Each week we find a new topic for our readers to learn about in our AI Education column.
“Information is the oil of the 21st century, and analytics is the combustion engine.” – Peter Sondergaard, Gartner Inc.
“Data is the sword of the 21st century, those who wield it well, the samurai.” – Jonathan Rosenberg.
We could dig up 20 more quotes that say something to the effect that data underpins the global economy, culture, social relations, et cetera.
Clearly, data is in vogue. In AI Education alone, we’ve talked about data governance (recently, in fact), data centers, data lakes, data warehouses, data ethics, poisoned data attacks and big data. Zooming out to all of the relatively short history of Digital Wealth News and AI & Finance, we’ve hit topics like data privacy (which we’ll briefly return to today), as well as data silos, data platforms, data management, clean data and data integration.
Here’s another great quote for you, from former Google CEO and chairman Eric Schmidt:
“There were 5 exabytes of information created between the dawn of civilization through 2003, but that much information is now created every two days.”
Schmidt said that towards the end of his tenure as executive chairman of Google, before the Alphabet days… sometime around 2015. So, we can rest assured that we’re making a lot more than 5 exabytes of information every two days now.
That’s a lot of data—and there’s still a trove of data topics to write about in the future. Data—and its quality and reliability—are so deeply intertwined with artificial intelligence both in theory and in practice that we’re going to continually revisit it. The quality of an AI system can never exceed the quality of the data that feeds it. This week, we’re going to talk about one specific facet of keeping data safe, in a topic that relates closely to data governance, data privacy and data ethics: Data integrity.
What Is Data Integrity?
Data integrity, simply put, is ensuring that one’s data retains its accuracy, quality and completeness as it is stored, moved and used. Data integrity is about consistency, not security, but in the process of maintaining integrity, organizations will also end up taking steps to protect data against breaches. Thus, ensuring data integrity looks conceptually a lot like quality control in a physical product business, where a manufacturer will take steps to ensure that both its raw materials and its finished products meet a baseline standard of quality.
We’re starting to ask AI to do a lot of work for us that involves decision making—and as it turns out AI is really no different from a manager or executive we’re employing: If we’re going to rely on AI to make good decisions, then we need to make sure it is basing those decisions on the best available data we can provide.
Before we zoom in, let’s step back: Truly ensuring data integrity will always be more difficult than it seems, especially if we allow ourselves the luxury of thinking on the timescales of geology, chemistry, physics—or the cosmos itself. The nature of data—even the data we store on hard drives—is to decay over time. Let us offer you an example inspired by author and astrophysicist Gregory Benford: Imagine if we were to irradiate part of the planet with an isotope carrying an enormous half-life—how would we craft a warning message to life tens or hundreds of thousands of years in the future? How can we make sure our message doesn’t degrade over time, and it can be seen or heard or received, not to mention understood, by intelligent living entities before they’re exposed to the mortal or morbid effects of radiation? How do we make the information useful to our distant descendats or others? As it turns out, trying to answer those questions requires giving thought to data integrity.
Types of Data Integrity
Physical Integrity
Physical integrity focuses on protecting data from physical threats and infrastructure failures. Hardware malfunctions, storage degradation, power outages, natural disasters, and cyberattacks can all damage stored information. Physical integrity measures help ensure that data remains accessible and intact despite these risks.
Entity Integrity
Entity integrity ensures that every record within a database can be uniquely identified. This is typically accomplished through primary keys and unique identifiers. Entity integrity prevents duplicate records and ensures that transactions can be accurately associated with the correct customer.
Referential Integrity
Referential integrity governs relationships between different data tables. It ensures that references between records remain valid and consistent.
Domain Integrity
Domain integrity ensures that values entered into a database conform to predefined formats, ranges, and rules. It prevents invalid or illogical data from entering systems.
User-Defined Integrity
Organizations often establish additional business-specific rules that reflect unique operational requirements. These constraints are known as user-defined integrity controls. A wealth management firm, for example, might require that every advisory account include a documented risk profile before investment recommendations can be generated. These rules reinforce business processes while improving data quality.
Threats to Data Integrity
Despite all the automation we’ve been implementing, human error remains the most prevalent risk to data integrity. Employees may accidentally enter incorrect information, delete records, use inconsistent formats, or improperly update databases. Even seemingly minor mistakes can propagate through interconnected systems. Poor data collection practices can create problems from the outset. If source information is incomplete, outdated, or inaccurate, downstream systems inherit those flaws. AI models trained on defective data simply automate bad decisions faster.
Software bugs, integration failures, synchronization issues, storage degradation, and system outages can also introduce corruption or inconsistencies into datasets. As organizations adopt increasingly complex cloud-based architectures, these risks often multiply. Cybersecurity incidents present their own data integrity concern. Attackers may deliberately alter records, manipulate transactions, inject malicious data, deploy ransomware, or compromise databases. In such cases, maintaining integrity becomes just as important as maintaining availability.
Data silos represent another growing challenge. Organizations often maintain multiple versions of the same information across disconnected platforms. Without proper integration and governance, inconsistencies emerge, making it difficult to determine which version is authoritative.
Data Integrity & Finance
Few industries depend on data integrity as heavily as financial services. Virtually every financial transaction, investment recommendation, risk calculation, regulatory filing, and customer interaction depends on accurate information. Even small errors can create significant financial and reputational consequences.
Banks process millions of transactions daily. Data integrity ensures that deposits, withdrawals, transfers, and payments are recorded accurately and consistently across systems. Imagine a situation where transaction records become duplicated or partially corrupted. Customers could see incorrect balances, triggering unnecessary disputes, regulatory scrutiny, and operational disruptions. Maintaining data integrity allows banks to reconcile transactions accurately and preserve customer trust.
Financial advisors increasingly use AI-powered tools for portfolio construction, financial planning, and client engagement. A retirement projection model is only as reliable as the data it receives. If income figures are outdated, account balances are inaccurate, or risk tolerance information is incomplete, the resulting recommendations may be unsuitable. Data integrity directly influences the quality of advice delivered to clients.
Perhaps nowhere is data integrity more important today than in financial AI initiatives. Generative AI systems, predictive models, recommendation engines, and autonomous agents all depend on trusted information. Training models on inaccurate, incomplete, or biased data can amplify errors at scale. As AI adoption accelerates across wealth management, banking, insurance, and capital markets, data integrity increasingly becomes the foundation of responsible AI deployment.
