Majority of CEOs report zero payoff from AI splurge

More than half of CEOs report seeing neither increased revenue nor decreased costs from AI, despite massive investments in the technology, according to a PwC survey of 4,454 business leaders. The findings pour more cold water on the hyperbole surrounding AI and the benefits it supposedly brings to business, although the report …

[Eye Know]

GOOD

There is a big slowdown in interest from our clients, Microsoft pissed off a lot of people with their AI bloat in M365 forced on them like a cable TV company making you take a "bouquet" of channels.

[Anonymous Coward]

Re: GOOD

Someone from IT support who was looking at something on my PC asked why when typing in Outlook there wasn’t text being suggested. I said I’d turned it off and he asked why, so I said I found it bloody annoying and that it was insanely distracting when typing. He pointed out that we were paying for AI, to which I replied then you’re wasting the company’s money in my case. He mentioned other features I was missing out on and I said “No I’m not missing them at all thank you.”

One thing a colleague said he did find AI useful for was answering questions at the end of mandatory training. He said he’d not pay attention to the training and then when the question set was presented at the end he would just ask AI for the answer to each question. It wasn’t always 100% correct but was enough to obtain a pass and complete the training. When I asked what Compliance would think of that he said he wasn’t bothered what they thought.

[Anonymous Coward]

Re: GOOD

Just done a load of training. Wish I thought of AI for that

[Michael Hoffmann]

Re: GOOD

Damn, that's brilliant (bet an AI did NOT come up with that typically human kind of ingenuity).

Except ours are all videos (or worse: some are so janky they look like they're Shockwave or something else from the Dark Ages). Followed by multiple choice, including the hated "click all that apply", where you can't even simply guess the one that makes the most sense or seems the least wrong.

[M.V. Lipvig]

Re: GOOD

Ours is the same, but it also tells you the right answer and lets you take the test twice. Write the correct answers as you go, take again for 100 percent. But they are sneaky about preventing cheating. They'll ask the exact same questions... in a slightly different order!

[JimboSmith]

Re: GOOD

Yep the same thing happens at my place except they don’t let you know the correct answer. A colleague who had taken one test four times and failed, attepted to bribe me to do it for them. I weren’t happy to just give them the answers and instead told them I would teach them. After that they passed first time, and the year later when it reared its ugly head again. She said I should be doing the training not a computer program.

[Sir Sham Cad]

Re: GOOD

Slightly over a decade ago we were implementing a new Electronic Patient Record as part of the less than well managed NPfIT. All staff, regardless of role, were required to take a training session and pass a web-based test.

One right-click-view source later I noticed that the code for the options actually specified which of the options was "correct" when the test form was submitted. Yes I did tell the entire of my training cohort.

Of course it was such a well thought out and not at all rushed job that one of the required fields in the training environment for a Hospital EPR was "Species"

[Excused Boots]

Re: GOOD

"They'll ask the exact same questions... in a slightly different order!”

Genius!

Complete twats - that is all!

[David Hicklin]

Re: GOOD

"They'll ask the exact same questions... in a slightly different order!”

The other favourite was changing the question from something like "which of these apply" to "which of these do not apply"

I guess it is to at least make you read the question, I had years of screenshots on the end.

[Anonymous Coward]

Re: GOOD

Last year, there was an option to view the video's transcript making it accessible to those with hearing or sight difficulties. Expanding the transcript removed the need to let the video play out.

This year, that was missing, so I put a complaint into HR that the videos in our anti-discrimination training were themselves discriminatory. I doubt the transcripts will come back, but it's still fun to put a cat amongst the pigeons from time to time.

[IanRS]

Re: GOOD

Lucky you.

I recently had to sit through an hour long collection of video-based presentations on various DEI topics. The training system would not let you skip any of the video - you had to watch it start to end - otherwise it refused to present the end of section questions and you had to replay the video, from the start.

[David Hicklin]

Re: GOOD

> including the hated "click all that apply"

I wish I could upvote that one more than once

Worst thing is most of the "compliance" training was a total waste of time as the vast majority of us would never be in a position to be effected by any of it.

