Our new CTO was remarking that our engineering teams AI spend is too low. I believe we have already committed a lot of money but only using 5% of the subscription.
This is likely why there is a lot of push from the top. They have already committed the money now having to justify it.
Simple fact: AI is extremely powerful, in the hands of experts who invested time in deeply understanding it, and in understanding how to actually use it well. Who are then willing to commit more time to build an actually sustainable solution.
Alas, many members of the C suite do not exactly fit that description. They just have typed in a prompt or three, marveled that a computer can reply, and fantasize that it's basically a human replacement.
There are going to be a lot of (figurative, incorporated) dead bodies on the floor. But there will also be a few winners who actually understood what they were doing, and the wins will be massive. Same as it was post dot-com.
If we ignore coding or tech industry for a min. Other companies Keep demanding new Report on certain things, and AI is doing that. Is it productive? probably not. But do execs loves it? Yes.
In a non-start up, bureaucratic companies, these report are there to make cover ups, or basically to cover everyone's ass so no one is doing anything wrong because the report said so.
This reminds me of the internet in 2000. Lots of companies doing .COM stuff but many didn’t understand what they were doing and why they were doing it. But in the end the internet was a huge game changer. I see the same with AI. There will be a lot of money wasted but in the end AI will be huge transformation.
The claim that big US companies “cannot explain the upsides” of AI is misleading. Large firms are cautious in regulatory filings because they must disclose risks, not hype. SEC rules force them to emphasise legal and security issues, so those filings naturally look defensive. Earnings calls, on the other hand, are overwhelmingly positive about AI. The suggestion that companies only adopt AI out of fear of missing out ignores the concrete examples already in place. Huntington Ingalls is using AI in battlefield decision tools, Zoetis in veterinary diagnostics, Caterpillar in energy systems, and Freeport-McMoran in mineral extraction. These are significant operational changes.
It is also wrong to frame limited stock outperformance as proof that AI has no benefit. Stock prices reflect broader market conditions, not just adoption of a single technology. Early deployments rarely transform earnings instantly. The internet looked commercially underwhelming in the mid-1990s too, before business models matured.
The article confuses the immaturity of current generative AI pilots with the broader potential of applied AI. Failures of workplace pilots usually result from integration challenges, not because the technology lacks value. The fact that 374 S&P 500 companies are openly discussing it shows the opposite of “no clear upside” — it shows wide strategic interest.
the AI umbrella has been helpful to my BigCorp to justify more machine learning work, or discrete optimisation and scheduling problems
agentic ai which is a huge buzz in enterprise feels more like workflow and rpa (again) and people misunderstanding that getting the happy flow working is only 20% of the job.
For most companies AI is a subscription service you sign up for. Because of great marketing campaigns, it has become a necessary tax. If you don't pay, the penalty is you lose value because it doesn't look like you are embracing the future. If you pay, well it's just a tax that you hope your employees will somehow benefit from.
As a small business owner in a non tech business (60 employees, $40M revenue), AI is definitely worth $20/month but not as I anticipated.
I thought we'd use it to reduce our graphics department but instead we've begun outsourcing designers to Colombia.
What I actually use it for is to save time and legal costs. For example a client in bankruptcy owes us $20k. Not worth hiring an attorney to walk us through bankruptcy filings. But can easily ask ChatGPT to summarize legal notices and advise us what to do next as a creditor.
Looking at my own use of AI, and at how I see other engineers use it, it often feels like two steps forward and two steps back, and overall not a lot of real progress yet.
I see people using agents to develop features, but the amount of time they spend to actually make the agent do the work usually outweighs the time they’d have spent just building the feature themselves. I see people vibe coding their way to working features, but when the LLM gets stuck it takes long enough for even a good developer to realize it and re-engage their critical thinking that it can wipe out the time savings. Having an LLM do code and documentation review seems to usually be a net positive to quality, but that’s hard to sell as a benefit and most people seem to feel like just using the LLM to review things means they aren’t using it enough.
Even for engineers there are a lot of non-engineering benefits in companies that use LLMs heavily for things like searching email, ticketing systems, documentation sources, corporate policies, etc. A lot of that could have been done with traditional search methods if different systems had provided better standardized methods of indexing and searching data, but they never did and now LLMs are the best way to plug an interoperability gap that had been a huge problem for a long time.
My guess is that, like a lot of other technology driven transformations in how work gets done, AI is going to be a big win in the long term, but the win is going to come on gradually, take ongoing investment, and ultimately be the cumulative result of a lot of small improvements in efficiency across a huge number of processes rather than a single big win.
I've found it to be a significant productivity boost but only for a small subset of problems. (Things like bash scripts, which are tedious to write and I'm not that great at bash. Or fixing small bugs in a React app, a framework I'm not well versed in. But even then I have to keep my thinking cap on so it doesn't go off the rails.)
It works best when the target is small and easily testable (without the LLM being able to fudge the tests, which it will do.)
For many other tasks it's like training an intern, which is worth it if the intern is going to grow and take on more responsibility and learn to do things correctly. But since the LLM doesn't learn from its mistakes, it's not a clear and worthwhile investment.
Sounds like blockchain all over. Reminds me of an essay from two product managers in AWS that talked to clients all over US and couldn't get any business to clearly articulate why they need blockchain.
Note: AWS has a hosted blockchain that you can use. [1]
PS: If anyone has read that essay, please do share the link. I can't really locate it but that's a wonderful read.
Large language models are a deeply impressive technology, but they're not artificial general intelligences, because you need to supervise them. Like everything else that has been called 'artificial intelligence' since the 1950s, I think we'll find some niches that they're good for, and that'll be the end of the hype bubble.
The hype does serve a purpose, though: it motivates people to try to find more possible uses for LLMs. However, as with all experiments, we should expect most of these attempts to fail.
