Highlights from Craig S. Mullins’ New Book The Cost of “Good Enough” Data
For years, organizations have lived comfortably with data that was “good enough.”
Customer records with a few inconsistencies. Product descriptions maintained in multiple systems. Reports that never quite matched. Data pipelines patched together over time. Documentation that existed mostly in someone’s memory.
None of these problems seemed catastrophic. Business continued. Applications ran. Decisions were made.
Then artificial intelligence arrived.
In his new book, The Cost of “Good Enough” Data, Craig S. Mullins argues that AI has fundamentally changed the economics of enterprise data. Data quality is no longer simply an operational concern for database administrators or data architects. It has become a strategic business issue that directly influences the reliability, trustworthiness, and competitiveness of AI-driven organizations.
The Hidden Costs Have Always Been There
One of the book’s central messages is that poor data rarely causes immediate disasters. Instead, the costs accumulate quietly.
Organizations pay through:
- Inefficient business processes
- Duplicate and conflicting information
- Regulatory exposure
- Poor customer experiences
- Increasing maintenance costs
- Growing technical debt
- Reduced confidence in analytics
These expenses often remain invisible because they are spread across departments and projects. Each individual problem appears manageable. Collectively, they become enormous.
AI Changes Everything
The emergence of generative AI has dramatically raised the stakes.
Unlike traditional applications, AI systems consume vast amounts of enterprise data. Every inconsistency, missing value, outdated record, undocumented business rule, and governance gap can influence model outputs.
The book explains that AI systems are only as trustworthy as the data flowing into them. If organizations feed AI inconsistent or poorly governed information, they should not be surprised when the resulting answers are inconsistent or unreliable.
The AI Supply Chain
Perhaps the book’s most compelling concept is the idea of the AI supply chain.
Rather than viewing AI as simply a large language model answering questions, Mullins encourages readers to think about everything that happens before a prompt ever reaches an AI model.
The AI supply chain includes:
- Operational systems
- Transaction processing
- Database management
- Data integration
- Data quality
- Metadata
- Governance
- Security
- Data movement
- Semantic enrichment
- Retrieval systems
Every step affects the quality of AI outcomes.
This perspective shifts attention away from AI models themselves and back toward the enterprise systems that create and manage trusted business data.
Mainframes Are More Relevant Than Ever
One surprising theme throughout the book is its defense of enterprise systems of record.
For years, many organizations viewed mainframes as legacy technology.
Mullins argues exactly the opposite.
The systems that process financial transactions, airline reservations, insurance claims, healthcare records, and retail purchases have spent decades proving their reliability. They deliver consistency, auditability, integrity, and security—precisely the characteristics AI initiatives now require.
Rather than replacing these systems, organizations should recognize them as foundational components of modern AI architectures.
Governance Is Competitive Advantage
Another recurring message is that governance is no longer merely about compliance.
Well-governed data enables:
- More trustworthy AI
- Faster innovation
- Better customer experiences
- Lower operational risk
- Greater organizational confidence
Organizations that invest in governance are not simply reducing risk—they are creating better AI.
Architecture Matters
The book also challenges several popular assumptions about modern data architectures.
It questions whether copying data into countless data lakes and cloud platforms truly improves business value.
Instead, Mullins argues that every additional copy introduces additional cost, synchronization challenges, governance complexity, and opportunities for error.
Sometimes the simplest architecture—the one closest to the authoritative source—is also the most reliable.
Practical Rather Than Theoretical
One of the strengths of The Cost of “Good Enough” Data is its practical perspective.
Rather than presenting abstract theories, Mullins draws on decades of experience designing, managing, tuning, and troubleshooting enterprise database environments.
Readers encounter real-world situations familiar to DBAs, architects, IT managers, and executives responsible for critical business systems.
The writing consistently emphasizes that technology decisions have long-term operational consequences.
Who Should Read It?
Although database professionals will naturally appreciate the technical discussions, the audience is much broader.
The book will benefit:
- CIOs and technology executives
- Enterprise architects
- Data architects
- Database administrators
- AI leaders
- Data governance professionals
- Application developers
- Business leaders responsible for digital transformation
Anyone responsible for enterprise data—or planning AI initiatives—will find valuable insights.
The Bottom Line
For decades, organizations could often survive with data that was merely “good enough.”
That era is ending.
Artificial intelligence depends upon trusted, governed, consistent enterprise information. Organizations that continue to tolerate poor data quality, fragmented architectures, and weak governance will increasingly discover that the true cost of “good enough” is far higher than they imagined.
Craig S. Mullins’ The Cost of “Good Enough” Data makes a compelling case that data excellence is no longer optional. It is the foundation upon which successful AI, reliable analytics, and resilient digital enterprises are built.
The book serves as both a warning and a roadmap, reminding readers that the quality of tomorrow’s AI begins with the quality of today’s data.