[Lee D]

Re: GOOD

I get this all the time.

I don't even use spell-check or grammar-check, I hate it. Not because I'm perfect but because... it's wrong far more often than I am. It wants to reword sentences that I don't want reworded. If you're writting narrative and slip in some deliberately poor grammar or shorten some text ("a'right, luv") it's because you WANTED to do that! Many places try to "correct" my British English to American English (feck off!), and I seem to use a lot of words that are NOT in most dictionaries. Things like "github" and the like.

I asked our helpdesk provider to NOT suggest that my team use AI to respond to tickets automatically or have AI "summarise" (with an S!) tickets for them. It was an uphill struggle to get them to put in an off-switch.

I don't want code-completion in my IDE... unless I ask for it. If I half-type a word and press a shortcut key for it, fine. But otherwise I don't want lists of nonsense suggestions flitting round the screen as I'm typing, especially when they suggest ridiculous completions, and ESPECIALLY if they do that thing where they auto-complete what you're typing and move you onto the next word before you're finished, so you splat the rest of the text into a nonsense repeat. It's like trying to just type code out while being interrupted by an ADHD child trying to complete all your sentences as you're halfway through speaking.

And the crux of the matter is: Not only do I not want to use it, but I don't want it to be in my software at all. And I don't want to pay for it. It's THAT USEFUL to me, that I actually want it disabled or just not present whatsoever. Same for voice recognition. Same for face-unlock. Same for "I'll search all your storage for that file starting with A" because I started to type "A" as part of a filename in a file dialog. Same for all kinds of stuff. If you insist on bundling it, give me an off-switch for it.

Even AMD drivers are now starting to bundle a local LLM install to "help" with their software. 6GB of LLM running on your machine in case you have a question about your graphics driver. Just.... no.

There's a reason that my Christmas present to myself was a Framework laptop so I could run Linux as an officially-supported OS on some new hardware. My computer is fantastically BORING again. It just does what I tell it, and what's absolutely necessary, and nothing more. And not once has it tried to engage me in conversation.

[Like a badger]

Re: GOOD

My computer is fantastically BORING again

Nirvana.

Can't wait until the nipper has completed her finals, and I can expunge Microsoft from my small domestic computer fleet (currently it's all tied together by an Office 365 family subscription and need to stick with that for her coursework).

[Snake]

Re: AMD drivers

Wow, seriously? AMD is building LLM's into their DRIVERS?! And why aren't they being called out on that?

[David Hicklin]

Re: GOOD

> Many places try to "correct" my British English to American English (feck off!),

I used to hate outlook at work when an email came from someone with American spelling on and it insisted that my reply was going to be spelt the same way. I think I defeated it in the end and as I am retired and will never, ever have to endure outlook again.

[steviebuk]

Re: GOOD

Yep and if you use the Google one, I forget what its called, not ChatGPT, you can get it to look at your screen and speak and say "What is the answer to this" it will scan it and give the answer :)

Not that I've used it for mandatory training ;o)

[Anonymous Coward]

Re: GOOD

Pre AI, I answered multiple choice “training” by random selection.

Don’t even read the questions, just select any option.

Mash, mash, mash: Fail.

Mash, mash, mass: Fail.

Mash, mash, mash: Pass

It was an ethics course too.

[Jeff 11]

Re: GOOD

"Despite the CEOs' repsonses, PwC concludes more investment is required. It claims that "isolated, tactical AI projects" often don't deliver measurable value, and that tangible returns instead come from enterprise-wide deployments consistent with business strategy."

I wonder why

[Steve Hersey]

Re: GOOD

Was going to comment on that paragraph.

Sounds suspiciously like "we lost money on every sale, so we plan to make it up on volume." Or perhaps, "you're not losing money on us fast enough."

[David Hicklin]

Re: GOOD

"Despite the CEOs' repsonses, PwC concludes more investment is required"

So a bit like going over the top from the trenches in WW1, it did not work the first dozen times so we will just send even more this time - that will fix it!

