I have spent the last 8 years of my 17 year career working in financial services environments that are grappling with some of the most complex data challenges. Candidly, the very nature of the financial services industry lends itself to high data risk exposure. The high variability, velocity and volume of data coupled with an organizational propensity for silos, is a mean recipe for data inaccuracy, inaccessibility, and, correspondingly, unreliability. As if that internal complexity weren't enough, external forces like markets, politics and regulatory governing bodies add additional layers of pressure that make the business-as-usual data management approach untenable. Industries such as healthcare, government and pharmaceuticals face similar data management challenges.
The proliferation of intra-organizational cottage industries in data management has increased the technical debt within most organizations. This is a complex but not an entirely unwindable entanglement. In my experience, this work requires both a great deal of art and science. In my experience, the most successful data transformation efforts have been driven by those who are nuanced in their thinking and have an earnest commitment to achieving desired outcomes. Unfortunately, it has also been my experience that these qualities are rare and often discouraged by the organizational ethos.
I have learned a lot about data over the years. I have learned even more about what not to do when deciding to transform data management practice within large firms. As such, I have developed a number of key positions on the matter. Here are some of my thoughts:
I have always found it valuable to call a thing a thing. Along that vein, I have yet to find hand-wavy big shop consultants useful for any shape of technical delivery. Convince me to the contrary.
Find a compromise between the “boiling the ocean” and the “stuck in PoC” types of approaches. Yes, many would agree that the risks associated with going too big, too fast are fiscally and culturally inopportune for most firms. However, the latter, in my opinion, is far more injurious to the bottom-line as it demoralizes ambitious and innovative thinkers, causing your best people to extinguish or flee. Get rid of the zero-sum mindset.
Teams charged with implementation duties painfully want to deliver but are often ill-equipped to navigate the broader tense political climates within organizations. Good managers provide air-cover so folks can focus on getting things delivered.
Technical women are frequently and prematurely pushed out of the rooms that need them the most. Intellectual and experiential diversity positively impacts outcomes. I find diversity of thought particularly essential in AI and the area of Knowledge Representation. Engagement from a broad intersection of knowledgeable and skilled folks is absolutely essential for AI accuracy. Follow the data and get it corrected, leaders!
The most irksome phenomenon I have encountered is the broad deployment of a technical debt modus operandi. I have never seen this work! The negative impact of this approach is usually quite profound and often results in complete project implosion and/or someone(s) getting fired.
I founded Reclassify AI in response to many of the frustrations listed above. I have formed a company and a team that delivers on the principles that work. We are pretty nerdy about data and passionate about SemTech. We also receive a great deal of personal satisfaction in solving problems and delivering good work. No drama. No BS.