AI Absorption Ledger / BAC

Bank of America

BAC · Q2 2026 · reported 2026-07-14 · revenue $31.56bn

Assessment

In Q2 2026 Bank of America added a presentation slide on AI. It put about active users of general-purpose AI tools at more than prompts a day, more than use cases approved with live generative AI use cases, and role-specific tools: about developers with coding assistance raising productivity by more than , about call center agents receiving automated recommendations, and more than wealth professionals with AI meeting tools. None of these comes with a dollar figure, and the 10-Q still mentions AI only in its forward-looking factors.

Every size shown is the ledger’s own estimate. Internal build investment is the largest at , the 2026 initiatives budget converted to the quarter by the ledger’s fixed annual-plan band; the developer saving is , resting on management’s productivity floor with a low captured share because the CEO says the same money will get more code; operations is . Knowledge-work tools, call-center agent assistance and Erica are measured only in counts of users, prompts, agents and interactions, and are left unsized. Compensation and benefits rose on incentives while headcount fell , the direction the efficiency claims predict, but management names AI beside automation and calls headcount flat to slightly down, so the motives stay exploratory. Sales effectiveness is credited to AI beside advisor productivity and digital engagement with nothing to separate its part, and is left unsized.

Call-center agent assistance and capital raising and lending for the AI buildout open as channels; the coding assistants bill and developer productivity, silent in Q1, are first described in a covered quarter. The CFO says the buildout helps investment banking, markets and commercial loan growth without saying how much. The ledger reads the buildout at , a share of the rise in debt and equity issuance fees; the lending part, which the CFO calls broader than AI, is left unsized. The ledger reads this as the buildout’s money moving through a bank rather than absorption by the borrowers. Cyber risk from AI went unmentioned. Erica interactions rose to from , after a fall in Q1; the assistant stays relabelled, as the reference call already called it the company’s AI agent and explained its interactions by proactive alerts.

Sized channels against the income statement, Q2 2026

6 of 11 channels sized

Each blue mark is one channel's dollars for the quarter; a bar is the range of an estimate. Grey marks are the company's reported lines. The distance between them is the point: how large the AI channel is next to the line it sits in.

New money and old money, Q2 2026

2 new6 expanded3 relabelled

Each channel is tagged once for whether its money existed before language models, from the company's annual report and call at the start of the period. A bar splits one flow's sized dollars by that tag. The incremental total is the part that would not be there without the models: a new channel counts in full, an expanded one only for what AI added, a relabelled one at zero.

Paid for AI$50mn to $335mn sized

new $10mn to $35mnexpanded $40mn to $300mn

Incremental total $10mn to $335mnpoint $18mn$80mn in 1 channel has no traced baseline
Cost displaced by AI$9.1mn to $259mn sized

expanded $9.1mn to $259mnrelabelled not sized

Incremental total $9.1mn to $259mnpoint $46mn
Revenue arriving through AI$9.7mn to $121mn sized

expanded not sizedrelabelled $9.7mn to $121mn

Incremental total $0

The sized total counts every channel the company credits to AI, including relabelled money that existed before language models and the ledger's own estimates for it. The incremental total counts relabelled channels at zero. A flow is split by layer where its dollars sit at more than one: end use, compute sold to builders, and hardware. The same dollar can be a buyer's spend, a cloud's revenue and a chipmaker's revenue, so the layers are never added together.

Paid for AI3 channels · $50mn to $335mn sized · $10mn to $335mn incremental

vendor bill

Model access and general-purpose AI tool licences

0.03% to 0.08% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: described· motive: exploratory· before LLMs: new

The presentation counts about active users of general-purpose AI tools inside enterprise productivity applications, generating more than prompts a day, with access to multiple models in cloud hosting environments; the CEO says spending on AI will rise. No vendor or amount is given. Information processing and communications rose . The size shown is the ledger’s own estimate, : the Microsoft Copilot seats management counted on the 2025-CQ4 reference call () times an effective seat price, with model hosting for the use cases inside the high end.

Evidence: 6 quotes, 4 figures, 2 confounds, 1 from before coverage

What the company pays outside vendors for model access, model hosting and the general-purpose AI productivity tools its staff use, which management named as Microsoft’s Copilot on the 2025-CQ4 reference call (the CEO says all employees have access to AI; the Q2 2026 presentation shows general-purpose tools inside enterprise productivity applications and access to multiple models in cloud hosting environments, and the CEO names a sales force moving to an AI agent product). Coding assistants are read separately. The bill would sit in information processing and communications or professional fees; neither is split and no vendor or amount is disclosed.

Why this motive

Carried from Q1 on the same tells: broad access with use cases still being worked through (claims c2 and c17), and no measured return on the bill.

Before LLMs: new

The anchor names no model, token or AI licence bill; information processing and communications expense, the line that would hold one, was for FY2024 with no AI share named. Paying for model access and generative AI productivity licences could not exist without LLMs, so the channel is new whatever the anchor says. By January 2026 Microsoft’s Copilot assistant had been rolled out to employees.

