It’s been niggling me for a while. Every time a new technology arrives, organisations rush to lift and shift, taking what and how we once did something and simply placing it on the new tech.
It’s an expensive habit. We end up doing the same things in new systems instead of using new systems to do different, better things.
And now, it’s happening again. This time with AI.
Across industries, we’re seeing the same pattern: the urge to lift and shift existing tasks, processes, and decisions into the AI tech stack, without rethinking what could genuinely change.
AI governance is suddenly the hot topic in every board, exec, and leadership meeting. That’s progress of sorts. Digital governance took years to evolve as we learned how to manage digital platforms responsibly.
But Tony Fish makes a powerful point on LinkedIn:
Can we really predict the right governance guardrails for something as fundamentally different as AI?
We can’t simply copy the frameworks we built for digital. AI demands something else — governance that learns as fast as the models it’s meant to guide.
Every AI model and use case being deployed today isn’t a “set it and forget it” moment. It’s another operational and technical layer that must learn from real life.
That means your governance can’t be static either. It needs to evolve through feedback, ethics reviews, incident data, and new ways of working that emerge over time.
So ask yourself:
- Are you designing procedural guardrails and operational ones?
- Have you built in a way to continuously learn and apply those lessons to your AI models?
- And have you made that a business-as-usual responsibility, not something parked with the transformation programme?
Lift and shift isn’t only a technology habit. It’s a human one. We carry the old ways of working across because they feel safe, and we leave the people who have to live with the result to catch up on their own. That’s the human gap: the distance between the plan and what people actually believe, experience and do, and the Reframe lens asks the question that closes it: what are we not seeing?
We already have DevOps, SecOps, and FinOps. But where’s AiOps, the operational practice of keeping AI aligned, adaptive, and accountable once it’s in the wild?
If AI is to become a living part of how we work, we’ll need to manage it with the same rigour we apply to any other critical system — but with far more humility about what we don’t yet know.
Maybe AiOps isn’t just a function waiting to emerge. Maybe it’s the mindset shift we’ve been missing all along.
Whatever next?
