I once said to a group of 500 executives, “Customer first, data second, everything else after.”

In most cases, the fancy new AI model is rarely the problem.

In the majority of AI initiatives that stall, underperform, or quietly get defunded, the root cause is the same. The data was incomplete. The data was inconsistent. The data existed in silos that nobody had properly connected. The data reflected assumptions baked in years ago that nobody had examined since. Or the data simply was not there.

The Model is the Easy Part

AI is the part that gets the demo. It is the part with the name that everyone wants to see. The model is visible in a way that data infrastructure will never be.

But a powerful model fed poor data produces poor outputs, reliably and at scale. It does not produce poor outputs occasionally, or in edge cases, or only when the prompt is badly written. It produces them systematically, because the model is doing exactly what it was designed to do: find patterns, generate outputs, and reflect back whatever the underlying data contains.

Garbage in, garbage out was a cliche before most of today’s AI workforce was born. It is still true. What has changed is the speed at which the garbage gets amplified.

What Companies Actually Mean When They Say AI Strategy

Ask any leader what they believe is an AI strategy, and you will hear a version of the same answer. Updated models, internal pilots, exploring use cases, training teams, and learn what works.

Almost none of them lead with data.

They treat data as a precondition, something to check off before the real work starts. Clean the data, then do AI. In practice, that checklist item is where most initiatives quietly die. The data is messier than expected. The cleaning takes longer than budgeted. The governance questions around who owns which data and who is allowed to use it for what purpose turn out to have no clean answers. The pilot gets delayed. The executive sponsor moves on. The initiative loses momentum.

The organizations that avoid this failure mode are the ones that treat data strategy and AI strategy as the same conversation from the beginning. They are asking, before they ever select a model, what data we have, what data we need, what our data actually represents, and what gaps exist between those three.

The Switching Trap

A company adopts a model. The outputs disappoint. Rather than examining the data feeding the model, the team concludes the model was the wrong choice. They switch to a newer model. The outputs disappoint again, in slightly different ways. They switch again. They add RAG.

Each switch carries a real cost. Engineering time to integrate a new model. Retraining the team. Rebuilding the prompts. Renegotiating the contracts. And at the end of the cycle, the organization is no further along than it was, because the thing that was actually limiting performance was never the model. It was the data the model had to work with.

This is the switching trap. It is seductive because switching feels like progress. It feels like something was learned, and the team is advancing. You are doing something, making a decision, responding to the problem. The model is also a more comfortable target than the data, because the data problem implicates internal decisions, internal ownership, and internal history. The model came from outside. Blaming it carries less organizational friction.

A Data Strategy Has Different Questions

Shifting from an AI strategy framing to a data strategy framing changes the questions a leadership team needs to answer.

Instead of “which model should we use,” the primary question becomes “what do we actually know about our customers, our operations, and our outcomes, and how well is that knowledge captured in a form a system can use?”

Instead of “what use cases can we automate,” the question becomes “where is our data strong enough to support reliable inference, and where are the gaps that would make automation dangerous or misleading?”

Instead of “how do we move faster,” the question becomes “how do we build data assets that compound over time, so that the systems we build this year are meaningfully better than the ones we built last year?”

Does our data actually paint a story about our consumer?

Does our data actually communicate seamlessly across teams?

Those are harder questions. They require different stakeholders. They surface uncomfortable answers. They also tend to produce AI initiatives that actually work.

The Compounding Advantage

The most important thing about treating AI strategy as a data strategy is what it enables over time.

Data compounds. A company that has spent three years building clean, well-governed, well-labeled data assets can take any capable model and produce results that a company with equivalent technical talent but worse data simply cannot match. The model is a commodity. The data is the differentiator.

This is already visible in the market. The organizations achieving the most durable results with AI are rarely the ones with the most sophisticated models or the largest AI teams. They are the ones with the longest history of treating data as a strategic asset. They invested in data governance before it was fashionable. They built pipelines, documentation, and ownership structures that seemed like overhead at the time. Now those investments are compounding.

The companies still benchmarking models are competing on a dimension that is rapidly equalizing. The companies investing in data are building an advantage that widens every year.

What to do With This

If your organization has an AI strategy that centers on model selection, vendor evaluation, or use case identification, it is worth asking a harder question before the next planning cycle.

What is our data strategy?

If the answer is vague or the question is redirected back to the model conversation, that is the signal. The AI initiative is sitting on an unstable foundation, and eventually that instability will show up in the results.

The model you pick matters. It is just the last decision you should be making, not the first.

Customer first. Data second. Everything else falls in place after. And in most cases, the AI portion is last.

---

If you find this content valuable, please share it with your network.

Follow me for daily insights.

Book me to speak at your next event.

Chris Hood is an AI strategist and author of the #1 Amazon Best Seller Infailible and Customer Transformation, and has been recognized as one of the Top 30 Global Gurus for Customer Experience. His latest book, Unmapping Customer Journeys, will be published in 2026.