“under the 365 Copilot rollout, which is now out across a total of 200,000 teammates and using it and learning from it”
CFO, qa, earnings call, 2026-01-14

Figures

  • Active users of general-purpose AI productivity tools (presentation) · as-of 2026-06-30
  • Prompts per day, a floor (presentation) · as-of 2026-06-30
  • Information processing and communications · 2026-CQ2
  • Information processing and communications growth, year over year · 2026-CQ2

What else could explain it

  • line composition: The bill would sit in information processing and communications or professional fees, which hold all software, data and communications costs.
  • bundling: Model access may be bought inside cloud or productivity software agreements and never billed separately.

Quotes

“In addition, we added a slide on AI. It's slide 20, which shows how our over 200,000 teammates are actively using AI-enabled capabilities across our company. These range from productivity tools to more advanced agentic workflows and coding support. Our associates are generating more than 400,000 prompts a day, and as of last week, we had over 300 AI use cases approved, all of which have good economics, of which 114 are live generative AI use cases.”
c2 · CEO, prepared remarks, earnings call, 2026-07-14
“Remember, we've been at this for a little while now, but you can begin to see now in these general purpose productivity tools or the tools that are aimed at specific functions like the bankers or the wealth professionals or the software developers.”
c16 · CFO, qa, earnings call, 2026-07-14
“I think we are spending at a good clip overall in technology and also dedicating a lot of time in the company towards careful examination implementation, catalyst, people working to understand the projects in AI. We gave you the outline on slide 20. I think there's productivity increases. There's a lot of spending. We're going to be spending more of it.”
c18 · CEO, qa, earnings call, 2026-07-14
“~200K active users generating >400K prompts per day”
c25 · Filing, presentation, 8-K earnings release, 2026-07-14
“Level I General Purpose AI Productivity Tools Integrated into enterprise-wide productivity applications”
c26 · Filing, presentation, 8-K earnings release, 2026-07-14
“Access to multiple models and enabled cloud hosting environments”
c29 · Filing, presentation, 8-K earnings release, 2026-07-14

By quarter

  • Q1 2026described · our inference · exploratory · $9.0mn to $24mn
  • Q2 2026described · our inference · exploratory · $9.0mn to $24mn

vendor bill

AI coding assistants for developers

0% to 0.04% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: described· motive: exploratory· before LLMs: new

First described in a covered quarter; Q1 carried it unmentioned on the reference call’s account of AI in coding work. The presentation says about developers use real-time coding assistance, and the CEO says coding has become more efficient with these tools. No vendor, price or amount is given. The size shown is the ledger’s own estimate, , developers times an effective price per developer from the shared coding-tools reference class.

Evidence: 3 quotes, 2 figures, 2 confounds, 1 from before coverage

The outside bill for the real-time coding assistance the Q2 2026 presentation says the company’s developers use. No vendor, price or amount is disclosed; it would sit inside information processing and communications. Separate from the productivity it is meant to pay for. Registered from Q1: the 2025-CQ4 reference call already described AI techniques in the coding work of the company’s developers, so Q1 carries it unmentioned.

Why this motive

The CEO says the same money will get more code (claim c19): a tool bought for capacity, with no displaced line, the exploratory tell.

Before LLMs: new

The anchor does not mention coding assistants or any AI tooling for developers; the line that would carry the bill, information processing and communications, was for FY2024. Before coverage, on the 2025-CQ4 reference call, management said AI techniques had been applied to the coding work of people writing code.

“We have 18,000 people on the company's payroll through code. And using the AI techniques, we've taken 30% out of the coding part of the stream of introducing a new product or service or change. That saves us about 2,000 people. So that's how we're applying it. That was this year's statistic, meaning 25.”
CFO, qa, earnings call, 2026-01-14

Figures

  • Developers using real-time coding assistance (presentation) · as-of 2026-06-30
  • Developer productivity increase from coding assistance, a floor (presentation) · as-of 2026-06-30

What else could explain it

  • bundling: Coding assistance may come inside a broader software or cloud agreement and never be billed per developer.
  • line composition: The bill would sit in information processing and communications, which holds all software costs.

Quotes

“Remember, we've been at this for a little while now, but you can begin to see now in these general purpose productivity tools or the tools that are aimed at specific functions like the bankers or the wealth professionals or the software developers.”
c16 · CFO, qa, earnings call, 2026-07-14
“Whether that increases expenditures of technology development dramatically or not, it really has to do with a couple of things. One is the shifting of spending towards it. Secondly, even the coding process has become more and more efficient using these tools. The same amount of money in 2027 will get us more code, for lack of a better term, in 2028.”
c19 · CEO, qa, earnings call, 2026-07-14
“~19K developers leveraging real-time coding assistance, increasing productivity >20%”
c27 · Filing, presentation, 8-K earnings release, 2026-07-14

By quarter

  • Q1 2026not mentioned · inscrutable · exploratory
  • Q2 2026described · our inference · exploratory · $1.1mn to $11mn

engineering

Internal investment in AI capability

0.13% to 0.95% of the quarter’s revenue

Incremental total: counts at zero at the point and in full at the high end, with no traced baseline.

