The AI War Chests: How Chase, BofA, and Citi Are Buying the Future
By Laurie McLachlan, Chief Marketing Officer at Revio Insight
Simple Subscribe
Subscribe Now!
Executive Summary
- The Big Three are spending billions to build deep competitive advantage in AI. They believe that mastery of AI at scale goes beyond greater efficiency to industry dominance.
- Nonetheless, Chase, BofA, and Citi are also pursuing significantly different AI strategies, reflecting their differing attitudes toward investment, ROI and risk.
- Smaller institutions risk getting left behind, but can find insight and inspiration for their own AI strategies by looking closely at what the Big Three are doing.
If you lead a bank or a credit union, here is the uncomfortable truth. While you sit in meetings asking if the ROI on AI is there, the megabanks are quietly locking in their forever edge.
The Big Three are already spending billions to build moats no one else can cross. JPMorgan Chase, Bank of America, and Citigroup are not testing the waters. They are betting that whoever masters AI at scale will not just get more efficient. They will seal a competitive advantage that others cannot touch.
Here’s a look at their strategies:
JPMorgan Chase: Building the AI Empire
Chase does not dabble. It invests about $18 billion a year in technology, with roughly $3 billion for AI. CEO Jamie Dimon has said AI’s impact could rival the printing press, electricity, and the internet.
Chase has more than 400 AI use cases in production. These range from fraud detection and marketing to compliance and wealth planning. They also built OmniAI, an internal platform that standardizes how models are deployed and governed. That means a new fraud model in credit cards and a customer churn model in retail both run through the same guardrails.
Chase CEO Jamie Dimon puts it plainly: “We took AI and data out of technology. It’s too important … we put AI at that management table.”
In its offshore engineering hubs, Chase reported developer productivity gains of 10 to 20% after rolling out its proprietary coding assistant. That was an early pilot applied to thousands of engineers, not the entire workforce, but in a bank with more than 60,000 technologists, the lift is still massive. Beyond engineering, Chase has pointed to revenue capture in sales teams, fraud detection savings, and client retention during volatility as evidence that AI is already delivering value.
| Strengths | Weak spots |
|---|---|
| Massive scale with capital and reach | Legacy systems are costly and complex, but Chase is unusual in that it is pouring billions into modernization rather than just layering AI on top of outdated cores. Retiring technical debt is painful in the short run but a long-term advantage. |
| Integrated platforms and unified toolsets | Integration across dozens of silos is still difficult even with new platforms. |
| Executive alignment at the highest level | Once the obvious use cases are deployed, incremental gains may be harder to capture. |
| Ability to absorb failures and iterate | Measuring the exact ROI of AI initiatives across such a vast operation is not always straightforward. |
Bank of America: The Guarded Innovator
Where Chase is bold, Bank of America is deliberate. It spends around $13 billion a year on technology with approximately $4 billion directed to new technology initiatives. It has earned a reputation for being methodical. Its AI brand is “Erica,” a chatbot launched years ago that now handles millions of customer interactions. In 2020, Bank of America launched Erica for Employees, an AI-driven internal virtual assistant which was rapidly adopted during the pandemic by employees seeking technology support in areas such as mobile device password reset, device activation, and many others. To date, more than 90% of the bank’s employees now use the AI assistant to improve efficiency, and every deployment is designed with human oversight, transparency and accountability.
BofA’s new tech leadership has also emphasized job-specific AI tools that sharpen customer service and employee productivity rather than broad, flashy experiments. The strategy is scale with discipline: give every employee AI, but only in ways that tie directly to measurable outcomes.
Bank of America has reported around 20% developer efficiency gains with generative AI tools. This is an average across its technology teams, according to the bank’s own reporting, and it highlights measurable improvements in delivery speed. Combined with Erica, which has handled more than 3 billion client interactions, BofA can point to proof of both internal efficiency and customer engagement.
“Erica has been learning from our clients for many years, enabling us to leverage AI today at scale, globally,” said Hari Gopalkrishnan, chief technology and information officer. “Our early and ongoing investments in AI demonstrate our commitment to delivering innovative experiences and value to clients.”
There’s a focus on intellectual property: One of the quieter but telling parts of BofA’s strategy is its investment in patents. The bank has more than 8,100 inventors in 14 countries and 42 U.S. states. By locking down intellectual property, BofA is signaling that it sees AI as a long-term differentiator worth protecting, not just a tool for short-term efficiency.
If Chase is constructing a palace, Bank of America is carefully restoring a historic building, checking every beam before moving forward.
| Strengths | Challenges |
|---|---|
| Stable, risk-aware execution | May move too slowly to capture disruptive moments |
| Deep IP and patent activity | Risk of missing first mover advantages |
| Culture of disciplined rollout | Revenue payoffs from AI are not yet as visible as efficiency wins |
| Ability to scale internal gains before expanding outward | Scaling from internal productivity to customer-facing impact will require bolder steps |
Citigroup: The Scrappy Experimenter
Citi has leaned into generative AI as a differentiator. Its leadership recently made major structural moves: three senior executives now co-sponsor Citi’s AI work globally, integrating AI strategy directly into business, operations, and technology.
