You’ll notice an odd contradiction if you walk into practically any corporate office right now. Usually, there’s a slide show called “AI Transformation 2026.” The marketing department is conducting a pilot. Perhaps a customer service chatbot. a hastily hired data science team. However, when you ask those who actually use these tools if the company’s operations have changed fundamentally, they almost always hesitantly respond, “not really.” The Harvard Business Review recently attempted to address that hesitation, and the story is more damning than it initially appears.
In an article published in early April, Thomas C. Redman made a straightforward point that is quietly making the rounds among executives: before considering what AI can do for your business, consider whether it is truly prepared for AI. His five guiding principles—acting on facts, customer centricity, process focus, continuous improvement, and quality through people—are not novel. For many years, they have been the silent foundation of total quality management. All of them would have been familiar to Deming. In a way, that’s the point. Despite its novelty, AI does not eliminate these foundations. Any preexisting foundation is amplified.
| Snapshot | Details |
|---|---|
| Publication | Harvard Business Review |
| Article Title | “To Succeed with AI, You’ve Got to Nail the Basics” |
| Published | April 6, 2026 |
| Author | Thomas C. Redman, President of Data Quality Solutions |
| Core Argument | Quality fundamentals matter more than AI adoption speed |
| Five Timeless Principles | Customer centricity, process focus, act on facts, continuous improvement, quality through people |
| Reported Failure Rate of AI Projects | ~85% |
| Common Root Cause | Weak data, broken processes, poor governance |
| Referenced Institutions | Harvard Business School |
| Related HBS Research | Ashley Whillans — executive interviews across global enterprises |
| Key Quote From Author’s Recent Book | “People and Data: Uniting to Transform Your Business” (Kogan Page, 2023) |
| Data Quality Reference | American Society for Quality (ASQ) |
Redman’s framing succeeds because it conveys an uneasy feeling that most executives are aware of but choose not to express. Many businesses are adding AI to already flawed processes. Consciously or unconsciously, it is expected that a clever model will somehow overcome years of disorganized data, ambiguous ownership, and inconsistent customer definitions. It hardly ever does. Actually, it’s quite evident from the literature that about 85% of AI projects fall short of their goals. Boardrooms ought to be alarmed by that figure, and in some cases they are. However, in most cases, the solution is to purchase additional AI tools rather than to address the underlying issues.
The speed at which this thesis has gained traction among practitioners who have actually served in the trenches is intriguing. After reading the article, HCLTech leader Mohanraj Chinnapaiyan posted on LinkedIn that the root cause was nearly always the same after 20 years of being placed in “red” accounts—the underperforming programs that no one wanted. procedures not adhered to. Governance is not upheld. Data is not trustworthy. Seldom tech, seldom complexity. “AI doesn’t fix broken systems,” he wrote. The phrase “It amplifies them.” has been making the rounds.
I believe that part of the reason this advice is overlooked is that it has an almost boring quality. Presenting a new agentic workflow at a leadership offsite generates more energy than discussing data quality. While subtly misaligning their incentives, most businesses pay lip service to the concept of customer centricity. Patience is necessary for continuous improvement, and since OpenAI used each quarterly earnings call as an AI litmus test, patience has been scarce. Even when they are the ones creating something long-lasting, executives who adopt a cautious approach run the risk of being called laggards.
We seem to be at a familiar stage in the hype cycle as we watch this develop. When businesses finally realized in 2015 that “lift and shift” to AWS did not automatically modernize their stack, the cloud experienced its own version of this. Billions of dollars were spent, there was little change, and eventually there was a quieter stage of actual architectural work. AI appears ready for something comparable. In three years, the businesses that currently perform the unglamorous tasks—cleaning their customer data, streamlining their procedures, and properly training employees—will likely appear as the quiet winners. Pilots will continue to be run by the others.
It’s difficult to ignore how little of this is discussed at the largest AI conferences. Seldom do keynote speakers say, “Actually, your data warehouse is a mess, and until you fix it, none of this will work.” The kind of truth that doesn’t sell tickets is like that. However, at this moment, it may be the most crucial thing anyone could say to a CEO. Harvard just expressed it in a courteous manner, as is customary. The more difficult question is whether anyone pays attention, which will likely determine which businesses emerge from this decade with something tangible to show for all the money spent on AI.