SAP stops most travel and hiring because of AI's soaring cost | Hacker News Reader

SAP stops most travel and hiring because of AI's soaring cost

404media.co

Having worked at some giant firms that heavily depend on SAP, I would think their position as "system of record for procurement/production" would get stronger with AI.

e.g. AI + a good database of what your company does seems like an excellent combination.

Or AI can be a threat: If AI can spit out good enough special purpose tools a lot of the SAP adaption business goes away. If AI then also assists in data migration the moat goes away.

EAI has existed for twenty years...Tibco, webMethods, IBM AppConnect...SAP is still what it is. How is AI going to make a dent on this?

I've been impressed so far with Snowflake's offerings of "AI + a good database"

The same argument can be made about snowflakes and other providers that handle the data layer. If they can leverage AI in a safe contained environment they would become the clear winner

LOL

Although they may internally use LLM, I wanna say said they were not going to allow their core offering to be integrated with LLM

Although maybe they reversed course?

Would a technology famous for hallucinating data often enough, and in many context unable to guarantee it won't assume and hallucinate when it feels "confident enough" despite a robust system prompt, be really considered for a "system of record for procurement/production" usage?

(yes yes RAG exists)

In the ERP space, AI is being used mostly for repetitive tasks, documentation and forecasting-type activities or where a non-deterministic answer might be useful. Scanning in documents and using OCR to create an order, asking what page I use to do X, historical demand modelling and constraint scheduling. Also have seen it used to do order flow type stuff like suggesting products to sell if they buy ABC. I'm not really seeing it being used to do heavy analysis and they've generally been careful not to put it in places where a hallucination could result in a bad business decision.

> Scanning in documents and using OCR to create an order,

I remember this being done more than 10 years ago.

The other uses cases you describe also were being done. Of course with AI maybe it’s become easier and cheaper?

Yes but it generally expected documents to be in a certain layout and the user had to fill in a lot of the gaps. It's more able to identify elements on an incoming PO, match a supplier invoice, or create a receipt based on a packing list with inference.

EDIT: The other activities, yes I think it made it cheaper and easier. They both required human intervention to make an accurate forecast or execution plan. Forecasting requires humans to do things like apply smoothing factors, identify abormal demand over huge datasets, or identify seasonal demand cycles, etc. Constraint scheduling requires a human to intervene in dispatching and resource allocation as well. AI can do 95% of the work with a proposal, or even control the process. Supposedly much better than historical algorithms. I haven't implemented this part of it, but I've seen it done.

The documentation part is much more robust. You can ask it complex questions, like "How do I setup a phantom blow-through part in a MS level 1 BoM?" and it walks you through the entire thing. With the sales suggestions, it can identify things like "The customer is buying spaghetti and pancetta" AI: "They are probably making carbonara, how about some garlic bread?"

Take your favorite OCR software.

Then take a photo of a document and throw it into ChatGPT, or Gemini, along with the word "Transcribe".

ChatGPT wins, by a landslide. BUT AI's advantage doesn't stop there.

"Take this picture of some idiot filling in form 49, I've also attached the PDF, fill in the PDF fields, provide a database record according to the schema attached and flag if there are any obvious problems with the entries".

That works too. "Produce a latex document of this kid's math homework and flag any problems" - works. "Produce a MS word document of this letter" - works. "Read this bill and produce a JSON version following the schema from this example" ... and so on and so forth.

More than that, it is starting to work pretty well with Gemma 31B local model (will still do 30 document analyses on cpu only, at Q4 on a DDR4 or higher machine. Yes it's mostly memory speed that matters) at this point. I mean, that's GPT 5.0 or so quality (95% correct with the occasional problem), but hey, Qwen 3.8 27B may be coming out next week ...

30 document analysis in parallel

Actually 30 is sequential on an hosted Xeon server. No GPU, about 2 minutes for a ~2000 token answer. I should try parallel since you're quite right, I should be able to do faster that way.

> forecasting-type activities

AI/ML yes, LLM's no (unless someone is naively using LLM's for forecasting)

The big risk for these platforms is that AI reduces the cost to switch to another ERP. IF AI can map all of the data between two different competitors, the whole business model is finished

That is not what ERPs are used for.

Would love to hear how AI is going to help here. This is like a young engineer thinking the CDC machine they just bought, is going to make Airbus and Boeing obsolete...

It’s well known that the moat of system of record is how difficult it is to extract your data into another system.

ERPs are often used as a lot more than a system of record: they’re directly engaged with as the LOB software point-of-contact by many expert employees, and function as a sort of encoding of organizational behavior.

They’re a lot more than a data store or system whose primary value is ingesting info and producing reports.

For a lot of business, replacing the ERP is on par with replacing Excel. It’s not about getting the data into another system; it’s about all of the complexity and expertise required to use it.

> It’s well known that the moat of system of record is how difficult it is to extract your data into another system.