[the Jim bloke]

Re: GOOD

When the purpose of the exercise is to downsize your army - sending them towards someone elses machine guns is an effective method..

And that is how the AIs and MBAs will be learning from history.

[Anonymous Coward]

Re: GOOD

I've tried (because work was banging the AI drum) to use copilot a couple of times. It's hit and miss, mostly miss because it keeps getting the numbers wrong which requires a whole bunch of work to check before I'd even get to fixing it.

No, just no.

[VoiceOfTruth]

A fool and his money are soon parted

AI snake oil salesmen: Look, we have a shiny golden AI widget.

Buyers: Oh, it's shiny on the surface. Let's 'invest' in this wonderful magic. Everyone is doing it, we need to too.

Time passes.

MostBuyers. This shiny AI widget is not quite so amazing after all. It seems more of a distraction. It doesn't save me either time or money.

SomeBuyers: It works for us in our particular narrow use cases.

SurveyPeople: MostBuyers are doing it wrong. They may be holding a pair of threes, but they need to go ALL IN. They need to spend more money. They also need to outsource all their IT and join it up so anyone can access anything.

[retiredFool]

Re: A fool and his money are soon parted

Similar to a tax increase that was resounding voted down for my city. The mayor's response to the no vote for the council's need for more money was, "We presented the issue to the public incorrectly". It wasn't we hear you, we will make due with the current tax revenue. I'm sure next year council will go to the pig trough for more money again.

Or years back a company I'd worked for was the result of a merger. One used offices, one used cubes. Management pinky sweared that they would implement the result of the poll of do we go offices or cubes for the new merged co. Even though the office side of the merger was about 1/3 of the people vs the cube side was 2/3, the poll was a resounding 70/30 offices. The pinky swear broke, and management said we are going with cubes because they are better. (IE cheaper)

Conclusion, Upper crust always does what they want.

[Doctor Syntax]

the report cautions that "clearly, we're in the early stages of the AI era."

The caution is more likely to enable them to plausibly deny their report burst the bubble.

[Anonymous Coward]

A lot of these people seem to think OpenAI invented LLMs.

[Doctor Syntax]

A lot of which people?

And does it matter? It's all money down the drain when the bubble burst. There's not much point arguing whose drain it is.

[Anonymous Coward]

My point was that we're not in the early stages, we're a decade in already for large models available out of the lab.

[Anonymous Coward]

https://en.wikipedia.org/wiki/ELIZA

1964-1967.

[Like a badger]

And does it matter? It's all money down the drain when the bubble burst. There's not much point arguing whose drain it is.

Doesn't matter whose drain it is, but it matters that the bubble is so over inflated that it will be OUR money bailing out all the shitty corporations, banks, private equity and pension funds when it all turns to ash.

[Excused Boots]

By ‘our money’ you mean tax-payers and pension funds - ‘oh yes you have been paying in to a fund for all of your working life, but we invested it in AI stocks and Alchemy (look they absolutely promised they could turn lead into gold, it was too good to pass up). Sorry about that, yes you were looking towards a nice, comfortable retirement - but about that, how about you work into your 80s? And this time it’ll be absolutely fine.....

Brick wall, blindfolds, firing squads......

[David Hicklin]

> By ‘our money’ you mean tax-payers and pension funds

Had my annual review with my financial adviser yesterday, all those tech stocks and other stock market growth made 2025 an amazing year for pension growth - pretty much a record year not likely to be seen again. Of my portfolio of investments (spread across many areas) there is on average a 70% gain/30% loss ratio each year , in 2025 only one single investment lost - everything else went up. The end of 2024 to retire was definitely a good time as my other final salary pension means my private one here does not need to be touched - it just goes up and up.

Yes there will be a correction but you also have to consider your "blend" of investments and if you are with a good pension provider they will have spread it around a bit, it also depends which level of risk/reward you put yourself in.

And yes, everything is overvalued at the moment which will make the bubble bursting even more painful as I am sure the fingers are posed on the "get out of here" buttons when it does go.

[Anonymous Coward]

There is hardly any content for my shiny old 3DTV, and I can’t seem to find a new 75” one for some reason.

