The big speed bump on the road to GenAI

It seems that every event I moderate, regardless of the topic, will devolve into a discussion of generative AI and the excitement of intelligent systems.  The enthusiasm for this innovative technology is irrepressible.  Ideas fly around, grandiose plans are discussed, and everyone can’t wait to get going.  However, as the discussion moves back to more thoughtful interaction, it soon becomes clear that the “Highway to AI” has one very large speed bump in the road.  Almost without exception, none of the companies in attendance have a data foundation to support it. 

AI apps without a solid, accurate, and complete data set aren’t worth much.

The natural question is, why isn’t the foundational data in place?  There are many reasons for this.  One attendee noted, “Building a data foundation is expensive, and it’s not sexy.  Management doesn’t get excited because it’s kind of like plumbing.”  When CEOs get together, they don’t brag about building a corporate data asset.  And it’s a bit invisible.  It doesn’t show up as a cool new feature on your website.

Another hurdle to building the data foundation is getting all the compliance, security and regulatory issues resolved.  This isn’t a trivial exercise.  There are data locality issues, protection demands, and more.  And if you are a global firm, the complexity grows exponentially.  The strategic approach in some organizations is to anonymize everything that could be private or protected to get past the stipulations.  However, total or widely used anonymization may make the data less or even non-useful for some AI applications.  As one of my attendees noted, “If we anonymize everything, how do we improve John Doe’s CX? That’s our goal.”

And this doesn’t even consider tasks such as merging databases, developing up with APIs to support the integration of transactional data, or myriad other potential issues. 

So, what to do?