After a wild and turbulent 2020, the new year has ushered in a renewed commitment to establishing or improving corporate governance. Yet, positive energy aside, our traditional approach to endorsing governance of data, analytics, or AI remains fraught. As a result, governance initiatives springing from an earnest desire to do right (e.g., responsible AI), as well as the need to not do wrong (e.g., regulatory/compliance), struggle to enlist broad coalitions of the willing.
The fault lies with our collective understanding about what governance is and does. Too often, governance is still seen as a necessary evil—one that imposes bureaucratic overhead by mandating activities that slow progress, inhibit innovation, and prohibit teams from rapidly delivering on their commitments. Sadly, it is often unwittingly promoted as such even by ardent supporters.
Putting Governance to Work
To put governance to work, we must change the mindset of our internal and external constituents. Yes, governance is a means to an end. But the endgame is not about compliance, perfunctory completion of checklists, elimination of risk, or passing the buck. Rather, the purpose of governance is to increase the cadence and quality of our collective and individual decisions.
This is true whether the deliverable is a simple BI dashboard or a complicated AI algorithm. From providing the proper interpretation of a business metric to deciding if an AI application can be deployed equitably and safely, governance serves as a universal translator and navigational beacon—one which ensures that everyone, regardless of role, is speaking the same language and working toward the same outcome.
Governance should, therefore, result in better- informed and more confident executives, managers, analysts, developers, and ultimately—consumers. This is perhaps counterintuitively accomplished by making time for mindful discussion of common, clearly defined goals. More specifically, governance encourages and indeed mandates collective, and therefore more robust, consideration of questions such as the following:
- Who is the application designed for? Intended to benefit?
- What is the target population or environment?
- What are the expected operating conditions?
- What decisions or actions will result from this application?
- What other interactions may impact this decision/action?
- What are the limitations of the information/system?
- Can and should we act within those limitations? If so, when and how?
- Will this achieve the intended outcome?
- Is “it” doing what we thought it would do? Running amok? Veering into unexplored territory?
- How will/do we know?
Equally important, governance creates mechanisms to probe and refine the collective understanding over time as systems and processes are developed and deployed. As such, governance is continuous learning in practice. All governance activities—from informing go/no-go decisions to profiling data to instituting robust ModelOps—operate in the service of a singular goal. Namely, the continuous enrichment of the team’s knowledge and ability to make informed decisions. This increases the probability that desired outcomes are achieved while unintended consequences are expeditiously identified and addressed. Governance can also go one step further by soliciting consideration of not just whether the intended goal is achievable but whether it is the right goal at all.
It has been well established that effective governance escalates innovation, improves product and services offerings, enhances the user experience, and reduces risk. Success often comes from avoiding unforced errors before grievous harm occurs or too much time and money are wasted in pursuit of an unrealizable or unwise goal. But, in order to benefit, the organization must engage.
Garnering broad, willing engagement requires shifting the narrative. Done right, governance encourages curiosity, increases knowledge, and helps teams uncover what they need to know to build better, do better, be better. If the activities mandated by your governance process cannot be articulated in these terms, a rethink is in order.