Matched line moved : the whole line, not this channel.

our inferencedisclosure: described· motive: exploratory· before LLMs: expanded

Q2 added the AI slide: more than AI and machine learning use cases approved, live generative AI use cases and fully deployed. The CEO said the company spends at a good clip on technology, that spending is shifting toward AI and will rise, and gave no figure; the CFO called it something for the future. The size shown is the ledger’s own estimate, , on the same basis as Q1: the quarter’s share, by the fixed annual-plan band, of the 2026 initiatives budget of more than times an AI share set by management’s several hundred million for AI projects.

Evidence: 14 quotes, 6 figures, 2 confounds, 5 from before coverage

Internal money spent building AI into the company: the approved use cases, the corporate-wide effort the CEO calls catalyst, and the technology investment the CEO says is shifting toward AI. It sits in compensation and benefits (technologists) and information processing and communications, and part may be capitalized software. Separate from the outside vendor bills.

Why this motive

The CFO calls AI something for the future that the company is working its way through (claim c17), and the CEO says spending will rise (claim c18): investing ahead of a measured return, the exploratory tell in management’s words.

Before LLMs: expanded

At the anchor the CEO put annual technology investments at , and the annual report explained rising expense by investments in people, operations and technology and named AI among the emerging technologies the company must apply. No AI share was given then or since, so no quarter of the activity before LLMs can be traced. Before coverage, on the 2025-CQ4 reference call, management put technology initiatives for 2026 at more than , all new code, and AI project spending at several hundred million dollars. The size is the whole of an activity that existed before.

“And we've gone from $3 billion to $3.8 billion in annual technology investments across the last couple of years, more branches.”
CEO, qa, earnings call, 2024-04-16
“investments in people, operations and technology”
Filing, mdna, 10-K periodic report, 2025-02-25
“our businesses may be negatively impacted if we, or our third-party providers, do not timely development and apply emerging technologies, like AI and quantum computing, or if our initiatives in these areas are deficient or fail”
Filing, risk factors, 10-K periodic report, 2025-02-25
“Yeah. So Mike, we'll be up on initiatives this year, 5%-7%, I think, types of numbers. Total spending $13 billion plus $4 billion plus in initiatives. That's all new code.”
CFO, qa, earnings call, 2026-01-14
“So there's 20 different projects going on in the company. I don't know off the top of my head the total expenditure, but it's $ several hundred million.”
CFO, qa, earnings call, 2026-01-14

Figures

  • AI and machine learning use cases approved, a floor (presentation) · as-of 2026-06-30
  • Live generative AI use cases (presentation) · as-of 2026-06-30
  • Generative AI use cases fully deployed (presentation) · as-of 2026-06-30
  • Information processing and communications · 2026-CQ2
  • Information processing and communications growth, year over year · 2026-CQ2
  • Compensation and benefits growth, year over year · 2026-CQ2

Reported line it is matched to

Information processing and communications rose , more than a year earlier; the 10-Q attributes the expense increase to revenue-related expenses and investment in people, marketing and technology without naming AI.

2026-CQ2: 2025-CQ2:

What else could explain it

  • line composition: Neither compensation nor information processing and communications is split by purpose; most technology investment is not AI.
  • transformation program: Technology modernization and digital infrastructure are named beside AI-related initiatives in the same program of growth investments.