I think that is one aspect but not the biggest issue.

The specific shape of data and shape of functionality gets built around for decades due to incompleteness of any system, expanding dependencies and changing business needs.

Decades of building around the system results in a mountain of work when trying to replace the system, typically a larger body of work than implementing the base system itself.

most of this complexity is because 1000 different types of businesses are jammed into the same database schema and ERP. Its all non sense

Maybe but I work for a company that integrates with a lot of ERPs and I don't know why but the rando IT guy who can't get themselves out of a paper bag is still a thing. You'd think AI would help them but man ... doesn't seem like it.

Telling that SAP now exfiltrates all your company data to the U.S. AI cartel:

https://www.sap.com/products/ai-platform.html

"Daddy, vodka got more expensive now, does it mean you will be drinking less?"

"No, kids, it means you will be eating less."

I sure am glad my employer isn’t a drunk daddy

So will AI be what finally kills SAP, making world a little bit better? ;-)

Seems like they're doing fine: https://en.wikipedia.org/wiki/SAP#mwAeo

Does it? Those are big numbers, but it looks like in 2024 they sold €4b of assets to juice income for 2025.

They didn’t need to? They wouldn’t have a particularly bad year.

So they are using AI to run the company, and AI allocated all the money to AI?

it may just be actually incredible what they are doing. on the other hand it could also be absolutely insane...

Sounds like a paperclip maximizer to me.

Seriously, why the duck aren't they running their own AI systems in house???

Like their systems aren't shitty enough.

TL;DR:

“Under the policy, only Al-related hiring, customer-facing travel and trips directly linked to the company's Al initiatives or mission-critical employee Al training are exempt, reflecting SAP's focus on "be disciplined in how we spend" while expanding its Al capabilities.”

no-paywall: https://news.ycombinator.com/item?id=49222851

I would pay to understand what mission critical employee Ai training is that can't be provided in other ways

Executive meet and greet

Or sales/FDE deployment

In other words, AI is now driving their budget decisions.

Wait, first they stopped hiring because AI is going to replace everybody. Now they stopped hiring because AI, which was going to replace everybody, costs too much.

You can’t make this stuff up.

Sometimes buying a tool to make 10 people 15% more effective is a better choice than hiring 1 person.

This is less of an issue than the headline suggests. I worked for a publicly traded US company remotely. The company was based between Ft. Lauderdale and West Palm Beach Florida. It was a roll of the dice each spring if the CFO would pull the trigger on a travel block for about 6 weeks around spring break when flights and hotels were most expensive. Approvals were available for critical projects. It was only ever about not exceeding a quarterly budget number; nothing else.

Anecdotally, I know several companies that have given the directive of "Use more AI instead of hiring more people". I wonder whether this is part of the story here too?

If you believe that AI does add to productivity gains, say equivalent to adding 10-20% employees, would you do anything differently than SAP?

If AI was actually making you more profitable, you wouldn't need to be cost cutting to avoid bankruptcy.

I'm not sure who is getting bankrupt?

No one, because they cut costs to avoid it.

These are enough gains that you don't need to believe, they should be measurable

Not really. A company can work and do things without having to run RCTs to prove everything. The market itself is the RCT.

If you doubt that you have gained something as high as 20% gains, it probably just means you got nothing.

And also, individual productivity isn't the same as organisation productivity, especially at such a large company.

  > gained something as high as 20% gains

all other things being held constant, i suppose it should be as easy as +20% profit rate? tho im not an executive so what do i know...

SAP also made some big acquisitions that cost a lot of cash..

teach everyone they live in a lobster bucket, and suddenly everyones a lobster.

This tells me we’re getting close to “peak bubble.” We saw these same sort of stupid headlines just before the dot-com bubble popped, too. When companies are distorting regular business to do the new thing, they’ve gone past the point of reasonableness on the new thing and we’re ready for a snap-back.

It will be interesting to see if SAP can turn into the basis for companies that use AI heavily or if companies that use AI heavily will prefer to replace SAP.

Who are the contenders that start fresh with focus on AI?

One that comes to mind is IONOS. They seem to be in the process of building something like a "SAP in the times of AI" from the ground up. Starting with the demands of SMEs. It looks like an interesting approach to me: Having a "knowledge base" for each customer, initially filled with the data from the website. Then let the customers add more data. Then offer agents that do work for the customers. Answer calls and emails and do marketing for example.

I have in mind what I think is a very strong framework to allow companies to replace legacy ERPs quickly with AI help. I guess I should get on with it and publish something!

> It will be interesting to see if SAP can turn into the basis for companies that use AI heavily or if companies that use AI heavily will prefer to replace SAP.

Considering the human and institutional complexity of implementing ERP, CRM, whatever the hell Concur is supposed to be, etc. I'd bet on most organisations implementing AI around and with those systems instead of replacing them with something vibecoded / "AI native" vendors.

Reply on news.ycombinator.com