The Operating Model Debate is a Trap
Centralized, distributed, hybrid…
Seems like every few months we twist ourselves into knots over which way of structuring our AI functions is most… functional.
For example the IBM Institute for Business Value recently published survey results showing that “More centralized operating models help organizations scale AI.” Of course, this was based on self-reported answers to just two questions. Oh, and they lumped hub-and-spoke models in with centralized models.
Deloitte chimes in with a view that, “To unlock the full value of AI, organizations must rethink their operating models and embed AI into the business core.” Got it. Another vote for centralization, right? Well, maybe not. They immediately follow with: “To keep pace, organizations must replace traditional structures with dynamic, adaptive organizational design.” Sounds pretty distributed to me. Which is it?
Not to be outdone, McKinsey go full agent mode: “The agentic organization may offer the key for the leaders to gain a competitive advantage by building decentralized outcomes-focused agentic networks.” So, decentralized, then?
Honestly, if you’re out there trying to build a new operating model for your business , or advise your leadership team about one, you must be so confused by all these…expert opinions.
Isn’t it a bit oversimplified to think one size fits all?
What I see working most often in large asset intensive industries, like mining, is a hub-and-spoke approach. The center in these businesses is very far from operations (the coal face, as it were). One company might have operations in several countries, which means different operating technologies, processes, and cultures. Centralized mandates in these situations often fail to drive real adoption. But without a unifying strategy and data governance framework, the system is inefficient and costly. So some bits of the system end up centralized, and some bits are distributed to those with the on-the-ground subject matter expertise and context to get things done.
But that doesn’t mean this approach is best for everyone. Surely it’s not a fit for scaling a consumer SaaS business, I imagine?
How are AI leaders to get real insights about what’s working for businesses like theirs - same industry, similar size, similar use cases? How can you learn which approaches might actually be worth trying in your business?