By Pat Dixon, PE, PMP
President of DPAS, (DPAS-INC.com)
The 4th industrial era makes industry more dependent on software than ever before. From the valve in your pipe to the financial system at the top of the corporation, the 4th industrial era is the potential for software to connect it all.
That is why the December 2025 article in IEEE Spectrum "The Trillion-Dollar Cost of IT's Willful Ignorance" should alarm everyone.
The author, Robert N Charette, has been writing about software failures for 20 years. His article says it is not getting any better. While global IT spending has tripled in that time "the business and societal costs of failure continue to grow as software proliferates". Charette pores cold water on the idea that artificial intelligence (AI) will fix this.
In fact, AI represents one of the sources of failure. We are in an AI boom, which may be followed by a bust. AI is certainly here to stay, but the failures come from the wrong goals and unreasonable expectations. If your goal is to use AI to cut your workforce, you may not understand what AI does and what your business needs. This error in target selection comes from listening to consultants instead of operators. Operators know why you aren't performing at your peak because they see it all day every day. If a consultant has never walked through your production and listened to operators, how do they know their AI promises will provide any benefit? If AI is meant to empower your engineers and operators to address your problems, it is far more likely to succeed. The notion that our facilities can replace humans with AI is very far in the future if at all, as addressed in an ISA article "What are the Prospects for an Autonomous Process Manufacturing Unit? (https://blog.isa.org/ask-the-automation-pros-what-are-the-prospects-for-an-autonomous-process-manufacturing-unit). Charette says unrealistic or unarticulated project goals are a prime driver for software failures. Cool technology and buzzwords without the right target and careful aim is a recipe for disaster.
Another source of failure is the assumption that better technology and techniques make software easy. It is absolutely true that AI can dramatically reduce software effort. It is untrue that this is a shortcut to quality. Many lines of code can be produced quickly, but matching it to the specific demands of our industrial operation and fixing bugs does not involve shortcuts. Some software development practices, such as Agile, can be very helpful. However, Agile is an odd fit for industrial processes. If all you need is a compiler, you can use Agile to develop an app efficiently. In our world we need motors, valves, power, steel and concrete, and many other dependencies. I have used hybrid Agile techniques to pull off some very challenging projects, but our automation projects require careful coordination with dependencies that don't exist in the IT world. Even with Agile, you don't have a guarantee. Charette says that some reports claim Agile projects have a 65% failure rate, which is actually an improvement over the traditional waterfall approach.
There are also no shortcuts to testing. If an app crashes, it is disappointing. If valves and motors don't do what they should, it can be a disaster. I have been involved in testing some of the leading control systems on the market, and of course there is factory acceptance testing before an automation project goes to the facility. There are some that think their bottom line will look better when they can downsize the quality control effort, but this is very short sighted. You might not have business that is sustainable in the long term if you leak bugs and have to respond to emergencies with software hotfixes. Experienced professionals know that testing carefully and early makes bug fixes much less costly than late surprises.
In 2005 Charette said "The biggest tragedy is that software failure is for the most part predictable and avoidable. Unfortunately, most organizations don't see preventing failure as an urgent matter, even though that risks harming the organization and maybe even destroying it."
In 2025, Charette says we are not learning.
