A Year of Bank AI: From Pilots to Production, Still a Cost Story

In a year, bank AI went from task forces to production, but almost entirely as a cost story. The banks now talking revenue, like Schwab, Citi, Truist, are the ones to watch.

We read the past four quarters of U.S. bank earnings calls, from Q2 2025 through Q1 2026, to see how management's AI narrative actually changed.

The one-year shift

Dimension	What changed over the year	Assessment & sources
Maturity of discussion	Banks moved from strategy, governance, and pilots toward production rollout and scaled adoption	More mature · · ·
Primary value proposition	Efficiency and operating leverage remained the dominant theme throughout	Still the core story · ·
Controls / governance	Governance stayed tightly linked to deployment, especially in fraud, AML, compliance, and risk	Caution remained high · ·
Employee adoption	Discussion became much more concrete, with firms citing usage rates, training, and enterprise tools	Adoption broadened materially · ·
Revenue / client use cases	Early in the year this was secondary; by early 2026 more firms were discussing client experience, lead generation, and revenue opportunities	Rising, but still earlier-stage than efficiency · ·

The pattern underneath all of it: everyone claims productivity, far fewer attach a number, and almost none mention revenue.

Quarter by quarter: how bank AI talk evolved

Fiscal quarter	What felt new vs. prior quarter	What it sounded like, with evidence
Quarter ended June 30, 2025	Banks were starting to move from broad ambition to named use cases, but most still sounded early	AI was framed mainly as enterprise capability-building: secure platforms, governance, task forces, pilots, and early workflow uses. Goldman described firm-approved GenAI tooling and piloting autonomous AI software agents . Bank of America cited 750 employees using AI agents in operations and 17,000 programmers using AI-enabled tools . F.N.B. emphasized a GenAI task force and responsible-risk controls .
Quarter ended September 30, 2025	More firms cited actual use-case counts, workflow embedding, and time savings	The narrative shifted toward broader employee adoption and measurable early benefits. Northern Trust said AI was embedded in 150+ use cases and saving tens of thousands of hours . BNY Mellon said AI was "for everyone, everywhere, and for everything," with digital employees already working alongside staff . Morgan Stanley described live and pilot use cases across coding, analytics, and lead distribution .
Quarter ended December 31, 2025	More firms cited deployment metrics, automation rates, or explicit productivity gains	AI discussion became more operational and more quantified. Customers Bancorp said every employee had been trained, over half the firm was using enterprise AI tools, and employees reported nearly 20% productivity gains . BNY Mellon said it deployed 130+ digital employees in 2025 . Bank of America cited a 30% reduction in coding time for product changes . First American said Sequoia AI was live in multiple markets with 40% automation in supported search/exam functions .
Quarter ended March 31, 2026	The newest shift was from "using AI" to "re-architecting how the firm works"	The conversation broadened from efficiency into enterprise operating models, client experience, revenue, and risk reduction. Texas Capital said ~80% of employees had accessed its AI platform in the prior four weeks and that agents were already in production for loan ops and fraud . Customers Bancorp cited 500+ agents/custom GPTs and 28,000+ hours saved . Schwab said all sales, service, and advice professionals were using AI daily, with client-facing assistants launching . Citi framed AI across revenue, productivity, and defensive risk uses .

The standouts: who leaned hardest into AI

These are the banks that leaned hardest into AI on their calls. It's a judgment call, not a scorecard. Ordered loosely, with the single best quote each one gave.

Bank	Why it stands out	Best concrete evidence
Bank of New York Mellon	Probably the clearest "enterprise AI at scale" story in the group	"218 AI solutions in production right now across the company — up four times year over year." Also said it has "digital employees working side by side with our teams."
Texas Capital	Very explicit production deployment plus broad employee usage	CEO said Ranger, its secure multi-LLM platform, is available to most employees; "about 80% of employees have accessed it in the last four weeks" and firm-wide agents are already in production for loan ops and fraud.
Customers Bancorp	Strong mix of adoption, internal buildout, and quantified productivity gains	CEO said 75% of team members have AI licenses and the bank has built "more than 500 agents and custom GPTs" while saving "more than 28,000 hours." ·
Banc of California	Broad rollout across the company, with especially strong developer adoption	Management said it had "nearly universal employee access" and "more than 80% of our developers using AI in their daily workflows."
State Street	Enterprise-wide embedding with scaling usage and productivity gains	Management said AI is "comprehensively embedded across the enterprise," with broad access, accelerating adoption, and usage becoming part of daily workflows.
First American Financial	Clear move from concept to production, with measurable process improvements	CEO said 25% of engineers are trained in agentic AI development and are moving "from concept to production in weeks rather than months." Also cited AI tools expanding QC capacity more than sixfold and reducing order processing time by roughly 30 minutes per file. ·
Charles Schwab	Strong client-facing and employee-facing adoption, plus monetization language	CEO said every sales, service and advice professional is using AI every day and that Schwab sees future monetization opportunities from AI-powered capabilities. ·
Citizens Financial	Very strong productivity evidence, including unusually large test results	Management said AI tools are already producing "30% improvement in productivity" and, in some tests, "5 to 10 times improvement in productivity."
Truist Financial	Broad application across consumer, wholesale, and advisory workflows	CEO said Truist is already deploying AI across Consumer and Small Business Banking and is using it in Wholesale to improve productivity, underwriting, and client engagement. ·
Northern Trust	Strong operating leverage and workflow integration, though less dramatic than the leaders above	Said AI is embedded in 150+ use cases and had saved tens of thousands of hours.