[captain veg]

The AI era is at least 50 years old. We might be said to be in the early stages of the current AI warm period.

https://en.wikipedia.org/wiki/AI_winter

-A.

[the Jim bloke]

report cautions that "clearly, we're in the early stages of the AI era."

And the vivisectionists, and doctors refusing to wash their hands between performing autopsies and birth deliveries were at the early stage of modern medicine...

Just because something seems new and shiny, doesnt mean that 5, 10, 20 years down the track it wont have turned out to be hideously stupid.

[BinkyTheMagicPaperclip]

Isolated projects *should* show value

If the technology doesn't depend on economies of scale (and AI does not), and you can't make money or reduce effort on a cherry picked, well defined task, then the technology isn't worth it.

I've not been using LLMs, but I'm fairly certain they're not completely useless based on friends and articles. Nevertheless there's a lot of evidence that there's sufficient disadvantages to using it, that overall it isn't enough of an improvement based on the cost, energy, and environmental impact.

We're not in the 'early stages' a few years in, the tech curve is heading for 'the wheels fall off the bus' stage, on to the rapid descent before 'it's used for what it's actually useful for, and what will make money'. The wheels haven't fallen off yet, but the bearings are making an unpleasant rattle. Sell the bus now.

Neither is it working in the consumer space, if they want to raise prices and increase adverts. Please add adverts and stick prices up, it'll starkly illustrate that consumers will play around when it's free, but once they have to pay it'll be dropped like a hot potato.

[Anonymous Coward]

Re: Isolated projects *should* show value

> However, pilot projects are by their very nature ...

... picking the low hanging fruit, and if you can't make it work here...

[Like a badger]

Re: Isolated projects *should* show value

it'll starkly illustrate that consumers will play around when it's free, but once they have to pay it'll be dropped like a hot potato

I wonder how many will fall for the Microsoft dodgy tactics of putting the price of Office 365 up by 33% "to include Copilot", but the thieving, dishonest bastards don't show as an option the cheaper version without Copilot, unless you click "cancel my subscription"?

[Mimsey Borogove]

Re: Isolated projects *should* show value

it isn't enough of an improvement based on the cost, energy, and environmental impact.... consumers will play around when it's free, but once they have to pay it'll be dropped like a hot potato.

I played around with it when it first burst out in, what, '22? But reading about all the costs has persuaded me that there's nothing I might want to do with "AI" that would be worth that. People need to be hit sooner rather than later with the need to pay their share for their "playing around."

[vtcodger]

Mandy Rice-Davies Rides Again

"Despite the CEOs' repsonses, PwC concludes more investment is required. It claims that "isolated, tactical AI projects" often don't deliver measurable value, and that tangible returns instead come from enterprise-wide deployments consistent with business strategy."

English Translation: "We know the product sucks, but we're hoping for a miracle."

File under They would say that, wouldn't they and/or You're Holding it Wrong

[Bebu sa Ware]

Re: Mandy Rice-Davies Rides Again

"File under They would say that, wouldn't they and/or You're Holding it Wrong"

Or under You're Holding someone else's… rightly or wrongly… more in Ms Keeler's line I suspect.

[Anonymous Coward]

Re: Mandy Rice-Davies Rides Again

PwC concluded: People should pay PWC lots of money to fix their AI business strategies, cause trust me bro.

Amazing! Who could possibly imagined they would come to this mind blowing conclusion.

[Rich 2]

Use case?

Another “AI” story. Another lack of an application.

Yet again, there is all this reference to “AI” without actually saying what the application is. It’s literally talking about a totally non-specific “thing” - it’s completely meaningless

[wolf2600]

Everybody.... back to blockchain!!

It's this amazing technology to decentralize transactions through an immutable ledger.

It's going to revolutionize the ${USER_INDUSTRY} industry!

[Tron]

They were paying for an education.

Next time they are offered snake oil by tech billionaires, they might keep their money in their wallets and not be such mugs.

[Doctor Syntax]

Re: They were paying for an education.

You'd think that. You'd hope that. History says they won't.

