The AI hype cycle in financial services is over.
Proof-of-concept pilots, innovation labs, and flashy chatbot demos no longer impress boards, regulators, or customers. The real question facing every banking executive today is brutal in its simplicity: Where is the value?
The days of experimentation without execution are finished.
Today, success demands something far more difficult: operationalizing AI at scale, with full trust, traceability, and a demonstrable return.
Banks that can't make that leap will be left behind. Fast.
From magic to mandatory
In the early stages, AI felt almost magical.
Banks explored use cases, toyed with new capabilities, and imagined what could be.
Today, imagination isn’t enough.
According to McKinsey, AI could add up to 15% to banking profits annually. But while the potential is extraordinary, very few banks are truly ready to capture it: Fewer than 10% of banks have built an enterprise-scale roadmap for AI transformation.
The gap between ambition and reality is now a chasm—and closing it requires a radical shift from pilot projects to enterprise platforms.
The winners will be those who stop playing with AI and start deploying it strategically, securely, and at scale.
Long live the thread
There’s a reason we’re talking about “following the thread.”
At the heart of this new AI-driven future is a simple truth: Trust is everything. And trust demands traceability.
Banks must be able to prove where their data comes from, how it’s transformed, and how decisions are made—from source to insight to action.
No thread? No trust.
No trust? No scale.
Regulators are moving to cement this expectation.
- The EU AI Act will demand transparency for high-risk AI applications
- BCBS 239 requires full governance and data lineage for risk and regulatory reporting
- UK regulators are taking a principles-based approach that demands clear accountability
Banks that can’t “follow the thread” will soon find themselves on the wrong side of the regulators—and the market.
Proof it can be done
It’s tempting to think full traceability at scale is impossible.
Most banks are struggling—and some industry leaders have openly admitted they have “no idea” how to fully comply with AI and data governance demands.
But success is possible.
A large ECB-supervised bank recently achieved something no other major bank has managed: After a formal review under BCBS 239, the bank was the only one to walk away without an action plan—meaning full compliance across the board.
Consider the scale:
- 1,100 KPIs actively governed
- 1.7 million data columns tracked and traceable
The bank didn’t just theorise about data lineage. They built it. They have successfully industrialised trust.
And they’re proof that “following the thread” isn't just a slogan—it's a competitive advantage.
The right tools (not just the big ones)
There’s a growing temptation in banking to throw one enormous large language model (LLM) at every problem.
It sounds simple. It looks impressive.
And it’s very expensive.
Compute costs for huge LLMs can spiral. Worse, using a giant, generalist model where a targeted solution would suffice often creates inefficiencies and increases risk.
Teradata’s view is clear:
- AI must be fit for purpose
- Use the right model for the right task
- Architect for explainability, efficiency, and scale—not just headlines
The banks that combine smart model selection with complete data traceability will not only scale faster—they’ll scale safer, cheaper, and smarter.
The new reality
The age of AI magic tricks is over.
Now comes the hard work and the real rewards.
Banks that can follow the thread—building AI that is transparent, traceable, and trusted—will win the next era of financial services.
The rest will be left wondering where it all went wrong.