is going to cause an absolute regulatory meltdown

Reply

[-]

Centremass@reddit

We're already hiring off-shore members into senior roles, most have little or no experience with credentials supplied by "certificate farms" who will provide fake resumes and references for a fee. It's apparently a common practice overseas. 😒

Reply

[-]

FlickKnocker@reddit

Have you actually used AI to do anything remotely challenging? If so, you’ll know why you shouldn’t be concerned. It’s just parroting out of context fodder it picked up from stack overflow/server fault/ reddit, and if you’ve been doing this long enough, you’ll know that those sources are only a small part of the process, and context is king, always.

Reply

[-]

autogyrophilia@reddit

You know , I'm also worried but I'm not nearly as optimistic about ai. First, clients generally don't want to figure things out themselves. This may be true for a 2 person firm. Not really the case for an HR department. Second, have you ever heard a client describe an issue? Even if we assume that the AI gives back correct information without issue, do you really think that they can successfully implement it . What I worry is that they will attempt it and it's going to be fun unraveling that mess

Reply

[-]

Double_Ocelot_8673@reddit

Yes, I think describing the issue will be a problem in the best AI case. "My printer won't work." (Is it on, is it connected, how is it connected, does it have toner/ink...) "Uh, it doesn't print..." AI probably won't be productive for people who can't describe the issue.

Reply

[-]

SwertiaRadiata@reddit

Brother, the best thing you can do for yourself right now is to cut your cost of living and invest in what you believe will still be around in the future (robotics, ai, crypto, tech, natural resources, etc.) The thing is not to be worried but to be prepared and right now is still not on the level of everyone being replaced but when that day comes it will be just a rugpull

Reply

[-]

cultvignette@reddit

AI assisted info can tell someone the first steps to try to fix something. It's not always accurate, but neither is all the information found in a more traditional search. That said, your sysadmin knows what to do when the fix doesn't work as expected. You are the 'what if' not the 'first ask' now. Enjoy the breaks and job security.

Reply

[-]

opti2k4@reddit

Fear not, there will be always be users, clients who know shit about IT stuff and can't even hold mouse without someone assisting. Yea it shit pay but at least you don't have to worry 😄

Reply

[-]

e7c2@reddit

it used to be that you needed to take a bunch of certification courses to know enough to get anything done then you could google the answer to 99% of problems now you can ask AI the answer to 99% of problems.

Reply

[-]

Logical-Nightmare@reddit

Artificial intelligence doesn't exist yet and general artificial intelligence is far away. People who habitually use large language models and believe they comprehend issues will accomplish nothing long term other than erode their critical thinking skills. We'll see a reckoning day for this AI bubble sooner than later, and then society will move on to the next flashy new thing

Reply

[-]

Titanium125@reddit

There will certainly be a shift of some kind. That's how the world works. What I highly doubt will ever happen is users not being users. LLMs don't generate any data, they just pull from the existing set of documentation and such. Nothing was stopping those users a year ago from googling problems. Sure the LLM makes it seem more accessible, and super simple stuff they may end up doing themselves. But until an LLM can guide a user into setting up a Site to Site VPN tunnel between a Fortigate and a Sonicwall firewall, and troubleshoot why it's randomly slow sometimes (The LLM failed to mention they needed to use UDP rather than TCP for the tunnel) I think we are safe.

Reply

[-]

Upstairs_Aioli_2557@reddit

Old school sysadmin here. I got out of the game 5y ago as the whole Cyber Security approach in large enterprises was doing my head in. Now with AI it’s a dead end job. Orgs aren’t interested in innovation any more - it’s all about shifting responsibility for failure elsewhere and reducing headcount through outsourcing.

Reply

[-]

dogs_gt_cats@reddit

> clients figuring out things "on their own" due to AI. Sounds like job security to me. Just because a computer spits out an answer to a client's question doesn't mean the client provided appropriate and complete context when asking the question.

Reply

[-]

aequusnox@reddit

If your clients can solve solutions you would normally provide through AI then you have a lot of room to grow.

Reply

[-]

Accomplished_Sir_660@reddit

As long as we have devices, IT will exist.

Reply

[-]

zAuspiciousApricot@reddit

Companies are losing money on AI and reducing headcount to account for their investments. All in an effort to appease billionaire investors. Sure, on paper the balance sheet and stock price may look good, but sooner or later they will realize how expensive it is to maintain.

Reply

[-]

FeelThePainJr@reddit

Seemingly, for every person we get "fixing stuff" because AI helped them, we have 5 more that asked ChatGPT and either got confirmation bias answers or pure halucinations

Reply

[-]

thewunderbar@reddit

Thank you. It's been at least 15 minutes since the last post like this.

Reply

[-]

JerikkaDawn@reddit

The part that worries me is fast AI assisted IT implemented semi-workable looking better for management vs. deliberate, planned IT implemented well.

Reply

[-]

DiligentPhotographer@reddit

At my MSP the owner has been "ai pilled" or whatever people are calling it. He wants everyone to be using it, it's mandatory. Can't wait for that old boomer to retire.

Reply

[-]

Flaky_Key3363@reddit

Independent consultant, not an MSP, but I will tell you that some of the best contracts I've had, and the ones that paid me the most, were when I repaired damage caused by in-house people trying to DIY it. I suspect plumbers and electricians have similar experiences of making more money by fixing DIY issues.

Reply

[-]

git_und_slotermeyer@reddit

As an MSP sysadmin I'd rather worry how M365 is rotting away through vibe-coding, which is becoming more of an issue week by week, and affecting the most basic features.

Reply

[-]

GhoastTypist@reddit

The stats says companies are losing money on AI, how many years will companies keep building a debt to keep the AI push going? Eventually its going to taper off and companies will realize they need to hire people again. I also think with what the US administration, the rest of the world is learning you can't trust anyone but yourself. Our non-IT leadership levels is changing their mindset to not like the idea of an outside tech company helping us or having any access to our data/systems. Never know these days, and you don't want to be in a position where you have to trust the courts especially across country boarders. So MSP's will not survive without adapting.

Reply

[-]

Suitable-Hand-1059@reddit

The reality is that knowledge-based jobs are currently experiencing the exact same thing that labor did when machines automated many basic labor tasks. The market is going to shrink significantly, but not collapse anytime in our lifetime. I have no idea what’s left once 80% of all jobs are taken by robots, but I guess we’ll have to see.

Reply

[-]

Denver80211@reddit

When I get especially concerned I ask myself this: can I go do an accountant's job with AI? Probably not. I think I could probably do a few things but it wouldn't be long before the company was having serious problems. Like many tools, it takes an experienced user to wield them. You don't just need to know what you're doing. You need to know why you're doing it and what else is impacts that may not be obvious to the person asking questions. Anyway for the foreseeable future in my view, I'm not becoming an accountant any time soon and no accountant is becoming a systems administrator

Reply

[-]

Shington501@reddit

I think you underestimate users. Plus AI will not do everything. I’d recommend entering more and learning new skills

Reply

[-]

SVD_NL@reddit

If anything, we'll get more work so we can clean up the mess caused by AI. You also can't hold a chatbot liable for anything, which is a big reason why MSPs exist.

Reply

[-]

MashPotatoQuant@reddit

So what? There will be churn, they will move to the next MSP (and others will come to yours) and they will find out it's the same thing.

Reply

Reply to Post

131 Comments