To Scale AI in Banking Centralize Resources
AI development is constrained by staff and by risk, banks must centralize both!
McKinsey must have read my book "Innovation Lab Excellence” because their solution that banks centralizing resources when developing AI is right from my book!
In a market with a highly constrained supply of AI or, in my book, digital talent, the only way to build a workable team is to make them a central resource.
This avoids dispersing people within an organization so they never reach critical mass to do anything!
AI also has another unique component: risk! Centralizing resources also allows for better risk control in a technology that hallucinates and may have built-in bias.
👉TAKEAWAYS
— Given the scarcity of top gen AI talent, centralization allows the enterprise to allocate talent in a way that is more likely to benefit the entire organization.
— In a rapidly changing environment where new large language models and gen AI features are regularly being introduced, a central team can stay on top of the evolving gen AI landscape better than several teams dispersed across an organization.
— A centrally led operating model is useful early on in an enterprise’s gen AI push, when it is necessary to make frequent and important decisions on matters such as funding, tech architecture, and LLM providers.
— Risk management and keeping up with regulatory developments are easier with a centrally-led approach.
The highly centralized AI approach produced the greatest percentage of AI in production. When talent is in short supply, spreading out critical resources to business units is a recipe for disaster.
👊STRAIGHT TALK👊
When I wrote Innovation Lab Excellence in 2019 banks faced a crisis in hiring employees for big data and digital transformation.
Now, a few years later, AI is the rage, yet the problem remains the same: how to hire a team when talent is in short supply.
The solution for banks trying to launch AI systems with a talent shortage is also unchanged: centralize resources. Whether you call it an AI lab, team, or group is immaterial; they all need to sit in the same place and learn to work with business units.
McKinsey may have read my book but came up with some great statistics that show how centralization works for AI development.
“About 70 percent of banks and other institutions with highly centralized gen AI operating models have progressed to putting gen AI use cases into production, compared with only about 30 percent of those with a fully decentralized approach.”
As I wrote in my book, the key to centralized resources is that they work with the business units but are not controlled by them. They must keep their independence to cut losses if things don’t work out.
I want to rewrite my book for AI just by doing a search and replace because all that is old is new again!
Will that get me paid like Mckinsey?