Beyond the Hype: Making GenAI Enterprise-Ready With Agile Governance
Generative AI — that broad term for the tool that creates all manner of content (codes software, automates industrial processes, composes music, produces deep research reports, and so much more) — is evolving at breathtaking speed. I’ve been programming since age seven and starting internet-powered companies since age 24, and I’ve never seen a technology evolve faster. I wrote about the stakes last month in an article about the nature of consciousness and our exponentially accelerating pace in AI, quantum computing, and robotics.
Now, businesses aren’t thinking about whether to use generative AI anymore — they’re thinking about how to make it work at scale. Of course we’ve seen this movie before. New tech comes along promising to change everything, but then hits a wall in the form of real-world business constraints.
That’s exactly what’s happening with genAI right now.
Yes, some companies are making it work beautifully. Bosch has integrated AI into manufacturing. Mayo Clinic is using it to transform healthcare. The drug discovery startup Insilico is leveraging AI to discover treatments for fibrosis. And there are many, many more inspiring examples. But here’s the kicker — the vast majority of companies just aren’t ready. In fact, would you believe that 96% of technology leaders said they aren’t AI ready from a data perspective? That’s a staggering statistic from Gartner last year. Data are the fuel for enterprise AI; data are the facts about a business or organization. How do you manage a business well without managing your facts well?
Why are so many companies struggling?
The barriers are very real, so let’s break them down:
Many organizations are dealing with messy, incomplete, or biased data. What I call Fuechel’s Law — named for the IBM instructor George Feuchel, who coined the term “GIGO” for “Garbage In — Garbage Out” back in the 1960s — still applies. Messy, disorganized, and incomplete data makes AI outputs that are, well.., messy, disorganized, and incomplete. In other words, useless or worse.
Then there’s the headache of legacy IT systems that fight against AI integration, forcing companies to invest in expensive middleware solutions just to make things work.
Don’t forget about the human element — fear of job displacement and resistance to change are slowing adoption across the board.
And hanging over everything there’s regulatory uncertainty and compliance risks that keep legal teams up at night.
One good place to deepen your understanding of the challenges you face is with our AI Readiness Survey, which we unveiled a few weeks ago and about which I share more detail below.
As one example, let’s take BMW. The famous German automaker is using generative AI to optimize components and materials, but that’s only working because they’ve invested heavily in their digital infrastructure and data governance. Most companies simply haven’t built that foundation yet. They’re still struggling with outdated data management practices, which means their AI initiatives either stall out or produce results nobody can trust.
Here’s a sobering statistic from McKinsey: 42% of enterprises using AI for financial forecasting experience accuracy failures due to poor data governance. Again, AI is only as good as the data feeding it, yet most companies are still treating data like an IT problem instead of a critical business asset.
From bottleneck to business accelerator
Let me be straight: If your data governance process feels like a trip to the DMV — slow, frustrating, and wrapped in red tape — it’s holding back your AI initiatives. Traditional rigid governance models were designed to minimize risk, not to support the agility that AI needs.
We need a shift in thinking. Data governance needs to evolve from a box-checking compliance function to an agile approach that actually enables AI-driven innovation. That means making three critical changes:
1. Treat data as a product, not as an infrastructure problem
Instead of siloed data stores, businesses need data products — curated, self-contained assets that deliver real business value, a topic I wrote about two weeks ago. Think about a well-designed customer churn model, for example. That’s a data product that directly informs decision making. Adopting data mesh principles can help companies break free from outdated architectures.
2. Leverage knowledge graphs and data catalogs
You know what I hear a lot? Executives assuming that AI can replace data catalogs. That couldn’t be further from the truth, as I wrote about last month. Modern data catalogs, powered by knowledge graphs, act as the “brain and nervous system” of enterprise AI. They provide:
- Discoverability: AI can only work if it knows what data exists
- Governance: AI models must comply with regulations like GDPR
- Collaboration: AI works best when business and data teams work together
3. Move to bottom-up, agile data governance
Traditional “top-down” governance slows everything down. Instead, agile data governance empowers teams at the edges — business units, data scientists, and engineers — to make governance decisions iteratively. Companies that adopt this approach see faster ROI on their AI initiatives.
What are the winners doing differently?
Leaders in AI adoption don’t just use the technology; they govern it smartly. Let me share some more real-world examples:
Indeed’s data governance success: Indeed, the world’s largest job site and data.world customer, turned governance from a roadblock into an accelerator. Instead of imposing rigid policies from above, they engaged employees who were already doing informal data governance and empowered them to refine company-wide best practices. The results? Higher data adoption, improved AI accuracy, and increased efficiency across the board.
Rockwell Automation’s AI-ready infrastructure: Rather than trying to overhaul their entire system at once, Rockwell incrementally modernized their PLC controllers to be AI-compatible. This strategic, modular approach cut downtime by 70% while ensuring AI adoption happened smoothly.
JP Morgan’s confidential AI transactions: JP Morgan processes $6 billion in AI-driven transactions daily without exposing client data. How do they pull that off? Through federated learning and encryption-based AI models, ensuring they stay compliant with regulations while still enabling innovation.
These companies are succeeding because they recognize that AI governance isn’t an afterthought — it’s the foundation of being AI-ready.
How to become AI-ready: your game plan
If you’re looking to move beyond the hype and make generative AI a reality in your organization, you need a clear plan. Here’s where I suggest starting:
1. Assess your AI readiness: Take an AI Readiness Survey to benchmark where your organization stands. Most companies overestimate their preparedness, only to be blindsided by governance and infrastructure challenges later.
2. Prioritize AI-ready data governance: Shift from a compliance-heavy, top-down model to agile data governance that aligns with your business goals. This means investing in data products, knowledge graphs, and automation tools to reduce governance friction.
3. Measure ROI and iterate: AI success isn’t just about deployment; it’s about measuring impact. Companies that track ROI from AI governance — whether it’s efficiency gains, risk reduction, or new revenue — are far more likely to succeed in scaling AI.
Remember, the goal isn’t just to implement AI — it’s to create business value through AI. And that starts with smart governance.
AI is moving faster than ever, but is your organization keeping up? The AI Readiness Survey — developed by my colleague Brandon Gadoci, our VP of AI Operations — helps you assess your strengths and gaps in key areas:
✅ Data Culture
✅ Governance & Compliance
✅ AI Strategy & Advanced Analytics
✅ Operations & Infrastructure
Most companies score well below industry averages in these areas. The only way to get ahead is to measure where you stand and build a strategy based on real insights.