“We are focusing on accelerating our AI strategy — connecting teams and partners, prioritizing resources and expediting use cases across our businesses and functions … In U.S. Personal Banking Operations, we’ve launched one of our first business GenAI pilots: Agent Assist and Enhanced IVR.”
Citi has rolled out AI tools to about 150,000 employees, with thousands already using them daily. In global locations, new tools like Citi Assist and Citi Stylus allow colleagues to search internal policies or summarize documents on the fly.
Tim Ryan, Citi’s Head of Technology and Business Enablement, described one of the tools this way: “It’s like having a super-smart coworker at your fingertips to help navigate commonly used policies and procedures across HR, risk, compliance, and finance.”
Early models in Agent Assist and enhanced IVR are showing tangible uplifts in call response times, reduced manual steps, and smoother agent workflows. The real test is scaling those pilots to millions of customer interactions.
Citi looks less like a cautious architect and more like a startup kid with a toolkit, trying ten projects at once and doubling down on what sticks.
| Strengths | Risks |
|---|---|
| Strong coordination at the executive level | Integration into legacy systems is hard |
| Rapid rollouts of enterprise AI tools | Many pilots yet unproven at full scale |
| Pilots in front-line client operations | Execution must match the ambition |
| Willingness to experiment and iterate | |
Head to Head
| Dimension | Chase (JPM) | Bank of America | Citi |
|---|---|---|---|
| Scale & Ambition | Extremely high, 400+ use cases | High but measured | Medium with rapid pilots |
| Platforms / Internal Tools | OmniAI standardizes model deployment | Strong internal tools, steady rollout | Multiple productivity tools, more siloed |
| Client Focus | Embedded into products and risk, expanding outward | Primarily internal, cautious expansion to clients | Boldly customer-facing (IVR, Agent Assist) |
| Research & Frontier | Bets Dedicated AI research, synthetic data, explainability | Heavy on patents, selective research | Early bets on agentic AI, customer pilots |
| Governance & Safety | Built into strategy and leadership | Conservative, risk-first | Industry standards involvement, but riskier |
| Results to Date | 10–20% developer lift in offshore hubs plus business gains in sales and fraud | ~20% average developer efficiency across teams, 1.5B+ client interactions through Erica | Faster call resolution, efficiency in trade processing, 150K employees equipped with AI |
Notes: BofA’s developer productivity figure is broader but self-reported. Chase’s figure is from a specific deployment and has been verified externally. Chase has also highlighted AI results in multiple business areas beyond engineering productivity.
How Can You Counterattack? Turn Data into Weapons That Work
1. Modernize the data foundation and make it usable.
Chase is investing billions in modernization because old cores are a dead weight. Smaller banks and credit unions cannot spend that much, but they can demand better access to their core data.
The key is not building predictive models in-house but using AI to translate raw transaction feeds into intelligence, revealing when customers or members are sending money to competitors, what products they hold elsewhere, and how big those relationships are.
Without that clarity, even the best frontline teams are selling blind.
2. Inventory and prioritize ROI bets first.
Start with use cases where the ROI is obvious and measurable. Chase’s COiN saved hundreds of thousands of legal hours. Its coding assistant freed up developer time. For smaller institutions, the equivalent could be simple: knowing which households are paying mortgages to another bank or moving deposits to Schwab. Those insights have an immediate dollar impact and can be tracked over time.
3. Spin up a “core AI platform” team early.
Even if you cannot match Chase’s OmniAI, you can avoid spaghetti pilots by creating a lightweight shared capability for interpreting and distributing insights. The goal is consistency and reuse, not a hundred disconnected vendor projects.
4. Begin with internal productivity before leaping to customer-facing tools.
Chase launched coding assistants and call center Q&A bots before rolling out client-facing AI.
Smaller banks and credit unions can follow the same pattern: Start with internal intelligence tools for bankers, marketers, and branch staff. Once those teams see the value, you can layer in AI-driven personalization in customer channels.
5. Elevate AI oversight to leadership, not IT backlog.
Jamie Dimon put AI and data “at the management table.” That is a signal. Institutions that bury AI under IT will treat it like plumbing. Institutions that elevate it to strategy will tie it to growth goals, deposit retention, and customer expansion.
6. Measure with rigor and kill fast.
Chase has been ruthless in measuring ROI: hours saved, productivity gains, new revenue captured. Smaller banks and credit unions should hold vendors and internal teams to the same standard. If a project cannot show measurable value, it should not survive.
