HSBC To Investors: If India Couldn't Build an Enterprise Software Challenger, Neither Can AI (x.com)
- Reference: 0180831770
- News link: https://it.slashdot.org/story/26/02/20/167244/hsbc-to-investors-if-india-couldnt-build-an-enterprise-software-challenger-neither-can-ai
- Source link: https://x.com/refsrc/status/2024747695689376032
The bank's analysts contend that enterprise software competition turns on factors that have little to do with the ability to write code -- sales teams, cross-licensing agreements, patented IP, first-mover lock-in, brand awareness, and go-to-market infrastructure. If a massive, low-cost, domain-expert workforce couldn't crack the market over several decades, HSBC argues, the idea that AI-generated code will do so is, in the words of Nvidia's Jensen Huang that the report approvingly cites, "illogical."
[1] https://x.com/twitter/status/2024747695689376032
AI is closer to the customer, not further (Score:2)
Indian firms failed because they were further from the customer than incumbents.
AI will succeed because it's closer to the customer than incumbents.
Joe in accounting has a spreadsheet that the whole company depends on. He can tell the AI what the company needs, and more critically can iterate very quickly to tweak it so it is exactly what they need; no more, no less.
The incumbents have an advantage over the Indians because they have better access to Joe, but it's only imperfect one way information flow.
Re: (Score:3)
I agree with your headline. But success in software is not primarily due to the code, but to the business behind the code. A badly-run business with good code, will fail, while a well-run business that has (or inherits) bad code, is far more likely to succeed.
Re: (Score:1)
In that case Joe becomes an amateur programmer. Even if he gets AI to code something that works good enough, it's probably full of undocumented gotcha's, poorly normalized data, and requires special fudgy steps at times that only Joe understands or knows are needed. If Joe leaves the company, the company is F'd.
I've worked with too much amateur code or designs to know how they (don't) think.
Re: (Score:2)
> In that case Joe becomes an amateur programmer.
Joe has always been an amateur programmer. Spreadsheets are functional programming. Joe has been writing complex macros, wrestling databases and refining OLAPs for decades, as Joe built a mini-IT system inside his department, much to the chagrin of the actual IT dept. Now Joe has a powerful clanker to write scads of code.
> it's probably full of undocumented gotcha's, poorly normalized data, and requires special fudgy steps at times that only Joe understands
That's another one of Joe's days that end in "y". Joe doesn't care. Joe's bosses don't care.
> If Joe leaves the company, the company is F'd.
Now the company has clankers to wade through the mess Joe made. So that's not the problem i
Creating complex software is.. (Score:3)
..difficult, regardless of how powerful the tools are
AI tools turn a specification into code
Writing a complex specification is hard, really hard
Reminds me of the old "waterfall" approach to software design
Re: (Score:3)
> ..difficult, regardless of how powerful the tools are
> AI tools turn a specification into code
> Writing a complex specification is hard, really hard
> Reminds me of the old "waterfall" approach to software design
You can still do incremental development of the spec with the "AI" just building the software to let you find out what your spec actually meant and then rapidly regenerating your software as the spec changes. I think the problem really is that since the LLM isn't actually intelligent it isn't actually learning how to explain to you the problems it finds and can't invent new explanations for things that only it has had the chance to "understand" where usually you'd have a developer who actually was intellige
Enterprise software is bought with blowjobs anyway (Score:5, Interesting)
OK, I can't prove that whoever picked the enterprise software we use was fucking the sales rep, but that would be the most logical explanation. It can't be on the merits of the software...it's GARBAGE, from a top vendor, and just fucking sucks...when you analyze it, you can see it's using 20 year old technology and hasn't kept up with the times. It doesn't need to actually work, just convince the decision maker it did.
I got my career start as a software engineer at consulting company that specialized in SAP deployments. It was well known that the sales were never based on merit, but on bribing and charming the decision makers.
However, consulting company reps look a lot more like pharmaceutical salespeople than software experts. It's not old nerds with industry experience...we hired lots of women under 30 who were communications majors...they had breast implants and mastered the art of wearing business suits that showed them off...not enough to look like porn stars or that you'll notice immediately, but that you'll pick up on the cleavage during a conversation. They were sorority party girls that can pass for smart. The men were the male equivalents...charming bros that take you to the NBA/hockey game. I got to go to a few of those events.
The women mastered the craft of wearing their hair and makeup so that a routine straight guy wouldn't notice they were trying to look hot, but they were...they smiled the right way, they charmed the customer...your stories were always interesting..."oooh, you went golfing where?...that's so amazing!!!...oh, you're thinking about getting a boat?...that sounds like so much fun."
India, as famously corrupt as the country is reputed to be, their outsourcing firms aren't into bribing with sex...at least to my knowledge. The Indian outsourcing companies I've worked with, as much as I despise them, were really trying to compete with their work. The SAP/PeopleSoft consulting firms were more lavish in perks, dinners, hot bimbos, charming bros...trips to your see your local NBA/NHL/MLB/NFL team, etc.
Enterprise software is a depressing dumpster fire of corruption.
Re: (Score:3)
Happens all the time. A friend spent a full year flying back and forth from Southern California, staying in hotels, to meet with a cross-company team to figure out how to use the new software they'd licensed from a Perot company. After the full year (or more), they decided the software just wasn't going to work out, so they scrapped the project and the whole effort was for nothing.
Re: (Score:2)
God dammit I miss the hot bimbo sales girls.
What in the DEI has the world come to?
Booth babes and salesgirls suck anyway (Score:3)
> God dammit I miss the hot bimbo sales girls.
> What in the DEI has the world come to?
I'm autistic AF and even I could see they're full of shit. :) If you want to see hot women, go to a strip club or if you're charming and single, tinder. It's just awkward and weird. I don't like getting erections at work and I don't want bimbos interfering with my project. If want my project to succeed so I can get a promotion, not having it derailed and over budget, but hey....at least Becky will take me out to lunch a few extra times and pretend my jokes are funny and stories aren't boring.
It's jus
Re: Booth babes and salesgirls suck anyway (Score:2)
The whole point of their existence is to get things other people don't because they're special.
Re: (Score:2)
The world doesn't cater to your desires because you rode the short bus and can't control your sexual urges.
No one else likes you getting erections at work, either! WTF? But, that's an issue for you an HR. It's not a Hot bimbo sales girl issue.
Re: (Score:2)
> consulting company reps look a lot more like pharmaceutical salespeople than software experts....
> India, as famously corrupt as the country is reputed to be, their outsourcing firms aren't into bribing with sex...at least to my knowledge.
The Bollywood remake of [1]"Love and Other Drugs" [imdb.com] would be mighty interesting.
[1] https://www.imdb.com/title/tt0758752
Re: (Score:1)
> India,...their outsourcing firms aren't into bribing with sex
Unleash the Kama Sutra!
I think it's worse than that (Score:4, Interesting)
> ... the idea that AI-generated code will do so is, in the words of Nvidia's Jensen Huang that the report approvingly cites, "illogical."
Entirely aside from the socioeconomic connections. domain knowledge, and early-to-the-market advantages that Oracle and SAP enjoy, there's another reason that AI coding isn't going to crack the market open. Excepting cases where AI is a time-saving assistant to a good programmer who carefully vets the code produced, AI-generated code is shit.
I have no direct experience of what I just wrote; but almost everything I see and read online - aside from puff-piece propaganda - says that AI doesn't program well. It might give good snippets and a decent place to start, but for anything large and complex it seems to create barely-functional dreck.
And based on my (admittedly very limited) experience with programming, the really difficult parts of projects of this magnitude aren't things that AI can currently do. When it comes to understanding business logic and process flow - and adapting to changes in those factors, and making sure the code is both commented well and self-commenting where that's possible - AI just isn't there. So let's assume a wildly high figure of a 25% increase in programming efficiency across the board. Is that enough to replicate a decade or two of work experience in a few years or less? I don't think so.
As always, IANAP so I look forward to being educated and contradicted if what I'm saying here is wrong.
I agree (Score:2)
You have it about right. AI can code fairly simple programs and do a halfway decent job. It can't write complex software, it falls apart quickly as features are added. As a programmer, you can use AI to write the chunks of code needed to make a complex piece of software.
A podcast I listened to used AI to write a Linux/Qt based broadcast clip management program. It didn't do all of it by itself, but he wrote up requirements and asked it to write the pieces of code that filled those requirements, and piece
Re: I think it's worse than that (Score:3)
and making sure the code is both commented
well and self-commenting where
thatâ(TM)s possible -
I can and do agree with you 100% on the rest, but this made me LoL. Enterprise code being documented is a unicorn.
