The conversation about getting your data in order has been happening inside companies for years. It's not new. Back in 2018, the Harvard Business Review published a widely circulated piece called "Data Is the New Oil," drawn from the book Prediction Machines by economists Ajay Agrawal, Joshua Gans, and Avi Goldfarb. The metaphor resonated. It showed up in board decks, strategy sessions, and vendor pitches. And it's still showing up today, most recently in Super Bowl ads from major enterprise platforms positioning data as the foundation of everything AI.
The emphasis on data isn't misplaced. But the conclusion many organizations have drawn from it is. For a lot of companies, "we need to get our data in order" has quietly become the reason they haven't started doing anything meaningful with AI.
A Worthy Pursuit Without a Finish Line
Every organization should invest in its data estate. Good governance, clean pipelines, consistent formatting, clear lineage: all of it matters. That's not the debate.
The problem is treating data readiness as a gate. Getting your data in order is not a project with a deadline. It's an ongoing discipline, more like maintaining a house than building one. There will always be another system to integrate, another data quality issue to resolve, another governance gap to close. If the plan is to wait until the data estate is "ready" before pursuing AI, the plan is to wait forever.
And that's exactly what's happening at a lot of companies right now.
Progress Can't Wait for Perfection
The organizations making real headway on AI have figured out something important: data improvement and AI adoption are parallel tracks, not sequential ones. You work on your data foundations while simultaneously putting AI to work on the problems you can solve today.
There is a wide range of AI applications that don't require a pristine data estate. AI can summarize meeting transcripts, draft customer communications from information already sitting in your CRM, review contracts, generate first drafts of reports, and help employees work faster on routine tasks. These are real, practical use cases that create value now, with the data you already have.
The more advanced applications, things like predictive analytics, cross-functional AI agents, and automated decision systems, do benefit from cleaner, more structured data. But you build toward those. You don't sit idle until they're possible.
The Cost Nobody Talks About
There's a cost to waiting that rarely shows up in data strategy conversations: lost organizational learning. AI adoption isn't just a technology challenge. It's a people challenge. Your teams need time to learn how to work alongside AI. Your leaders need to develop instincts about where AI creates value and where it doesn't. Your workflows need to be tested, refined, and rebuilt around new capabilities.
None of that learning happens while you're waiting for the data estate to be perfect. Every month spent exclusively in "data readiness mode" is a month your competitors are building the muscle memory and institutional knowledge that compounds over time.
The companies that will be furthest ahead in two years aren't necessarily the ones with the cleanest data today. They're the ones that started moving, learned from the messiness, and kept improving on both fronts.
Two Tracks at the Same Time
If your organization has been stuck in the "fix our data first" loop, here's a more productive framing. Run two tracks simultaneously.
The first track is your data estate. Keep investing in governance, quality, and infrastructure. This work is real and it matters. Don't stop doing it.
The second track is AI adoption. Identify the use cases that work with the data you have right now. Empower your employees with AI tools that make them more productive today. Find AI-powered solutions to problems that have been sitting on the backlog because nobody had the bandwidth. Start building the organizational knowledge that only comes from doing.
These two tracks aren't in conflict. They reinforce each other. The AI initiatives reveal where data gaps actually cause problems (as opposed to where they theoretically might), and that insight makes your data investments sharper and more targeted.
Don't Let the Perfect Be the Enemy of the Possible
The data estate matters. Nobody serious about AI would argue otherwise. But it was never supposed to be a prerequisite for all forward motion. It's a foundation you build and improve over time, not a gate you need to pass through before anything else can happen.
Start with what you have. Improve as you go. The organizations that figure this out will look back in a few years and realize that the decision to move on both fronts at once was one of the best strategic calls they made.