Unveiling the Chinks in Data Management and AI Adoption in Businesses: Debunking Unrealistic AI Fantasies

In the past decade, billions have been poured into data infrastructural tools — behemoth-sized warehouses, real-time pipelines, high-tech machine learning platforms. Yet, the return on these massive investments often falls short of expectations. The crux of the problem? It isn’t the data but rather the approach towards it, the absence of effective product thinking.

The ‘data-as-service’ model, wherein data teams act as in-house consultants, attending to requisitioning queries, was manageable in the early stages. However, as companies have pursued a data-driven method, the model has been pushed beyond its limits. A prime example is Airbnb. Before introducing its metrics platform, teams would pull their unique versions of metrics like ’nights booked’ or ‘active users’; a setup that resulted in operative inefficiencies and discrepancies in decision-making. The root of this dysfunction is not technical but fundamentally a product failure.

On another front, the AI realm is swarming with overstated expectations and unrealistic fantasies. Autonomously functioning AI systems solving limitless problems without the requirement of guardrails and constraints remain a moonshot dream. For context, even a 99% accurate AI in food delivery implies one in every hundred orders landing at the wrong address. Such discrepancies may not only result in expensive consequences but could also batter credibility and leave a mark difficult to erase in front of a client or regulator. Thus, the focus shouldn’t target all-encompassing AI functions but rather, solutions to well-defined problems within clear constraints.