Quotes

“Third, we manage cost while we continue to invest in a franchise, our brand, our people, our technology, and our AI-enabled productivity.”
c1 · CEO, prepared remarks, earnings call, 2026-07-14
“In addition, we added a slide on AI. It's slide 20, which shows how our over 200,000 teammates are actively using AI-enabled capabilities across our company. These range from productivity tools to more advanced agentic workflows and coding support. Our associates are generating more than 400,000 prompts a day, and as of last week, we had over 300 AI use cases approved, all of which have good economics, of which 114 are live generative AI use cases.”
c2 · CEO, prepared remarks, earnings call, 2026-07-14
“34 of those cases are fully implemented, and we see new capabilities coming on every week. These tools are designed to help our customer relationship managers prepare more thoroughly for the client meetings. Our bankers automate the research and presentation materials. Our developers code more efficiently, and all our teammates improve productivity, consistency, and client service while creating significant opportunities ahead of us.”
c3 · CEO, prepared remarks, earnings call, 2026-07-14
“Non-interest expense on slide 10 was approximately $18.6 billion, up roughly $100 million from the first quarter and $1.4 billion from the second quarter of 2025, reflecting continued investment in technology, sales teams, financial centers, and brand marketing.”
c5 · CFO, prepared remarks, earnings call, 2026-07-14
“Client engagement remained broad-based with activity across capital markets, strategic transactions, liquidity management, we continued our program of growth investments, including technology modernization, digital infrastructure, and AI-related initiatives. We're also using AI-enabled tools to help bankers accelerate their research, prepare materials, and identify relevant client opportunities more efficiently.”
c10 · CFO, prepared remarks, earnings call, 2026-07-14
“Investments in technology and AI are helping teams deliver insights faster, operate more efficiently, and further strengthen our competitive position.”
c11 · CFO, prepared remarks, earnings call, 2026-07-14
“At this point, we put it at the top of the page just so you could see it, but you can see the number of approved model cases at this point is 300. You can see the number we've got in here that we're using, 114. This is going to be something for the future, and we're just working our way through it.”
c17 · CFO, qa, earnings call, 2026-07-14
“I think we are spending at a good clip overall in technology and also dedicating a lot of time in the company towards careful examination implementation, catalyst, people working to understand the projects in AI. We gave you the outline on slide 20. I think there's productivity increases. There's a lot of spending. We're going to be spending more of it.”
c18 · CEO, qa, earnings call, 2026-07-14
“Whether that increases expenditures of technology development dramatically or not, it really has to do with a couple of things. One is the shifting of spending towards it. Secondly, even the coding process has become more and more efficient using these tools. The same amount of money in 2027 will get us more code, for lack of a better term, in 2028.”
c19 · CEO, qa, earnings call, 2026-07-14
“What we learn from also is not only what happens outside, we see the impact on our company and the ability for us to use it effectively, relatively quickly. It's a very powerful tool. It has great utility. It has to be carefully managed. You have to have your data perfect.”
c24 · CEO, qa, earnings call, 2026-07-14
“>300 AI / machine learning use cases approved | 114 live generative AI use cases with 34 fully deployed”
c34 · Filing, presentation, 8-K earnings release, 2026-07-14
“The increases were primarily driven by higher revenue-related expenses during both periods, as well as continued investments in the business, including people, marketing and technology for the three-month period”
c37 · Filing, mdna, 10-Q periodic report, 2026-07-31
“the risks related to the development, implementation, use and management of emerging technologies, including artificial intelligence and the ability to achieve expected or potential benefits, such as increased productivity and cost savings”
c38 · Filing, risk factors, 10-Q periodic report, 2026-07-31
“Noninterest expense of $5.8 billion increased 4%, driven by investments in technology and brand”
c43 · Filing, press release, 8-K earnings release, 2026-07-14

By quarter

  • Q1 2026described · our inference · exploratory · $40mn to $300mn
  • Q2 2026described · our inference · exploratory · $40mn to $300mn

Cost displaced by AI5 channels · $9.1mn to $259mn sized · $9.1mn to $259mn incremental · 3 not sized

back office · expensive to verify

Operations and manual processing displaced by AI

0.01% to 0.31% of the quarter’s revenue

Incremental total: counts in full.

Matched line moved : the whole line, not this channel.

our inferencedisclosure: direction only· motive: exploratory· before LLMs: expanded

The CFO said AI-enabled tools are now embedded in operations, risk, finance and technology workflows and have helped reduce manual work, and the presentation says a document intelligence platform reduces manual work and lowers operating costs. Headcount fell over the year to ; the CFO calls it flat to slightly down in recent quarters. No saving is measured. The size shown is the ledger’s own estimate, .

Evidence: 8 quotes, 5 figures, 3 confounds, 5 from before coverage

Payroll in operations, servicing, risk and finance processing that AI tools displace or avoid: the manual work the CEO says the application of AI helps reduce, and the document intelligence platform the Q2 2026 presentation says reduces manual work and lowers operating costs. The line it would show in is compensation and benefits; headcount is the visible measure.

Why this motive

The exploratory tell in management’s words: the CFO calls AI something for the future that the company is working its way through (claim c17). In January the CFO described AI and digitalization as eliminating work in operational support while client-facing staff were added and headcount held flat (claim bac-anchor-c18): redeployment, not a shrinking line. Headcount fell over the year and compensation rose, so no displaced line visibly shrinks; carried from Q1.

Before LLMs: expanded

At the anchor the CEO credited an Operational Excellence platform and a headcount managed down as work was eliminated, and the annual report already named automation, AI and robotics among the technologies applied to data processes; compensation and benefits was for FY2024 with about employees. LLM tools in that work change its cost per unit, not the existence of the line. The size is the change AI made, not the whole line.

“First, our Operational Excellence platform continues to deliver and improve processes. These savings from that growth help fund the future growth in the company and lower the risk. Second, we managed headcount as we eliminated work.”
CEO, prepared remarks, earnings call, 2024-04-16
“take steps to leverage emerging technologies, such as automation, AI and robotics”
Filing, risk factors, 10-K periodic report, 2025-02-25
“investments in people, operations and technology”
Filing, mdna, 10-K periodic report, 2025-02-25
“Continued digital engagement savings and operational excellence initiatives should help us offset other cost increases for people and technology through the back half of the year.”
CFO, prepared remarks, earnings call, 2024-04-16
“We saw productivity improvements through AI and digitalization more generally, and those enabled us to add client-facing associates as we eliminated work and roles in our operational support areas.”
CFO, prepared remarks, earnings call, 2026-01-14

Figures

  • Headcount at quarter end · as-of 2026-06-30
  • Headcount a year earlier · as-of 2025-06-30
  • Headcount, change over the year · 2026-CQ2
  • Compensation and benefits · 2026-CQ2
  • Compensation and benefits growth, year over year · 2026-CQ2

Reported line it is matched to

Headcount fell by over the year to , the direction the claim predicts, while compensation and benefits rose on revenue-related incentives.

as-of 2026-06-30: as-of 2025-06-30:

What else could explain it

  • transformation program: The CFO names AI and automation together, and headcount management by attrition predates LLMs.
  • line composition: Compensation and benefits holds revenue-related incentives, which rose with investment banking, asset management and trading revenue.
  • operating leverage: Processing cost growing slower than volumes is ordinary scale at a bank of this size.

