Why Our Current AI Approach Needs Reevaluation

In discussions about AI, we often emphasize a single metric: productivity. This has been the focus of technological advancements since the dawn of the modern tech era.

Reflecting on my early days as an external tech analyst, particularly around the Windows 95 launch, the promise was that it would significantly boost productivity, delivering a return on investment (ROI) within a year. Ironically, the first year saw so many issues that productivity initially declined.

AI’s ROI could be even more problematic, largely because our challenge today revolves around poor decision support, not a lack of productivity or performance.

At a Computex prep event last week, productivity was a constant theme. My concern remains that while we can drastically increase speeds, if we don’t also enhance the quality of decisions, we risk making irreversible errors at machine speeds.

In this discussion, I’ll also share my Product of the Week—an airline that made my recent trip to Taiwan exceptionally pleasant, especially compared to United, which I usually fly for international journeys. This contrast highlights why many non-U.S. airlines offer a significantly better experience.

Productivity vs. Quality

As a former IBM employee, I was part of a select group that went through the company’s executive training program, where the importance of quality was deeply ingrained in us.

One notable class was from the Society of Competitive Intelligence Professionals (SCIP), which emphasized speed vs. direction. The instructor highlighted that companies often prioritize speed but fail to define the right direction, leading to faster but misguided efforts.

During my time at IBM and Siemens as a competitive analyst, I frequently saw decision support recommendations ignored or opposed, leading to catastrophic failures. This was because executives preferred to appear correct rather than be correct. Eventually, my unit was disbanded because calling out failures based on sound advice embarrassed executives who ignored it.

As an external analyst, I noticed my guidance was more likely to be followed since executives did not view me as a career threat.

Executives have access to vast amounts of data, yet many still make poorly founded decisions with catastrophic results. AI should focus first on improving decision-making, and only then on productivity and performance. If we don’t ensure the decisions are correct, we’re likely to accelerate mistakes rather than successes.

Decision-Making Challenges

AI can help us make faster decisions in both personal and professional realms, but the quality of these decisions is degrading. If you look at Microsoft and Intel, key supporters of modern AI technology, you’ll notice they’ve made poor decisions that cost them significant leadership changes this century.

My friend Steve Ballmer, despite being highly intelligent and successful in certain areas like the Xbox, faced persistent bad decisions that led to critical failures and his eventual dismissal.

Similarly, John Akers at IBM was failed by those who prevented important information from reaching him. Although my efforts to resolve company issues were acknowledged, the barriers imposed by those wanting to maintain status and access led to his downfall.

Both CEOs were denied critical information by trusted individuals who prioritized status over the company’s success.

The Dual Problem with AI Decision-Making

Firstly, AI results, though impressive, are often inaccurate or incomplete. The Wall Street Journal recently evaluated top AI products, finding that Google’s Gemini and Microsoft’s Copilot were often of low quality, despite their widespread use.

Secondly, even if AI’s accuracy improves, executives may not use it, trusting their instincts over any system’s advice. This bad behavior is reinforced by current AI quality issues, hindering the technology’s potential to enhance business and government success.

Concluding Thoughts

Currently, our desire for speed (productivity, performance) is less important than the need for trustworthy and reliable technology. Even if we address AI’s accuracy problems, Argumentative Theory suggests internal advice will still be seen as a threat to jobs, status, and image.

This concern has some merit; if decisions are AI-driven, people might fear being seen as redundant.

We should shift our focus from AI-driven productivity to high-quality decision support to avoid being overwhelmed by rapid, poor decisions. Training people to accept valid advice and rewarding effective AI use can help us advance responsibly at machine speeds without risking job security.

AI has the potential to improve our world, but only if it delivers quality results and we’re committed to using those results in our decisions.

Starlux Airlines

I’ve almost stopped flying with United Airlines due to poor experiences, from long delays in remote airports to first-class tickets turning into coach seats due to operational failures.

My experiences with non-U.S. airlines have been much better. On a recent trip to Computex, I flew with Starlux, a Taiwanese airline. The experience was far superior.

In United’s business class, I often feel like an inconvenience. Starlux, on the other hand, prioritized my comfort. They went out of their way to provide a special meal on request and resolved my Wi-Fi issues promptly and with genuine concern for my experience.

Traveling frequently for work, I’ve grown to dread it, which is unfortunate since I loved flying as a child. Starlux helped me regain some of that enthusiasm, making my 13-hour flight enjoyable and leaving me eager for my return journey.

I’ve noticed this level of excellence with other international carriers like Singapore Airlines and Emirates, among others. So, Starlux Airlines is my Product of the Week.