Big benefit from coding Agents: for things to work it better have documentation.
Which humans usually aren't given, so anything which forces documentation is good.
one of the great benefits of AI so far has been the push for more plain text documentation and opening up API access via MCP. Let's enjoy it while it lasts until we are forced back into walled gardens and micro transactions
Well, as anecdotal data, have you folks noticed ads lately pushing Gemini/Claude/xx on both legacy media and online? If AI (and these products) is sooo great, why do these companies have to advertise to sell their wares?
And Google and Microsoft are hellbent on pushing AI into everything. Even if users don't want it. Nope, we're gonna throw the kitchen sink at you and see if it sticks.
In the non-tech world, nobody gives a shit about AI. People and businesses go about their daily lives without thinking about things like "Hmmm...maybe I could have prompted that LLM a different way..."
The problem with AI is how confidently wrong it is. In Lisbon I uploaded a picture of myself on some church steps and asked ChatGPT where that is.
I came up with a place I was sure I’d never been. Then I asked if it could be part of some other place and it said sure it’s inside the main church. The pic was clearly outside. Next it gave me a random famous stair that is so clearly different a human could never be fooled.
Each of these lies were extremely elaborate citing sources and describing all the things that matched.
The only matching experience was with a taxi in Delhi some 20 years ago where the driver pretended to know where he was going and when I further questioned him he said the 40 story hotel I am looking for has been demolished 5 years after opening. At least he has monetary interest in lying to me so I enter his cab.
25 comments
[ 3.8 ms ] story [ 60.3 ms ] threadEnterprise is way too cozy with the big cloud providers, who bought into it and sold it on so heavily.
0: https://fortune.com/2025/08/18/mit-report-95-percent-generat...
This is likely why there is a lot of push from the top. They have already committed the money now having to justify it.
Alas, many members of the C suite do not exactly fit that description. They just have typed in a prompt or three, marveled that a computer can reply, and fantasize that it's basically a human replacement.
There are going to be a lot of (figurative, incorporated) dead bodies on the floor. But there will also be a few winners who actually understood what they were doing, and the wins will be massive. Same as it was post dot-com.
I said a couple years ago that the big companies would have trouble monetizing it, but they'd still be forced to spend for fear of becoming obsolete.
In a non-start up, bureaucratic companies, these report are there to make cover ups, or basically to cover everyone's ass so no one is doing anything wrong because the report said so.
The thing that changed it was smartphones (~7 years later). Suddenly, the internet was available everywhere and not just a thing for nerds.
Not sure that AI is quite there yet, currently trying to identify what will be the catalyst that makes it seamless.
It is also wrong to frame limited stock outperformance as proof that AI has no benefit. Stock prices reflect broader market conditions, not just adoption of a single technology. Early deployments rarely transform earnings instantly. The internet looked commercially underwhelming in the mid-1990s too, before business models matured.
The article confuses the immaturity of current generative AI pilots with the broader potential of applied AI. Failures of workplace pilots usually result from integration challenges, not because the technology lacks value. The fact that 374 S&P 500 companies are openly discussing it shows the opposite of “no clear upside” — it shows wide strategic interest.
agentic ai which is a huge buzz in enterprise feels more like workflow and rpa (again) and people misunderstanding that getting the happy flow working is only 20% of the job.
I thought we'd use it to reduce our graphics department but instead we've begun outsourcing designers to Colombia.
What I actually use it for is to save time and legal costs. For example a client in bankruptcy owes us $20k. Not worth hiring an attorney to walk us through bankruptcy filings. But can easily ask ChatGPT to summarize legal notices and advise us what to do next as a creditor.
I see people using agents to develop features, but the amount of time they spend to actually make the agent do the work usually outweighs the time they’d have spent just building the feature themselves. I see people vibe coding their way to working features, but when the LLM gets stuck it takes long enough for even a good developer to realize it and re-engage their critical thinking that it can wipe out the time savings. Having an LLM do code and documentation review seems to usually be a net positive to quality, but that’s hard to sell as a benefit and most people seem to feel like just using the LLM to review things means they aren’t using it enough.
Even for engineers there are a lot of non-engineering benefits in companies that use LLMs heavily for things like searching email, ticketing systems, documentation sources, corporate policies, etc. A lot of that could have been done with traditional search methods if different systems had provided better standardized methods of indexing and searching data, but they never did and now LLMs are the best way to plug an interoperability gap that had been a huge problem for a long time.
My guess is that, like a lot of other technology driven transformations in how work gets done, AI is going to be a big win in the long term, but the win is going to come on gradually, take ongoing investment, and ultimately be the cumulative result of a lot of small improvements in efficiency across a huge number of processes rather than a single big win.
It works best when the target is small and easily testable (without the LLM being able to fudge the tests, which it will do.)
For many other tasks it's like training an intern, which is worth it if the intern is going to grow and take on more responsibility and learn to do things correctly. But since the LLM doesn't learn from its mistakes, it's not a clear and worthwhile investment.
You can’t convince me otherwise, we just haven’t found a ‘killer app’ yet.
Note: AWS has a hosted blockchain that you can use. [1]
PS: If anyone has read that essay, please do share the link. I can't really locate it but that's a wonderful read.
[1]. https://aws.amazon.com/managed-blockchain/
The hype does serve a purpose, though: it motivates people to try to find more possible uses for LLMs. However, as with all experiments, we should expect most of these attempts to fail.
And Google and Microsoft are hellbent on pushing AI into everything. Even if users don't want it. Nope, we're gonna throw the kitchen sink at you and see if it sticks.
In the non-tech world, nobody gives a shit about AI. People and businesses go about their daily lives without thinking about things like "Hmmm...maybe I could have prompted that LLM a different way..."