Quotes

“Non-interest expense on slide 10 was approximately $18.6 billion, up roughly $100 million from the first quarter and $1.4 billion from the second quarter of 2025, reflecting continued investment in technology, sales teams, financial centers, and brand marketing.”
c5 · CFO, prepared remarks, earnings call, 2026-07-14
“AI-enabled tools are now more embedded in workflows across operations, risk, finance, technology, and our client-facing teams. That's helped reduce manual work, improve speed, and enhance consistency for clients and teammates.”
c6 · CFO, prepared remarks, earnings call, 2026-07-14
“We also see meaningful opportunities to continue using AI and automation ourselves to improve productivity, strengthen client engagement, and support disciplined growth across the company.”
c13 · CFO, prepared remarks, earnings call, 2026-07-14
“You're asking a question that's really about sustainability going forward. Can it help us with our own operations? The answer is yes.”
c15 · CFO, qa, earnings call, 2026-07-14
“A lot of the incremental expense growth from second quarter this year, last year was due to incentives and BC&E clearing expenses.”
c20 · CEO, qa, earnings call, 2026-07-14
“I think if you look at the core, the best core measure is probably headcount. Our headcount discipline over the last six quarters has been excellent. It's flat to slightly down.”
c21 · CFO, qa, earnings call, 2026-07-14
“Enterprise document intelligence platform searches, classifies, extracts, and translates documents to accelerate client servicing, reduce manual work, and lower operating costs”
c30 · Filing, presentation, 8-K earnings release, 2026-07-14
“The increases were primarily driven by higher revenue-related expenses during both periods, as well as continued investments in the business, including people, marketing and technology for the three-month period”
c37 · Filing, mdna, 10-Q periodic report, 2026-07-31

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026direction only · our inference · exploratory · $4.4mn to $99mn

back office · cheap to verify

Knowledge-work productivity from general-purpose AI tools

Not sized

The only measures are counts: active users and prompts a day (claim c25), departmental templates (claim c45) and training courses (claim c44). No productivity rate is given for general-purpose tools, and a count sizes nothing (the methodology rule for counts: users, employees, interactions, documents and deals size nothing).

described, no sizedisclosure: direction only· motive: exploratory· before LLMs: expanded

The presentation counts about active users of general-purpose AI tools, more than prompts a day, more than departmental AI templates and more than AI training courses completed. Nothing is measured, and compensation and benefits rose , mostly on incentives. The measures are counts, so the channel is directional and unsized; the former estimate is no longer used.

Evidence: 15 quotes, 6 figures, 2 confounds, 4 from before coverage

Staff time saved by the general-purpose AI productivity tools available to all employees (prompts, departmental templates), which may show up as fewer hires or more output per head inside compensation and benefits. Excludes operations processing, developers, call-center agents and the advisor and banker tools, each read on its own channel.

Why this motive

Carried from Q1: broad use of general-purpose tools (claim c25) with no displaced line measured.

Before LLMs: expanded

The work is the existing staff cost inside compensation and benefits, for FY2024; the annual report already spoke of applying automation and AI, and the anchor measured no productivity effect from AI tools. The size is the change AI made, not the whole line.

“First, our Operational Excellence platform continues to deliver and improve processes. These savings from that growth help fund the future growth in the company and lower the risk. Second, we managed headcount as we eliminated work.”
CEO, prepared remarks, earnings call, 2024-04-16
“take steps to leverage emerging technologies, such as automation, AI and robotics”
Filing, risk factors, 10-K periodic report, 2025-02-25
“under the 365 Copilot rollout, which is now out across a total of 200,000 teammates and using it and learning from it”
CFO, qa, earnings call, 2026-01-14
“We saw productivity improvements through AI and digitalization more generally, and those enabled us to add client-facing associates as we eliminated work and roles in our operational support areas.”
CFO, prepared remarks, earnings call, 2026-01-14

Figures

  • Active users of general-purpose AI productivity tools (presentation) · as-of 2026-06-30
  • Prompts per day, a floor (presentation) · as-of 2026-06-30
  • Departmental AI templates in use, a floor (presentation) · as-of 2026-06-30
  • AI training courses completed, a floor (presentation) · as-of 2026-06-30
  • Compensation and benefits · 2026-CQ2
  • Compensation and benefits growth, year over year · 2026-CQ2

What else could explain it

  • line composition: Compensation and benefits holds revenue-related incentives, which rose with investment banking, asset management and trading revenue.
  • other: A productivity gain may be taken as more client capacity rather than lower cost.

