there is no end of jobs better done by ai, but it will be slow and messy as usual. however, I recall back in the day, looking at a product line of appliances engineered back in the fifties. It still took tewenty more years to replace them.
an engineer freind took over a meter company over twenty years ago. The hardware used an 8 bit chip, but was good enough. anotherr twenty years before better hardware used.
my point is that good enough matters and change after that remains hard and risky.,
Change all Jobs? Change is Hard and Risky
February 28, 2026 by Brian Wang
https://www.nextbigfuture.com/2026/02/change-all-jobs-change-is-hard-and-risky.html#more-208826
IBM losing 13% because Anthropic did some initial COBOL analysis is absurd. It was also an absurd part of the $1 trillion stock losses from the boogeyman story about AI job losses.
1. IBM only makes a tiny fraction from the COBOL business
2. BS the banks etc… will change the cobol running the ATMs etc… when they have not done it for 60+ years. They were scared to change before and they are scared to change now. They did not know what is in all of the undocumented code and even if AI tells them how long will it take to test and verify. Maybe each big bank can save a few hundred million per year if they change it. Or they can focus on making some new crypto-AI apps and services to make a few billion per year on new products.
Screenshot
Screenshot“Modernizing COBOL has been a technically solved problem for a while,” Matt Brasier, analyst at Gartner, told VentureBeat. “The real problem is that the costs of modernization are high and the ROI is low.” Senior data and infrastructure engineers will spend the next few weeks fielding questions from executives who saw the headlines and assumed the hard problem just got solved. It did not.
“It’s COBOL, but there are numerous applications tied to it,” Joshi said. “It’s not like you transform millions of lines and somehow you are ready to go to cloud. It’s a massive risk assessment, dependencies and all those things.” Everything other system is connected to the existing systems. The hard parts are extracting institutional knowledge, reworking processes and controls, change management, and containing operational risk in systems that cannot break. AI can compress the “analysis and translation” work, but it does not eliminate the governance and accountability burden.
Growing is Far More Rewarding
Full transformations often yield 0-20% net gains (or losses) after failures. New ventures (or add-ons) in boom eras delivered 5-1000x scaling. If growing is better and more rewarding then jobs are not killed. Productivity increases. The pie is expanded.
When Ford originally grew its car company. The employees were paid more to create more customers for the new industry.
US retail ecommerce went from ~$27B (2000) to $1.2T+ (2024). Global ~$6.3T (2025 est., +8.8% YoY). ADDED channels drove 50%+ of some retail growth. They added and integrated with bricks and mortar stores.
Risks and Failure Rates in Extreme Change Management
Large-scale change initiatives (restructuring, full digital overhauls, or change all jobs reengineering) fail at rates of 60-70%, a figure consistent across decades of studies.
70% Failure Rate for Change Programs
Harvard Business Review (2000, still widely cited) finds ~70% failure rate for change programs.
McKinsey traditionally 70% fail due to employee resistance and lack of management support. Even today, radical reinvention is tough.
Maybe better practices can flip odds to 70-80% success in prepared organizations [This is called consultant fantasy where an executive is being sold by a consultant company that will get huge money to try to implement the change].
Errida & Lotfi (2021) meta-analysis and others finds one breakdown showed 50% outright failures, 16% mixed, 34% successes.
Result of Failed Change Autopsies
Employee resistance ~37-70% of failures.
Poor communication and leadership. 25% of leaders unable to execute the big change.
Change fatigue when employees face ~10 planned changes/year
Without full process mapping and verification, disruptions cascade (knowledge loss, supply chain breaks, customer churn). Not knowing you did not know how important things really worked.
Failed transformations destroy value, erode trust, and can lead to bankruptcy.
Even vastly more efficient new models don’t automatically kill incumbents due to switching costs, brand inertia, regulations, customer habits, and legacy moats.
During the 1990s-2000s internet boom and later cloud era, outsourcing isolated non-core departments (BPO/IT) delivered reliable cost and efficiency gains without full overhaul, while adding new channels like ecommerce or SaaS drove the biggest revenue upside. Full core gutting/switching was rarer and riskier.
That is Old, Why is It Still Around?
Newspapers have been around for 300+ years in US (colonial era).
Daily circulation peaked ~62.8M (1987), ~55.8M (2000) → ~21M combined print/digital (2022, -8-10% YoY).
Radio starts in 1920s, Survived TV and then later internet/podcasts. Weekly reach stable at ~80-90% of peak.
Walmart versus older retail and Sears. Niches in older retail survive and it is taking decades for Sears to go away. Sears could have adjusted and competed far better.
Amazon versus Walmart and Barnes & Noble. Physical book stores got hit hard but are now coming back. Walmart continues to grow.

.webp)


