Can AI Generate Value Without Data Governance?

Pegasus Airlines Chief Technology Officer Barış Fındık spoke about generating value from artificial intelligence without data governance. Fındık stated the following:

Of course, the concepts of “big data,” “data is the new oil,” are older than AI; we’ve been talking about this for even longer. Perhaps its history goes back more than 10 years. Regulations are also very important for the aviation sector. Because, perhaps, they are also important for banking and telcos. I also have experience in banking and telcos in the past, but aviation is a business that needs to be done entirely according to regulated, world standards. Therefore, European Union regulations, different regulations around the world, and Türkiye’s own regulations are all very binding for us. In terms of both KVKK (Personal Data Protection Law) and GDPR (General Data Protection Regulation), the security of data, how it is processed, and how it is managed are very valuable to us. Therefore, our governance models, processes, and structures that we will consider when developing projects are all defined. We are trying to operate these structures effectively at Pegasus. But structuring our data is an issue that needs to be addressed before artificial intelligence projects. We now have a program called “DNA” that we are working on. Within our “DNA” program, we are creating a data catalog and a data dictionary. In other words, we’re ensuring the organization speaks the same language. What’s the purpose of this? For example, a new concept has emerged: “Ask your data.” You ask your data, and it answers you. But the entire organization needs to understand the concept from the question I ask my “data” in the same way so that the answer I receive is consistent and comprehensive. We are trying to achieve equalization at every level, from the general manager to our experts, regarding these concepts. This work, which we’ve been doing for 3-4 months, will continue for over a year. All the company’s concepts and data are being cataloged in a dictionary, with a consensus. All our AI projects are actually built upon this catalog. I believe that establishing this underlying foundation correctly is invaluable for AI projects. Of course, aligning these with regulatory definitions, addressing them with a proper “governance” module, logging data we haven’t logged, thus enriching the data, feeding our “datamarts” richly from different sources, and building AI projects on top of it is only possible. Let’s say that deriving value from these is only possible in this case.