Quotes

“Third, we manage cost while we continue to invest in a franchise, our brand, our people, our technology, and our AI-enabled productivity.”
c1 · CEO, prepared remarks, earnings call, 2026-07-14
“In addition, we added a slide on AI. It's slide 20, which shows how our over 200,000 teammates are actively using AI-enabled capabilities across our company. These range from productivity tools to more advanced agentic workflows and coding support. Our associates are generating more than 400,000 prompts a day, and as of last week, we had over 300 AI use cases approved, all of which have good economics, of which 114 are live generative AI use cases.”
c2 · CEO, prepared remarks, earnings call, 2026-07-14
“34 of those cases are fully implemented, and we see new capabilities coming on every week. These tools are designed to help our customer relationship managers prepare more thoroughly for the client meetings. Our bankers automate the research and presentation materials. Our developers code more efficiently, and all our teammates improve productivity, consistency, and client service while creating significant opportunities ahead of us.”
c3 · CEO, prepared remarks, earnings call, 2026-07-14
“AI-enabled tools are now more embedded in workflows across operations, risk, finance, technology, and our client-facing teams. That's helped reduce manual work, improve speed, and enhance consistency for clients and teammates.”
c6 · CFO, prepared remarks, earnings call, 2026-07-14
“We also see meaningful opportunities to continue using AI and automation ourselves to improve productivity, strengthen client engagement, and support disciplined growth across the company.”
c13 · CFO, prepared remarks, earnings call, 2026-07-14
“You're asking a question that's really about sustainability going forward. Can it help us with our own operations? The answer is yes.”
c15 · CFO, qa, earnings call, 2026-07-14
“Remember, we've been at this for a little while now, but you can begin to see now in these general purpose productivity tools or the tools that are aimed at specific functions like the bankers or the wealth professionals or the software developers.”
c16 · CFO, qa, earnings call, 2026-07-14
“A lot of the incremental expense growth from second quarter this year, last year was due to incentives and BC&E clearing expenses.”
c20 · CEO, qa, earnings call, 2026-07-14
“I think if you look at the core, the best core measure is probably headcount. Our headcount discipline over the last six quarters has been excellent. It's flat to slightly down.”
c21 · CFO, qa, earnings call, 2026-07-14
“~200K active users generating >400K prompts per day”
c25 · Filing, presentation, 8-K earnings release, 2026-07-14
“Level I General Purpose AI Productivity Tools Integrated into enterprise-wide productivity applications”
c26 · Filing, presentation, 8-K earnings release, 2026-07-14
“The increases were primarily driven by higher revenue-related expenses during both periods, as well as continued investments in the business, including people, marketing and technology for the three-month period”
c37 · Filing, mdna, 10-Q periodic report, 2026-07-31
“the risks related to the development, implementation, use and management of emerging technologies, including artificial intelligence and the ability to achieve expected or potential benefits, such as increased productivity and cost savings”
c38 · Filing, risk factors, 10-Q periodic report, 2026-07-31
“>600K AI training courses completed”
c44 · Filing, presentation, 8-K earnings release, 2026-07-14
“>2K departmental AI templates in use leveraging general purpose AI productivity tools”
c45 · Filing, presentation, 8-K earnings release, 2026-07-14

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026direction only · described, no size · exploratory

engineering · cheap to verify

Developer productivity from AI coding assistance

0.01% to 0.51% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: bounded· motive: exploratory· before LLMs: expanded

The presentation says about developers use real-time coding assistance, increasing productivity by more than , and the CEO says the same money will get more code. This is the first AI rate in a covered quarter, not the first the company has given: on the 2025-CQ4 reference call management said AI took out of the coding part of the change process, saving about people in 2025, and Q1 carried that unmentioned. The size shown is the ledger’s own estimate, , built on management’s developer-equivalents rather than on the slide’s floor, with the captured share kept low because management describes capacity.

Evidence: 5 quotes, 3 figures, 2 confounds, 3 from before coverage

Developer time saved by real-time coding assistance, which the Q2 2026 presentation says has raised the productivity of the company’s developers, with a floor on the gain; on the 2025-CQ4 reference call management said AI took a share out of the coding part of the change process in 2025 and saved about a stated number of developers. Registered from Q1, which carries it unmentioned. The CEO says the same money yields more code, so the gain may show as more output rather than lower cost. The line it would show in is compensation and benefits.

Why this motive

The presentation measures a developer productivity gain (claim c27), the start of an efficiency tell, but the CEO frames it as more code for the same money (claim c19) and no displaced line shrinks. The tells conflict, so the less durable motive stands: exploratory.

Before LLMs: expanded

Software development by the company’s own technologists was part of the technology investment the CEO put at a year at the anchor, paid largely inside compensation and benefits ( for FY2024); coding assistance tools are not mentioned. Before coverage, on the 2025-CQ4 reference call, management said AI took out of the coding part of the change process in 2025, saving about of its people writing code. The size is the change AI made, not the whole line.