The one-year shift

Dimension	What changed over the year	Assessment & sources
Maturity of discussion	Banks moved from strategy, governance, and pilots toward production rollout and scaled adoption	More mature · · ·
Primary value proposition	Efficiency and operating leverage remained the dominant theme throughout	Still the core story · ·
Controls / governance	Governance stayed tightly linked to deployment, especially in fraud, AML, compliance, and risk	Caution remained high · ·
Employee adoption	Discussion became much more concrete, with firms citing usage rates, training, and enterprise tools	Adoption broadened materially · ·
Revenue / client use cases	Early in the year this was secondary; by early 2026 more firms were discussing client experience, lead generation, and revenue opportunities	Rising, but still earlier-stage than efficiency · ·

The pattern underneath all of it: everyone claims productivity, far fewer attach a number, and almost none mention revenue.

Quarter by quarter: how bank AI talk evolved

Fiscal quarter	What felt new vs. prior quarter	What it sounded like, with evidence
Quarter ended June 30, 2025	Banks were starting to move from broad ambition to named use cases, but most still sounded early	AI was framed mainly as enterprise capability-building: secure platforms, governance, task forces, pilots, and early workflow uses. Goldman described firm-approved GenAI tooling and piloting autonomous AI software agents . Bank of America cited 750 employees using AI agents in operations and 17,000 programmers using AI-enabled tools . F.N.B. emphasized a GenAI task force and responsible-risk controls .
Quarter ended September 30, 2025	More firms cited actual use-case counts, workflow embedding, and time savings	The narrative shifted toward broader employee adoption and measurable early benefits. Northern Trust said AI was embedded in 150+ use cases and saving tens of thousands of hours . BNY Mellon said AI was "for everyone, everywhere, and for everything," with digital employees already working alongside staff . Morgan Stanley described live and pilot use cases across coding, analytics, and lead distribution .
Quarter ended December 31, 2025	More firms cited deployment metrics, automation rates, or explicit productivity gains	AI discussion became more operational and more quantified. Customers Bancorp said every employee had been trained, over half the firm was using enterprise AI tools, and employees reported nearly 20% productivity gains . BNY Mellon said it deployed 130+ digital employees in 2025 . Bank of America cited a 30% reduction in coding time for product changes . First American said Sequoia AI was live in multiple markets with 40% automation in supported search/exam functions .
Quarter ended March 31, 2026	The newest shift was from "using AI" to "re-architecting how the firm works"	The conversation broadened from efficiency into enterprise operating models, client experience, revenue, and risk reduction. Texas Capital said ~80% of employees had accessed its AI platform in the prior four weeks and that agents were already in production for loan ops and fraud . Customers Bancorp cited 500+ agents/custom GPTs and 28,000+ hours saved . Schwab said all sales, service, and advice professionals were using AI daily, with client-facing assistants launching . Citi framed AI across revenue, productivity, and defensive risk uses .

The standouts: who leaned hardest into AI

These are the banks that leaned hardest into AI on their calls. It's a judgment call, not a scorecard. Ordered loosely, with the single best quote each one gave.

Bank	Why it stands out	Best concrete evidence
Bank of New York Mellon	Probably the clearest "enterprise AI at scale" story in the group	"218 AI solutions in production right now across the company — up four times year over year." Also said it has "digital employees working side by side with our teams."
Texas Capital	Very explicit production deployment plus broad employee usage	CEO said Ranger, its secure multi-LLM platform, is available to most employees; "about 80% of employees have accessed it in the last four weeks" and firm-wide agents are already in production for loan ops and fraud.
Customers Bancorp	Strong mix of adoption, internal buildout, and quantified productivity gains	CEO said 75% of team members have AI licenses and the bank has built "more than 500 agents and custom GPTs" while saving "more than 28,000 hours." ·
Banc of California	Broad rollout across the company, with especially strong developer adoption	Management said it had "nearly universal employee access" and "more than 80% of our developers using AI in their daily workflows."
State Street	Enterprise-wide embedding with scaling usage and productivity gains	Management said AI is "comprehensively embedded across the enterprise," with broad access, accelerating adoption, and usage becoming part of daily workflows.
First American Financial	Clear move from concept to production, with measurable process improvements	CEO said 25% of engineers are trained in agentic AI development and are moving "from concept to production in weeks rather than months." Also cited AI tools expanding QC capacity more than sixfold and reducing order processing time by roughly 30 minutes per file. ·
Charles Schwab	Strong client-facing and employee-facing adoption, plus monetization language	CEO said every sales, service and advice professional is using AI every day and that Schwab sees future monetization opportunities from AI-powered capabilities. ·
Citizens Financial	Very strong productivity evidence, including unusually large test results	Management said AI tools are already producing "30% improvement in productivity" and, in some tests, "5 to 10 times improvement in productivity."
Truist Financial	Broad application across consumer, wholesale, and advisory workflows	CEO said Truist is already deploying AI across Consumer and Small Business Banking and is using it in Wholesale to improve productivity, underwriting, and client engagement. ·
Northern Trust	Strong operating leverage and workflow integration, though less dramatic than the leaders above	Said AI is embedded in 150+ use cases and had saved tens of thousands of hours.