According to Forbes, Vijay Mehta, Chief Data & Technology Officer at Experian, argues that enterprise AI success depends not on models but on the “plumbing” – things like model drift detection, compliance automation, and governance infrastructure. He stresses that organizations must start with intent rather than jumping straight to building models, focusing on data flows and decision points before AI implementation. Experian has launched the Experian Assistant for Model Risk Management, an AI-powered solution that automates documentation and compliance audits while aligning with global regulatory standards like SR 11-7 and SS1/23. The biggest reason AI pilots fail is that fundamentals like clean data and operational pipelines aren’t in place, leading to what’s called “pilot purgatory” where projects show promise but never scale into production. Success comes when AI becomes invisible in workflows, driving measurable business outcomes rather than just technical metrics.
Sponsored content — provided for informational and promotional purposes.
The plumbing-first reality
Here’s the thing that most companies don’t want to hear: the exciting part of AI – the models, the predictions, the “magic” – is actually the easiest part. The real work happens in the boring stuff. Data pipelines. Governance frameworks. Version control. Compliance automation.
Vijay’s point about treating AI as an engineering discipline rather than magic really hits home. I’ve seen so many companies get excited about building a fancy model, only to realize they have no way to monitor it, update it, or even explain how it works. And in regulated industries like banking and healthcare? That’s not just inconvenient – it’s potentially catastrophic.
Escaping pilot purgatory
That term “pilot purgatory” is so accurate it hurts. Basically, companies build something that works in a controlled environment with clean data, then try to scale it to real-world conditions and everything falls apart. The model that was 95% accurate in testing becomes unreliable when faced with messy, incomplete, or changing data.
And here’s the kicker: Vijay notes that organizational silos are often the real culprit. Data scientists build something amazing, but without buy-in from operations, compliance, and IT? It’s going nowhere. We’re still treating AI like it’s a technical problem when it’s actually a business transformation problem.
When success looks different
The most interesting part of Vijay’s perspective is how he defines success. It’s not about accuracy metrics or dashboard numbers. It’s about adoption. When people stop thinking about AI as something special and it just becomes part of how work gets done? That’s the goal.
But here’s what I think many leaders miss: this requires being willing to kill projects too. If a model isn’t delivering real business value, or if the complexity outweighs the benefit, you need to have the discipline to walk away. How many companies are stuck maintaining AI systems that don’t actually move the needle?
The governance imperative
With regulations like SR 11-7 and SS1/23 becoming the global standard, companies can’t afford to treat AI governance as an afterthought. These aren’t just nice-to-have guidelines – they’re becoming mandatory requirements for anyone using AI in critical decision-making.
The platform approach that Experian is taking makes so much sense. Building governance and monitoring directly into the model lifecycle from day one, rather than bolting it on later. Because let’s be honest – if you try to add compliance and auditing after the fact, you’re already behind. The organizations that get this right will be the ones scaling AI successfully while others remain stuck in perpetual pilot mode.
