A couple weeks ago I put a post out on LinkedIn about AI Governance. It got a few thousand eyeballs on it. It deserves a more thorough "unpacking" because it’s a big topic.
If you Google “AI Governance” you will get an amazing number of results. Go on, try it, I’ll wait. Every vendor, government agency, quasi-government agency, consulting firm, your grandmothers knitting club , now has an AI governance framework. Some of that is because we need to be able to answer the questions when our clients and prospects ask. Some of it is because we don’t want to miss the bandwagon. Data governance has never been that popular, but you want to know a secret? They aren’t that different.
We find ourselves on the cusp of such a compelling set of technology. Granted AI has been around for decades, but it is now that it is becoming more ubiquitous and commercialized. The governance rules that are happening today across and between nations demand things like transparency, accountability, and explain-ability to name a few. There’s one problem with these standards. Most people don’t understand enough about AI or machine learning to appropriately govern. A lot of governance structures put layers between the people that know and the people that are most affected by them. Much like we’ve done in data governance for decades.
As we all teeter on this cliff that is AI getting salient advice from people that have experience is paramount. The trouble is, very few people have “been there done that” with AI governance. Of those that are out there talking about it, precious few have done any work in any governance. What I have found in working in data governance for the last twenty-five years and specifically the last five almost exclusively is that governance and democratization (a fancy way of saying usage) are two radically different concepts. Data Governance historically has either tried to stop or slow, not manage, or guide. And while most of my governance friends would say that is not the spirit of the function it is without question the result. It is also the intent of the new AI governance functions that are popping up. But in the case of AI, for good reason. Where does this leave us?
We need to clarify the scope of data governance functions. So that we can appropriately govern aspects of AI that overlap with data. If we don’t, we will fail at both. Conflating too many of the critical functions of governing AI and ignoring basic functions that support the use of data in an organization.
Comments