[MachDiamond]

Re: They were paying for an education.

"Next time they are offered snake oil by tech billionaires, they might keep their money in their wallets and not be such mugs."

Yes, but it makes sense to the kids and we are afraid of looking like decrepit old dotards.

I'm not going to say that it "won't" work, but I'm not seeing a good reason why I should buy it. I'm also seeing a rising tide of "no-buy"/"Zero-buy" memes on YouTube to tame bloated households and budgets.

[Anonymous Coward]

Removed

For whatever reason I was added to a copilot pilot

I instantly asked to be removed but was told I needed to wait 6 weeks and would get removed if I didn't use it

Today is just over 7 weeks, but it is being REMOVED

Got a survey too -- hope it helps them to save money

[Anonymous Coward]

Re: Removed

A while back, Github "gifted" me a Copilot subscription because I've contributed to various OSS projects.

Their support were quite perplexed when I contacted them and asked them to take it back.

[martinusher]

AI is actually quite useful, just not a universal panacea

Apart from the headline stuff (protein folding etc.) stuff there are situations where AI has proven useful. For example, an accountant friend (yes, they're real people) told me recently how its saved a huge amount of effort in his business making spreadsheet pivot tables. A simple and useful application, one of many. The problem is that this doesn't translate to the kinds of RoI on billions in investment that AI goldrush is expecting. He just sees this use as a natural evolution, the way that the word processor superseded the typewriter -- saves time and money but not that kind of money.

[Anonymous Coward]

Re: AI is actually quite useful, just not a universal panacea

The problem is that you only save money if you don't spend time checking its work.

[Adair]

Re: AI is actually quite useful, just not a universal panacea

And there we have it: what price UNRELIABILITY + TIME WASTING?

[frankvw]

Re: AI is actually quite useful, just not a universal panacea

"...there are situations where AI has proven useful..."

This. Exactly this. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

I can already hear the flamethrowers being fired up in the background, but I'm going to say it again: I struggle to understand the intense, vitriolic hatred against all things AI-related among the Reg commentariat.

Yes, AI is over-hyped. That's a problem with the IT industry and IT manglement, not with AI itself.

Yes, AI is no more intelligent than a Windows Solitaire game. The fact the some people think it's actually clever is a problem with people's perception of it, mainly due to the fact that AI is a misnomer. We're not dealing with Artificial Intelligence but only with Simulated Intelligence. If it were referred to by the latter, there would be a lot more realism surrounding it.

Yes, LLMs (which are only a subset of AI as a whole) are being trained on content in ways that are at odds (to put it kindly) with existing copyright law as we know it. That proves we need new copyright paradigms that accept the reality of content being incorporated in an LLM, as opposed to content being redistributed as-is. AI is disruptive technology. This is nothing new. The media industry tried to make war on digital audio and video being distributed on the Internet when that technology first became available; yet now we're looking at streaming services obliterating previous means of content distribution. Similarly, AI will also necessitate some radical paradigm shifts. But that problem is one of inertia within the content industry and its reluctance to respond to radical changes, not with AI itself.

Yes, AI is still unreliable and prone to failures in catastrophic (and often hilarious) ways. Of course it is: the technology is relatively new and still evolving. That's a problem with over-eager early adoption of new tech and a lack of caution by the user, not with AI itself.

Yes, manglement has become delusional to the point where they use AI to decimate the workforce and in some cases rehiring said workforce at a lower wage. That's a problem with incompetence and greed in the boardroom, not with AI itself.

Yes, some believe that buzzcoding can replace real devs. That's a problem with corporate stupidity, not with AI itself.

And so on.

Nor is this anything new.

Let's set the Wayback machine to the late 1990s, when the "New Economy" was the be-all and end-all. Tiny start-ups were given millions to play with and companies who had never made a single product, let alone any profit, went IPO and became worth billions overnight. I wondered out loud in those days how that sort of nonsense could ever be sustainable. I was met with scorn: my questioning the viability of the New Economy clearly illustrated my complete lack of understanding of how it worked. It was all about the Internet, and E-commerce, and the Virtual Workplace, and Java replacing regular applications, and Wearable Computers, and the words "dot com" which, when tacked onto anything at all, would ensure Instant Magic.

