Emily Steele writes in The Financial Brand on why AI results track operational connectivity rather than model selection
Banks and credit unions have spent years on modernization. Systems were replaced, digital channels improved, customer access expanded. Much of that work delivered real progress. Writing in The Financial Brand, Savana President Emily Steele argues that AI is now revealing how much of it stopped short.
Her distinction is between modernizing and operationalizing. Those are not the same thing, and production AI makes the difference obvious. Pilots tolerate workarounds. Live customer journeys and banker workflows do not.
The Gaps Predate the AI
Steele’s observation is that AI initiatives across the industry keep running into the same conditions. Fragmented systems. Workflows that don’t connect. Customer context that changes depending on which channel you ask. Handoffs where knowledge quietly disappears.
None of those are AI problems. They are operational problems that were there the whole time. Modernization digitized parts of the institution while leaving the silos underneath intact, and employees became the connective tissue holding the gaps together.
“The gaps were always there. AI simply made them impossible to ignore.”
The difference, as Steele frames it, is that people can compensate for missing context through experience and institutional knowledge. AI cannot. It depends entirely on the operational infrastructure around it, which is why it surfaces the gaps almost immediately.
Why Results Vary So Widely
This is also Steele’s explanation for why AI outcomes differ so much between institutions running comparable technology. The differentiator is rarely the model. It is what the model has to work with.
Institutions seeing measurable gains tend to share the same traits: integrated systems, unified access to information, orchestrated workflows, consistent policy enforcement, and shared operational context. Layer AI onto a fragmented environment instead and the AI becomes fragmented too.
Steele also sees the industry moving from advisory intelligence toward operational intelligence. Advisory systems generate recommendations that someone has to review, govern, and reconcile against existing policy. Operational AI works inside existing workflows, approval paths, and permissions, inheriting the institution’s governance rather than requiring a new structure around it.
Her closing argument reframes the friction. Deployment problems are not failures. They are diagnostic, and each one identifies where operational maturity needs strengthening. Institutions that treat those signals as a roadmap get an improvement plan that benefits employees, customers, and every AI deployment that follows.
Share This: