Weaponizing Governance
Many data governance processes originally created as defensive measures driven by regulatory requirements are now being re-purposed and enhanced in support of offensive strategies around machine learning and artificial intelligence
A Necessary Evil
Different industries have matured into an understanding of, and respect for, data governance at different rates. It’s fair to say that the financial services and insurance industries, long driven by the collection and analysis of detailed information [think, actuarial tables and tick marks on clay tablets] are among the longest-standing practitioners of this function. Regulatory demands imposed since the crash of 2008 have pushed these programs forward aggressively during more recent years.
In this context, many of the policies and procedures that we now think of as within the realm of modern data governance were established on a defensive footing. Corporate leaders, especially those not personally engaged with the technology-side of their organizations, thought of governance as an expensive necessary evil in response to regulatory demands. Attributes of these programs were based tightly around the specific requirements defined by associated regulation, and the whole assemblage was treated — begrudgingly by many — as a cost center.
Solid foundation to build upon
That’s just fine, and there are many laudable parts of our work that we rack-up simply as “the cost of doing business”. Those of us in the trenches of problem-solving at large, data-rich enterprises, especially those with vast portfolios of historical content, have long recognized that within the jurisdiction of data governance lie many of the keys to success up and down the information supply chain throughout the firm. As Eric Aranow [ https://www.linkedin.com/in/eric-aranow-3ab34b/ ] has gone to such pains to teach me over the past year, if your data isn’t endowed with socialized, normalized, domain-specific meaning, you're basically nowhere. A few other ways in which the functions often associated with data governance are pivotal include:
- Creating a centralized dictionary for domain-specific semantic terms, and mapping to all data elements coming from systems of record and external sources to those meanings
- Establishing taxonomies and/or ontologies that provide data elements with self-referential context
- Defining a standard methodology and structure for curation of the data, and clear expectations about the ‘curation attributes’ of every data set available for usage
- Centralized documentation about what data assets are available, as well as their lineage
- Some subscription system so that the managers of the data can understand who is consuming it, and for what uses. This is clutch when the source data needs to change, and associated downstream impacts must be defined and mitigated
- Strict rules about the sourcing of data to be utilized for business decisioning, as well as passing that data downstream to additional consumers
“If your data isn’t endowed with socialized, normalized, domain-specific meaning, your’re basically nowhere.”
Feeding the ML and AI beasts
Children of the 1980s know that among the recurring themes of movies from that era was the parable of the wallflower suddenly catapulted into popularity. All of sudden, data governance finds itself playing the protagonist in a John Hughes film. Among the newest offensive systems available to large, global enterprises are machine learning [ML] and artificial intelligence [AI]. YES, we can have a debate about the fact that ML has existed almost since the first computer programmer got tired of writing code to manage every permutation of her application, as well as the reality that 80% of the modern use cases popularly identified as ‘artificial intelligence’ are a far cry from the strict academic definition of that capability. But that doesn’t change the fact that recent technology and organizational advances have brought these capabilities to the attention of senior business leaders, who are actively investing to leverage their potential benefits to the bottom line. This is all ‘offensive’ strategy — wielding new tools in order to create market differentiation from and/or wring more value from existing data assets than the competition. And that’s where the story wraps-back to data governance.
“All of sudden, data governance finds itself playing the protagonist in a John Hughes film.”
One of the first things these firms learn in their experimentation with ML and AI is that those technologies thrive on a steady stream of pristine data. The knowledge bases these systems utilize to solve thorny problems must first be established and endowed through direct synthesis of a broad and deep portfolio of known data. To understand what is meant by the term ‘known’ here, glance back up at the bulleted list above, associated with the functions of data governance.
In their enthusiasm for the promises of ML and AI [real and/or perceived], many corporate leaders who used to keep data governance on a starvation diet are looping back around bearing gifts to the very same functions within their orgs. They might not clearly understand the details around keeping enterprise data clean and well-governed, but they have learned those to be prerequisites to unleashing ML and AI for the advancement of their business goals. And that’s enough for them to become vocal boosters.
You’re Never Fully Dressed Without a Smile
Long-term data management practitioners can be excused for a skeptical eye roll over their newfound popularity. Welcome to the ‘profit center’ club. But they shouldn’t allow those emotions to preclude them from capitalizing upon the opportunity that this scenario represents. At the end of the day, our goal has always been to create and deploy the highest quality and most easily accessible data assets to the enterprise. Just because the funding to execute on those dreams suddenly flows from Daddy Warbucks and his newfound fascination with Hal 9000 doesn’t preclude the fact that we are all on the same team, and our own agenda is being serviced through this activity.
Great data governance has always been the ‘right thing to do’, and a deeply-held philosophy to that effect has long been a secret recipe of many performant modern enterprises. While their competition continually squeezed the governance function of funding in reaction to market pressures, those performant companies quietly recognized governance as the root system supporting every upward-reaching branch and leaf of their growing enterprises, and invested faithfully.
All of a sudden, due to the promise of ML and AI, even those less-self-aware organizations are becoming converts to the faith of data governance. Those of us on the data management side of the house should welcome them with open arms, provide them with honorary pocket protectors, and maximize the opportunity this represents for the good of the precious data in our enterprise portfolios. Our fifteen minutes will be up before we know it.