Frankly, Iâ(TM)ve thought about seeing what caliber of docs come out of handing some AI a codebase. Even if itâ(TM)s half-wrong, itâ(TM)d beat the nuthinâ(TM) I usually get.
Re: (Score:1)
The only way to get AI to work on big projects (and by work I mean say 80% of the way), it requires an existing, skilled, experienced programmer. The problem is not writing the code. The problem is writing the right code at the right time in the right place. None of the people attaching their existence to this train understand this. You can tell by how easily impressed they are by LLMs in general (and how they think they represent some form of consciousnessâ"oblivious that it's just math). This is sai
As a programmer and daily claude user, I agree (Score:4, Informative)
> Excepting cases where AI is a time-saving assistant to a good programmer who carefully vets the code produced, AI-generated code is shit.
> I have no direct experience of what I just wrote; but almost everything I see and read online - aside from puff-piece propaganda - says that AI doesn't program well. It might give good snippets and a decent place to start, but for anything large and complex it seems to create barely-functional dreck.
AI-generated code only sometimes compiles. It rarely works. It also is so slow it barely saves me time for easy tasks I know how to do. Today, I needed sample data for a unit test...PERFECT use case for AI...I prompted it to generate 3 examples of a specified class...it took several minutes in Claude Opus 4.6, so I just canceled it and did it myself. I retried it with another class and it didn't even compile. It is not remotely reliable. If it was, the world would be a different place, as I've ranted many times before at /.. A reliable code generation tool that could build nice software with prompts would set the world afire and these evangelists and puff piece purveyors wouldn't be writing about the glory of AI, but would be making vast fortunes chasing their software dreams.
Thats not really the threat (Score:2)
Or at least not the only threat. The barrier to entry was still very high, which meant the costs were high. You are going to get some large number of customers wondering why they are paying so much money for software services when new competitors are offering the same service for 1/10th the cost.
You are also going to get a lot more people asking why they are paying for services that are now not that difficult to just build in house.
These cases don't have to apply to everyone, but if it applies to 10%
HSBC must know (Score:3)
Fair. Here is the actually concise version.
Major HSBC scandals:
2012 — Money laundering & sanctions violations
Helped launder cartel money and processed transactions for Iran, Cuba, Sudan.
Fine: $1.9 billion.
2015 — Swiss Leaks tax evasion scandal
Swiss private bank helped wealthy clients hide assets offshore.
2019 — U.S. tax evasion case
Assisted Americans hiding assets in Swiss accounts.
Fine: ~$192 million.
2020 — FinCEN Files
Revealed HSBC moved large suspicious funds even after the 2012 settlement.
2021 — UK anti-money-laundering failures
Fine: £63.9 million.
Interest-rate manipulation (Euribor cartel)
Participation in benchmark-rate rigging.
EU fine ~€31 million.
UK government bond trader cartel
Illegal information sharing between traders.
Part of ~£105 million industry fines.
Deposit protection failures (UK)
Failed to correctly safeguard customer deposits.
Fine: £57 million.
Hong Kong disclosure breach (2025)
Failed to disclose conflicts in research reports.
Fine: ~$4 million.
Core pattern (one line):
Money laundering, tax evasion facilitation, sanctions breaches, market manipulation, weak compliance controls.
That’s the short list.
Re: HSBC must know (Score:3)
Yep.
Basically curruption is how contracts are won, not products. HSBC would know.
Probatio existentiae - Tesla One (aka Warp Drive) (Score:1)
Tesla One (originally Warp Drive) does it all, 100% developed in-house. So they're saying all of India's giant IT houses couldn't do what Tesla did internally? Yet another moat of Tesla's that few people understand. Ask Grok about it.
See:
https://www.youtube.com/watch?v=JzdziadEkzs
The Better Mousetrap never works (Score:2)
I agree with the assertions in this article. Even if AI builds a vastly better system than SAP or Oracle, companies won't buy it, and it comes down to an old saying. It's slightly different in every industry, but it comes down to this:
No one ever got fired by buying from GE/ThermoFisher/Oracle.