“And we've gone from $3 billion to $3.8 billion in annual technology investments across the last couple of years, more branches.”
CEO, qa, earnings call, 2024-04-16
“Yeah. So Mike, we'll be up on initiatives this year, 5%-7%, I think, types of numbers. Total spending $13 billion plus $4 billion plus in initiatives. That's all new code.”
CFO, qa, earnings call, 2026-01-14
“We have 18,000 people on the company's payroll through code. And using the AI techniques, we've taken 30% out of the coding part of the stream of introducing a new product or service or change. That saves us about 2,000 people. So that's how we're applying it. That was this year's statistic, meaning 25.”
CFO, qa, earnings call, 2026-01-14

Figures

  • Developers using real-time coding assistance (presentation) · as-of 2026-06-30
  • Developer productivity increase from coding assistance, a floor (presentation) · as-of 2026-06-30
  • Compensation and benefits per head in the quarter · 2026-CQ2

What else could explain it

  • other: The productivity gain is management’s own measure, with no definition given; more code is not the same as lower cost.
  • line composition: Compensation and benefits holds revenue-related incentives, which rose with investment banking, asset management and trading revenue.

Quotes

“In addition, we added a slide on AI. It's slide 20, which shows how our over 200,000 teammates are actively using AI-enabled capabilities across our company. These range from productivity tools to more advanced agentic workflows and coding support. Our associates are generating more than 400,000 prompts a day, and as of last week, we had over 300 AI use cases approved, all of which have good economics, of which 114 are live generative AI use cases.”
c2 · CEO, prepared remarks, earnings call, 2026-07-14
“34 of those cases are fully implemented, and we see new capabilities coming on every week. These tools are designed to help our customer relationship managers prepare more thoroughly for the client meetings. Our bankers automate the research and presentation materials. Our developers code more efficiently, and all our teammates improve productivity, consistency, and client service while creating significant opportunities ahead of us.”
c3 · CEO, prepared remarks, earnings call, 2026-07-14
“Remember, we've been at this for a little while now, but you can begin to see now in these general purpose productivity tools or the tools that are aimed at specific functions like the bankers or the wealth professionals or the software developers.”
c16 · CFO, qa, earnings call, 2026-07-14
“Whether that increases expenditures of technology development dramatically or not, it really has to do with a couple of things. One is the shifting of spending towards it. Secondly, even the coding process has become more and more efficient using these tools. The same amount of money in 2027 will get us more code, for lack of a better term, in 2028.”
c19 · CEO, qa, earnings call, 2026-07-14
“~19K developers leveraging real-time coding assistance, increasing productivity >20%”
c27 · Filing, presentation, 8-K earnings release, 2026-07-14

By quarter

  • Q1 2026not mentioned · inscrutable · exploratory
  • Q2 2026bounded · our inference · exploratory · $4.7mn to $160mn

customer support · cheap to verify

Call-center work assisted by AI recommendations

Not sized

The only measure is a count of agents using the tool (claim c28); the CFO’s remark on new AI capabilities gives no measure and is not tied to agents (claim c8). A count sizes nothing (the methodology rule for counts: users, employees, interactions, documents and deals size nothing).

described, no sizedisclosure: direction only· motive: exploratory· before LLMs: expanded

A new channel. The presentation says about call center agents receive automated recommendations in real time, and the CFO says new AI capabilities improved service and efficiency. No cost or volume is given. The only measure is a count, so the channel is directional and unsized; the former estimate is no longer used.

Evidence: 3 quotes, 1 figure, 2 confounds, 2 from before coverage

Call-center payroll avoided because agents receive automated recommendations in real time, which the Q2 2026 presentation counts among multi-process AI-enabled systems. Separate from the client self-service Erica provides. The line it would show in is compensation and benefits.

Why this motive

The CFO says new AI capabilities have increased efficiency and let staff focus on higher-value interactions (claim c8), with no displaced line measured: the exploratory tell.

Before LLMs: expanded

At the anchor client service by phone and chat was existing work inside compensation and benefits ( for FY2024), Erica already resolved a share of chat inquiries automatically, and the CFO counted digital engagement savings among the levers offsetting people costs. Recommendations to agents from AI tools change the cost per contact, not the existence of the work. The size is the change AI made, not the whole line.

“Erica has resolved 43% of the CashPro chat inquiries automatically, demonstrating more and more clients are able to self-solve.”
CEO, prepared remarks, earnings call, 2024-04-16
“Continued digital engagement savings and operational excellence initiatives should help us offset other cost increases for people and technology through the back half of the year.”
CFO, prepared remarks, earnings call, 2024-04-16

Figures

  • Call-center agents receiving automated recommendations (presentation) · as-of 2026-06-30

What else could explain it

  • transformation program: Agent tools, routing and digital self-service predate LLMs; the presentation does not say which recommendations are generative.
  • other: Freed agent time may go to higher-value interactions, as the CFO says, rather than to lower cost.

