What most banks call AI transformation is actually AI theater — here’s the difference
Most financial institutions are not transforming with AI. They are automating processes that should be eliminated, not accelerated.
Over 25+ years — building production AI/ML into a $120M+ portfolio across Commercial Banking, Capital Markets, and Private Bank — I have watched this pattern repeat. Here is what the difference looks like.
AI THEATER LOOKS LIKE THIS
You take a 52-step credit review process, identify 12 steps a model can handle, automate those 12, and declare transformation. The process still exists. The logic still belongs to 2009. You have made the inefficiency faster — digital paint on a structural problem.
Theater is also what happens when a bank adds copilots or workflow bots on top of a broken operating model. The work is still fragmented. The handoffs are still manual. The decision logic is still trapped in people, spreadsheets, and swivel-chair workflows. The institution gets a demo, not a redesign.
REAL TRANSFORMATION LOOKS LIKE THIS
1. Redesign the operating model before you deploy the model.
The harder question first: should this process even exist?
At Citizens Bank, we did not automate commercial lending. We rebuilt it from first principles — ingesting 400+ data elements, making real-time decisions, eliminating human review where the data was sufficient. Lending went from 2+ weeks to under 2 hours. We replaced the process, not the speed. Roles were eliminated because the work no longer existed — directly improving the efficiency ratio.
Same principle at Citi — compressing Digital Margin Lending from weeks to under an hour with same-day settlement.
2. The AI goes into production, not into a pilot.
Production means live transactions, at scale, with P&L accountability. Our credit decisioning models improved accuracy by 40% and cut review time by 65% — not in a sandbox, in the actual lending pipeline. The difference between a pilot and production is the difference between a lab experiment and a business outcome.
3. The board can see it on the income statement.
Theater generates slide decks. Transformation generates numbers your CFO can point to — $85M+ in operational cost efficiency, $15B+ in new lending volume, $25M+ in annualized infrastructure savings. If your AI investment cannot be traced to revenue, cost, or risk, it is not transformation. It is activity.
Institutions confuse deploying AI tools with redesigning AI-enabled operating models. One is procurement. The other is transformation.
When your board approves an AI budget, are they approving a tool purchase or a structural change to how your institution operates? Two completely different conversations — and most boards are only having one of them.