Then the bubble burst. And not a moment too soon.

But strangely, the Internet is still here, bigger than ever. Working remotely is still here, bigger than ever. Java has not replaced traditional software, but it has earned its place in the IT market; just look at large financial institutions or, closer to home, something like Apache Netbeans. Wearable Computer tech looks even more ludicrous now than it did back then, but everyone has a smartphone in their pocket, often with more computing power than the average desktop in 1995. E-commerce is also still here, bigger than ever: many of us order more goods and services online than what we buy through regular outlets.

My point (yes, there is one!) is that once the hype wears of, as it inevitably will, the core of technology that it sublimated around will continue to exist and function, and it will find its proper, useful application, whatever that may turn out to be. AI is no different. The bubble will burst, the hype will wear off, reality will set in, there will be a shake-out in which the more vapid players in the market will go under, and that is all as it should be. But AI itself will persist. It's here to say. Just look at history.

[anothercynic]

Re: AI is actually quite useful, just not a universal panacea

I don't think the El Reg Commentariat have a problem with AI/LLMs in general (i.e. LLMs being used for useful things), but rather the overblown hype involved in the CoPilot/ChatGPT/Grok/... purveyors who seem to believe that general AI will solve all the world's problems.

Most of us who have an aversion to the hype are all old enough to know that it won't, mostly because we've actually been through the whole development lifecycle for them and know how LLMs are constructed, and don't buy into the "all the AI for all the things" bollocks. And yes, many of us are of the age where we've seen the multiple dot-com bubble bursts and have become yet more cynical about the ability of these young hypers to actually think critically and avoid starting another bubble, which inevitably leads to yet another financial crisis because those tend to spiral out of control and affect people who had no skin in the game in the first place.

:-)

[Sorry that handle is already taken.]

Re: AI is actually quite useful, just not a universal panacea

I struggle to understand the intense, vitriolic hatred against all things AI-related among the Reg commentariat.

It's a defence mechanism against the grift.

[Benegesserict Cumbersomberbatch]

You what?

Yes, LLMs (which are only a subset of AI as a whole) are being trained on content in ways that are at odds (to put it kindly) with existing copyright law as we know it. That proves we need new copyright paradigms that accept the reality of content being incorporated in an LLM, as opposed to content being redistributed as-is.

So, a business model dependent on behaviour that is both illegal and immoral is tolerable because it is "disruptive"? We need to change the law to tolerate the fact that something is happening, rather than bring to justice any who have illegally enabled it? Organised crime is happening, but dealing in illicit drugs, extortion, threats and violence are still illegal, and for good reason.

[Pulled Tea]

Re: AI is actually quite useful, just not a universal panacea

Yes, AI is over-hyped. That's a problem with the IT industry and IT manglement, not with AI itself.

What's AI? Is it anything that computers can't do yet? Is it an ideological project to shift authority and autonomy away from individuals, towards centralized structures of power? Is it a blanket term applied to any system that claims to supplement, reproduce, or replace human actions, decision making, or reasoning?

Some of the things that are defined as “AI” are fundamentally useless (Tessler's definition literally means vaporware). Wilson's definition is slightly better, but means… anything, really — a system that's supposed to supplement, reproduce or replace human actions, decision making or reasoning is literally automation. Al-Khatib's definition? Talks about a system that sucks and needs to be resisted every step of the way, unless you fancy yourself the Spartiates of the bold new world that political project will bring into life, rather than the Helots being ground into the dust (you are likely not going to be a Spartiate).

Like, what utility is AI when, come the winter, everyone will start calling their projects “machine learning” or “computer vision” or whatever it is they called it before the band of swivel-eyed loons decided to inject centibillions into their projects to avoid the stink of insanity that happened when things were hot?

[Anonymous Coward]

Re: AI is actually quite useful, just not a universal panacea ... <=== Not true by a long chalk !!!

Lets start at the top ... IT is NOT 'Simulating Intelligence' ... Full Stop.

