"But if the business is chaotic, the software will never catch up."
In my fairly long but admittedly somewhat limited experience (I've been in the internal IT organization of an insurance/financial services company in several roles for about 35 years) this is the ultimate problem faced by all internal IT organizations. There should be no 'if' in this statement. From the IT organization's viewpoint, business is always chaotic. Also, in my experience as a consumer of package software, even software development companies don't appear to run their business under your idealized assumptions about change control and product development.
Also, again from the perspective of somebody who has been in the trenches of "slap our logo on generic software" projects, that's pretty much a fantasy. The fundamental problem is that every piece of software is designed with specific use cases (i.e. sales pitches) in mind, and it is highly unlikely that those use cases will align with all the existing business processes of a company. Sometimes this is just a matter of executive ego or 'we always did it this way' but you are either going to have to customize the software or you are going to have to change the business process flow to accommodate the way the software functions, and either approach generates its own set of problems.
"And while I know nothing about Volkswagen’s business processes, I would bet—based on their problems with software—that those processes are more chaotic than at other auto firms."
I was questioning something like this before I got to this sentence. First, are their software problems bigger than others? How does one measure that? Is Microsoft's software process chaotic? If not, did it used to be and at what point did they get beyond that? Is Tesla's car making more chaotic than other car companies? If so, what does that tell us about SpaceX's processes? My last thought was that it's unlikely you become one of the two biggest auto manufacturers with chaotic car building processes.
Here's a great tell that your observations are correct: the person identified as CTO of Cariad in the WSJ article has a publicly accessible LinkedIn profile. Hers is not the resume of a software person.
But, in most instances Volkswagen doesn’t want to be a Microsoft.
Safety critical systems need RTOs (real time operating systems) that are 100% reliable and respond instantaneously to sensor feedback.
Do you want your automotive ABS system running on Windows? How about the airbags? Asking the question provides the answer.
The issue here for VW appears to be primarily related the infotainment stack, which is just a fancy gui to change the thermostat, music, navigation, etc. The obvious workaround for this is Apple CarPlay or Android Auto, but many auto manufacturers don’t want to cede control of this to these potential future competitors.
Interesting tale, for it was VW the first to learn to their surprise after reunification in the Trabant car plant in the East, that everything was made in-house down to the nuts and bolts, with the result that the finished product was worth less than what had gone into making it. Stick to what you know seems best policy.
"But if the business is chaotic, the software will never catch up."
In my fairly long but admittedly somewhat limited experience (I've been in the internal IT organization of an insurance/financial services company in several roles for about 35 years) this is the ultimate problem faced by all internal IT organizations. There should be no 'if' in this statement. From the IT organization's viewpoint, business is always chaotic. Also, in my experience as a consumer of package software, even software development companies don't appear to run their business under your idealized assumptions about change control and product development.
Also, again from the perspective of somebody who has been in the trenches of "slap our logo on generic software" projects, that's pretty much a fantasy. The fundamental problem is that every piece of software is designed with specific use cases (i.e. sales pitches) in mind, and it is highly unlikely that those use cases will align with all the existing business processes of a company. Sometimes this is just a matter of executive ego or 'we always did it this way' but you are either going to have to customize the software or you are going to have to change the business process flow to accommodate the way the software functions, and either approach generates its own set of problems.
"And while I know nothing about Volkswagen’s business processes, I would bet—based on their problems with software—that those processes are more chaotic than at other auto firms."
I was questioning something like this before I got to this sentence. First, are their software problems bigger than others? How does one measure that? Is Microsoft's software process chaotic? If not, did it used to be and at what point did they get beyond that? Is Tesla's car making more chaotic than other car companies? If so, what does that tell us about SpaceX's processes? My last thought was that it's unlikely you become one of the two biggest auto manufacturers with chaotic car building processes.
Here's a great tell that your observations are correct: the person identified as CTO of Cariad in the WSJ article has a publicly accessible LinkedIn profile. Hers is not the resume of a software person.
The exception is JPM, who is well regarded as a formidable technology company even by other tech companies.
But, in most instances Volkswagen doesn’t want to be a Microsoft.
Safety critical systems need RTOs (real time operating systems) that are 100% reliable and respond instantaneously to sensor feedback.
Do you want your automotive ABS system running on Windows? How about the airbags? Asking the question provides the answer.
The issue here for VW appears to be primarily related the infotainment stack, which is just a fancy gui to change the thermostat, music, navigation, etc. The obvious workaround for this is Apple CarPlay or Android Auto, but many auto manufacturers don’t want to cede control of this to these potential future competitors.
Interesting tale, for it was VW the first to learn to their surprise after reunification in the Trabant car plant in the East, that everything was made in-house down to the nuts and bolts, with the result that the finished product was worth less than what had gone into making it. Stick to what you know seems best policy.
“Maybe release a ‘beta’ version of a car to a few customers before you decide on the final specs.”
While we are at it, let’s also consider trying beta airplanes and beta nuclear plants.
I would potentially add beta viruses to the list, but I think this already got tried a few years ago with mixed results.
One wonders if the lessons in this post are applicable to corporate adoption of LLM systems. Charles Hughes Smith has a piece today on this topic and asks whether LLMs will generate profits or merely add new costs: https://www.zerohedge.com/technology/what-if-ai-only-cost-and-not-profit-bonanza. One gets a general impression from the hype that there may be some chaos in the non-tech companies who may be rushing into this space in pursuit of early adapter gains. https://www.investors.com/news/technology/17-non-tech-firms-making-big-bets-on-artificial-intelligence/
And GM is revamping their car IT infrastructure to collect more user data. (They will rely on Google to build it.) https://www.msn.com/en-us/news/technology/why-ford-thinks-gm-is-wrong-to-drop-apple-carplay-from-future-models/ar-AA1b9icM
One of Tesla's advantage during the chip shortage was being able to rewrite software to be able to use less powerful chips that were more available.