Businesses are fixated on choosing the right AI models, but this focus is a distraction from the foundational data work required for success.
Gary Nakanelua, Managing Director of Product & Innovation at Blueprint Technologies, argues that the biggest threat to AI adoption is a failure of experience, where technical outputs are unusable for business teams.
True AI readiness is built on four pillars: data governance, accessibility, cleanliness, and consistency, not on the sophistication of the tech stack.
In the era of AI-noise, it’s often easy for C-Suite executives to focus on the “wrong parts” of AI. Believing the right tool will automatically guarantee a return on investment, many leaders fall into the trap of fixation on choosing the right models, newest protocols, or latest tech stacks. But what this technical arms race distracts from is a more fundamental truth: the long-term value of AI hinges more on the unglamorous work of data preparation than the model itself.
As the Managing Director of Product & Innovation at Blueprint Technologies, a data intelligence firm, Gary Nakanelua has a clear view of why so many expensive AI initiatives fail to deliver. He believes that to unlock sustainable ROI, organizations must first confront a simple, human-centric problem.
The biggest threat: For Nakanelua, the greatest challenge with AI adoption is a failure of experience. “Organizations tend to get so wrapped up in AI models and frameworks that they forget about the human on the other side,” Nakanelua says. “But why invest millions of dollars to generate outputs that business users often don’t understand? It’s just as important, if not more, than the tech itself.”
Nakanelua urges leaders to shift their focus from models to the quality of their data to bridge this gap between technical output and business utility. True AI readiness, he explains, is built on four pillars.
The four pillars: “Governance, accessibility, cleanliness, consistency—here is where discipline matters more than the algorithm,” Nakanelua says. “For cleanliness, consider relevance: focus on what matters and skip the rest. For consistency, reduce variation. But data accessibility can be a political challenge.” After years of cultivation, some departments can become “protective of datasets.”
The antidote is an experience so intuitive it becomes invisible. Here, Nakanelua points to an example that reset expectations for what a human-computer interface could be.
The ChatGPT lesson: “ChatGPT fundamentally changed our thinking about the user experience. Suddenly, non-technical users could use what was historically a very technical tool. Most could understand the interaction layer immediately because it was familiar, like sending a text message.” Now, organizations need to think about their use of AI in the same way.
For most, the biggest challenge is moving beyond experimentation. As a cautionary tale, Nakanelua advises companies to avoid pursuing projects with no real-world value. Instead, he suggests, start with a solid use case.
“AI can get very expensive, very fast,” Nakanelua says. “Often, nobody stops to ask, ‘Do we really need this? What problem does it solve? Who will find value in that? Meanwhile, organizations invest millions into initiatives that are so lengthy that they forget why in the first place. Eventually projects die, and organizations conclude, ‘AI is a waste.’ Many decide to wait until things are ‘more mature’ instead.” No longer a matter of tech maturity, successful AI is a matter of use case alignment and a solid understanding of ROI.