There is NO connection with 'Intelligence' in any way ... no matter how hard you twist & turn.

If it could 'Simulate Intelligence' it 'might' have some 'use' ...

It is 'pattern matching on steroids' BUT the pattern recognition is flawed and it completes the patterns with 'bad data' rather than say 'I don't know' !!!

'The intense, vitriolic hatred against all things AI-related' is because 'AI' is being mis-sold, not by accident BUT deliberately to gain profits for a thing that does not work !!!

What is the purpose of something that cannot be depended upon to give a correct answer at all times !!!

It is in use but the results are still not so useful that they can eliminate the 'Answers may be wrong, small print'.

Imagine if accounting software was sold with get out clauses stating that the calculation may be wrong and all accounts should be checked for accuracy !!!

'AI' is a long shot that has not worked ... the expectation was that the 'issues' would be 'ironed out' as the systems were developed and experince gained.

'AI' is NOT improving ... the marketing is accelerating and the claims are multiplying regarding its use and real-world value.

No ammount of enthusiasm or 'faith & Hope' will make 'AI' better, it has not got better as the $Billions have been spent.

The need to be successful is driven ONLY by the size of the spend to date ...

As the old saying goes ... 'you can fool ALL of the people SOME of the time BUT cannot fool ALL of the people ALL of the time !!!'

'AI' is a failure as a general purpose tool, it can be useful in a focused arena of knowledge with specifically curated data fed into the 'AI'.

The game is up and all efforts from now on are simply trying to 'wing it', hoping that the majority of people will live with a stochastic 'Knowledge Gatling Gun' ... a entertaining toy BUT no more !!!

It cannot write code, it cannot answer the majority of questions accurately & repeatably (answers change as the model 'learns' something 'new' from the questions and the last answers), it is not safe to replace 'real people & knowledge' ... a real harm is going to happen because someone is going to believe the 'garbage' answers and NOT check if they are correct/safe before use.

:)

[Anonymous Coward]

Re: AI is actually quite useful, just not a universal panacea

When a known Vatnik troll has been told to promote the AI bubble, we should all realise we are being set up for a fall.

[DS999]

CEOs: we're not seeing a payoff from our AI investment

PWC: that's because you're not investing enough, keep spending gotta keep the bubble going!!!

[IvaliceResident]

Re: CEOs: we're not seeing a payoff from our AI investment

Also PwC: Spend all the money on PwC services.

[Groo The Wanderer - A Canuck]

Oh lovely - the Reg is being spammed with what are obviously AI SLOP stories. This is almost word for word what Gartner claimed two weeks ago.

Go ahead. Belive in the Magic Silver Bullet like every SUCKER has every second decade for the past 80+ years of the computing industry. What is it about the 20 year cycle? The new generation being too damned stupid to learn from their parents?

[anothercynic]

The new generation being too damned stupid to learn from their parents?

Errr... yes. Exactly.

[ecofeco]

No problem!

They'll just raise prices to make up for the bad decision!

No problem!

[anonymousI]

Re: No problem!

Of course they'll have to raise prices.

All that new tech is very expensive to develop and install, y'know.

[M.V. Lipvig]

"increased revenue nor decreased costs"

Now ask how many saw decreased revenue or increased costs from AI.

[DJO]

It cannot be easy to disentangle "AI" losses or gains from all the other geopolitical and macroeconomic pressures at the moment, it's a horrible thought but maybe (but probably not) "AI" stalled the worst of the damage caused by the orange menace.

I wonder if there will ever come a time when "AI" does not need the inverted commas?

[Winkypop]

AI, not particularly profitable

Unlike consultancy reports, booming!

[MooseMonkey]

Business Cases

Very few businesses owners or profit & loss owners seem to understand simple business cases any longer, and are swept up by the IT department or vendor selling smoke and mirrors.

I was a programme manager in a large mobile phone company, and the director of the customer service division used to hand all the “business cases” from the IT dept to me to filter them. My favourite was “New TV screens for the contact centre”, that claimed that slightly bigger flatscreens on the walls of the call centre would reduce the agents time spent looking at the screen by 0.1 seconds per view, and save 6 minutes a day for all 2,000 advisors hence a huge saving.

