Why Your Data & AI Plans SHOULD NOT Rest on One Overworked Human

Your company’s data strategy probably looks a lot like that meme: a giant elephant (all your ambitions, AI plans, and reporting needs) delicately balancing on a beach ball… that’s being held up by a couple of very tired ants. In real life, those “ants” are usually a single data person who quietly keeps the whole thing from collapsing.

In many organisations, that one person has built custom data pipelines, automated dozens of Excel workflows, stitched together APIs, and spun up more dashboards than anyone can count—often with zero documentation. In the moment, it feels amazing. Work gets done quickly. They never say no. Every new request somehow gets squeezed into their day. They become the unofficial owner of every metric and every “quick data pull.” Before long, every project routes through them, whether it should or not.

Most businesses around the world can reasonably resonate with this message.

At the same time, leadership is busy talking about “expanding our AI strategy” and “unlocking value from data.” Maybe, if things go well, they’ll even approve headcount for one more data hire. But structurally, nothing changes: the organisation is still built around key person dependency—the risk that everything depends on the knowledge, context, and goodwill of one individual.

AMI sees this up close a lot. Over the past few years, around one-third of our team's projects have come from exactly this issue. A data lead moves on, gets promoted, or simply burns out. Suddenly, nobody really understands what they were doing. Dashboards start failing quietly. Automations that “just worked” stop running. External partners complain that they aren’t getting data refreshes. The business realises, often too late, that it never had a real data strategy—just someone willing to put in late nights and weekends.

Key person dependency isn’t just a staffing problem; it’s a governance problem. If your AI roadmap, reporting, and operational decisions all hinge on a single individual, you don’t have a scalable data function—you have a heroic workaround. And heroics don’t scale. They also don’t survive holidays, notice periods, or burnout.

A healthier model looks very different: documented pipelines, shared ownership of critical metrics, clear SLAs for data products, and a roadmap that separates “business-critical” from “nice-to-have.” It means investing in processes, not just people; in repeatable systems, not just clever one-off solutions. The meme is funny because it’s true—but if your entire data and AI story depends on one overworked ant, it’s also a warning.