For this particular case, given how critical business systems are, a wrong move can screw up the business royally. That means the executives and the CEO even, but all the way down to the buyers are on the hoo
U.S. Incumbents (Score:3)
None of them have built a competing product that gained meaningful traction against the U.S. incumbents
...and here I thought SAP was German.
Re: (Score:2)
Yes, quite!
Nobody gets fired ... (Score:2)
... for buying IBM. Or AI.
In fact, you might get fired for NOT buying it. Or at least have your stock downgraded. A lot of investors are depending on the equities markets to bail out their investments into LLMs. And that means having the markets lean on corporations to buy and install that garbage.
There are no equivalent iffy investments in Indian IT systems in such need of bailing out. Or with the political connections.
They identified the issue... (Score:3)
That the prevailing winners in the market are not winning based on code, but on marketing. Now amount of 'here's a cheaper knock-off' is going to overcome the scare factor. Saving money won't win any of these decision makers enough to warrant the risk of cocking up something important.
However, AI is pitching supreme bespokeness. The cover is not just about 'saving money', the emphasis is that it can be fit to your 'special' purpose.
It's all a marketing game, and India low cost labor didn't bring any marketing to the table, but the entire damn world is doing marketing for the LLM bros.
AI wins, no contest (Score:3)
In fact, anything is better than Indian code. Their culture is "do fuck all, bare minimum, dissipate the blame for failure, cheat, lie and weasel as much as require. Then, charge millions.
Reliably incompetent.
SO, their Infinite Monkey Theory test has failed? (Score:2)
And why would they admit this just to dis AI? I guess it's their main competitor. Certainly says something for their future.
Out with old, in with the new (Score:1)
Old system was write code, tweak it to fix bugs or add new features/adjustments. This is why AI code needs to be well documented/readable/understandable/well structured to make changes to it. It is none of these for foreseeable future.
New system is, generate what you need from a very detailed description. Test like crazy. It either does exactly what you want. Or you add description and generate it ALL over again from scratch. You need a new feature added? REDO THE DESCRIPTION. Yes it will generate 99% of th
A difference of kind or of degree (Score:1)
Is AI just a better challenger, or is it a new type of challenger? I think two things are both true. It is a new and perhaps better challenger, and in some ways it is no different. The core challenge remains: turn ideas into software reality. Coding is a small part of that, and while AI may do a better and better job of handling pieces than most India firms, it falls flat in complex, novel, or uncertain situations- which describes pretty much every corporate IT project.
Re: (Score:2)
Ii think AI is a new type of challenger because it scales better than we do. There's limits to what people can accomplish and there's limits to intelligence but self replicating individual sentient machines may very well redefine what we think of as evolution.
However, it's not AI per se I'm concerned with, it's AI empowered upper class people who use all of our technology to further their domination over the rest of us.
Re: (Score:1)
> self replicating individual sentient machines
And here we see the problem with what we are talking about. We really really need to stop calling what we have now "Artificial Intelligence" , because it isn't. What we have is "Large Language Models", "Deep Learning" and what it delivers is what I call "Artificial Skills". There is nothing sentient about any of the systems that are currently being delivered. We don't have artificial intelligence, which requires what the current shysters call "AGI" to disguise the fact that they haven't actually delivered AI
Re: (Score:2)
> The interesting model, though, is driving. Most of us think that this has been a complete failure. Musk set out to do it and failed, like many of his other enterprises. What we missed is that in fact there is a company that has delivered "full self driving" [youtube.com] by limiting the problem so it doesn't need intelligence.
There are at least two fully autonomous robotaxi companies operating in San Francisco. Waymo, in particular, has been wildly successful and is winning business away from the likes of Lyft and Uber. It will even give you a ride to the airport now.
Re: (Score:2)
Are we talking about the same Waymo that just had to admit their cars are very often remotely driven by Philipinos?
Re: (Score:2)
> Are we talking about the same Waymo that just had to admit their cars are very often remotely driven by Philipinos?
Sure, but that's actually a brilliant solution and we should respect it. The various robo-taxi companies have a relatively good safety record (compare with steam engines before the standardization of boilers which used to blow up regularly killing whoever was standing next to them), they manage to have more than 70%, and for some companies in many situations more than 95% of their distance driven by artificial intelligence and, by using humans where needed they deliver a service which is about as good as th