Obviously that one went in the bin, but before I worked there, many like this had been passed.

[Anonymous Coward]

Re: Business Cases

Bad rationale has been a thing since the Emperor invested in new clothes. I tried to block a business case for a new product line for a large UK energy company, because even the generously padded business case showed it was loss making for all seven years of sales projections. I was taken aside and told that I couldn't say no, because our German parent company had identified this technology as core to the group, so it was essential that the UK business also offered it even though we knew the UK market was tiny.

The same applies now, that everywhere ill-informed, technically illiterate CEOs are deciding that AI is a transformative technology whilst suffering from the usual C-suite FOMO, then insist that it is pushed into every potential use, regardless. I now work for government, and they're doing exactly that.

[TheServitor]

Jives with my experiences. Was writing the other day about a related theme, why AI works for some few (on some things) but doesn't seem to catch. It's hard to transfer effective AI use skills. Or even keep up with them. I struggle to consistently apply them to my own work, much less get colleagues set up.

It's so easy to just lazy prompt and settle into a single conversation flow instead of breaking down tasks efficiently, thinking ahead, working in parallel, which is where genuine productivity have happened for me. And many kinds of job simply can't be done in parallel or turned over to agents, regardless of whether I'm focused and disciplined in method.

So, yep, article sounds right to me.

[naive]

ROI with AI

It all depends on what an employer is doing. If an employer is producing bread toasters in a factory he inherited from his grandfather, AI will be less relevant than it is for companies like Palentir or a startup planning to produce drones capable to fully autonomously scan battlefields and remove all hostile combatants its sensors are able to pick up. AI will be with us until the moment it becomes greater than us. When that happens, we will be removed in a Darwinian manner. At least it generates funny and sometimes beautiful youtube content, enjoy it while you can.

[Jedit]

Looks like it's time for a few hits of cocaine

Yep, it's another episode of the game where we replace "AI" with "cocaine" and see if the meaning of the sentence changes or becomes less valid.

"More than half of CEOs report seeing neither increased revenue nor decreased costs from cocaine, despite massive investments"

""isolated, tactical cocaine projects" often don't deliver measurable value"

"So if your cocaine projects fail, you clearly just don't believe enough."

I'm giving this one 10/10 for accuracy and 8/10 for humour. Also I'm calling dibs on The Tactical Cocaine Project as a band name.

[An_Old_Dog]

Misreading

I originally misread the headline as, "Majority of CEOs report zero payoff from AI scourge".

[ABugNamedJune]

AI doesn't work, but just because you're not AI-ing hard enough

It claims that "isolated, tactical AI projects" often don't deliver measurable value, and that tangible returns instead come from enterprise-wide deployments consistent with business strategy.

Can't wait for my boss to start insisting I use CoPilot integration in Exchange Admin, an already futile application made more futile with a speak n' spell trying to create my groups for me.

[Taliesinawen]

Who knew :o

Who knew the AI vendors would wildly exaggerate and oversell the technology.

[Anonymous Coward]

The company I work for is always just a little behind in following hype trends. However, when they jump in, it is with both feet. This means that 2026 is looking very dismal when it comes to force-feeding AI down developers throats (2025 was the "AI experimentation and use is encouraged" year, 2026 will be "agentic AI is mandatory in all things, spending lots of money is mandatory, and if there is no payback it is R&D's fault for not trying hard enough").

I don't always agree with Cory Doctorow, but in this case his description of AI as centaur (relatively positive, machine assisting the human who is calling the shots, what was largely happening here in 2025, maybe even to slight benefit) vs. reverse centaur (entirely dystopic, human assisting the machine that is calling all the shots, what appears to be the plan for 2026) is completely aligned with my experience.

My only hope is a big hype collapse, and soon. Else I simply have to get out of here.

[mcswell]

Margin of error

"PwC warns that companies avoiding major investments due to geopolitical uncertainty underperform peers by two percentage points in growth and three points in profit margins." You can measure that???