Quotes

“We expanded our financial center network in new and growth markets, introduced new card products, and deployed new AI-enabled tools designed to enhance both the client and teammate experience.”
c7 · CFO, prepared remarks, earnings call, 2026-07-14
“Digital engagement remains a clear differentiator, with roughly 50 million active digital users, more than 24 million active Erica users, and digital sales representing 70% of total sales. New AI capabilities have improved service, increased efficiency, and allowed teammates to focus on higher-value client interactions.”
c8 · CFO, prepared remarks, earnings call, 2026-07-14
“~5K call center agents receiving automated recommendations with real-time results”
c28 · Filing, presentation, 8-K earnings release, 2026-07-14

customer support · cheap to verify

Client self-service through Erica

Not sized

The only measures are counts: active users (claim c39), interactions this quarter and a year earlier (claim c41) and CashPro chat interactions (claim c40). No cost, deflection rate or cost per contact is given, and the CFO’s remark on AI capabilities is not tied to Erica (claim c8). A count sizes nothing (the methodology rule for counts: users, employees, interactions, documents and deals size nothing).

described, no sizedisclosure: direction only· motive: product-defensive· before LLMs: relabelled

Active Erica users rose to , and interactions rose to from a year earlier, after a fall in Q1; the CFO said new AI capabilities have improved service and efficiency, without tying them to Erica. The channel stays relabelled: novelty is registered once, the rise is not attributed to a new capability, and before coverage the CEO renamed Erica the company’s AI agent while management explained its interactions by proactive alerts. The measures are counts, so the channel is directional and unsized; the former estimate is no longer used.

Evidence: 6 quotes, 7 figures, 3 confounds, 6 from before coverage

Service contacts that clients resolve through Erica, the virtual assistant in the consumer, wealth and CashPro platforms, instead of reaching a person. Erica predates LLMs; management names it in its AI answers and reports its users and interactions, not the cost avoided. The line the saving would show in is compensation and benefits.

Why this motive

Carried from Q1: an unpriced client feature; usage rose this quarter, but no cost avoided is given.

Before LLMs: relabelled

Erica was already in service before the anchor call, where the CEO put its cumulative interactions at more than and its automatic resolution of CashPro chat inquiries at . The covered sources credit AI with better service but attribute no measured change to it, so the tie-break reads the channel as the existing assistant described with AI named. On the 2025-CQ4 reference call the CEO renamed Erica the company’s AI agent, and management explained its falling interactions by proactive alerts, a cause other than AI.

“Erica, our virtual banking assistant, reached a key milestone of more than two billion interactions since its introduction about six years ago. It took four years to reach one billion interactions. It took just 18 months to reach the second billion.”
CEO, prepared remarks, earnings call, 2024-04-16
“Erica has resolved 43% of the CashPro chat inquiries automatically, demonstrating more and more clients are able to self-solve.”
CEO, prepared remarks, earnings call, 2024-04-16
“Continued digital engagement savings and operational excellence initiatives should help us offset other cost increases for people and technology through the back half of the year.”
CFO, prepared remarks, earnings call, 2024-04-16
“You should note there the impact of Zelle in the continued usage growth and also note the impact of Erica, our AI agent, and its use both across our businesses and with our teammates.”
CEO, prepared remarks, earnings call, 2026-01-14
“The reality of that is what we don't show you is the amount of alerts that we deliver. So you can set up alerts, which then has slowed down the need for Erica because the alerts are up to, I think, billions a quarter that are telling you when your balance is low and things like that to avoid you going and asking the question.”
CFO, qa, earnings call, 2026-01-14
“So as we told you at Investor Day, today's activity and Erica in our consumer business alone is worth thousands of teammates that we don't have to have to do the great work we do for the customers.”
CEO, qa, earnings call, 2026-01-14

Figures

  • Active Erica users · as-of 2026-06-30
  • Active Erica users, growth year over year · 2026-CQ2
  • Erica interactions in the quarter (presentation chart, last value, 2Q26) · 2026-CQ2
  • Erica interactions a year earlier (presentation chart, third value, 2Q25) · 2025-CQ2
  • Erica interactions, change over the year · 2026-CQ2
  • Erica interactions, growth year over year · 2026-CQ2
  • Interactions with CashPro Chat supported by Erica technology · 2026-CQ2

What else could explain it

  • relabel: Erica predates LLMs; the AI language may describe the same assistant.
  • other: The jump in active users may reflect a counting or product change; the release does not explain it.
  • other: Proactive alerts: management explained Erica interactions by alerts before coverage (claim bac-anchor-c16), a cause other than AI.

Quotes

“We expanded our financial center network in new and growth markets, introduced new card products, and deployed new AI-enabled tools designed to enhance both the client and teammate experience.”
c7 · CFO, prepared remarks, earnings call, 2026-07-14
“Digital engagement remains a clear differentiator, with roughly 50 million active digital users, more than 24 million active Erica users, and digital sales representing 70% of total sales. New AI capabilities have improved service, increased efficiency, and allowed teammates to focus on higher-value client interactions.”
c8 · CFO, prepared remarks, earnings call, 2026-07-14
“24.6 million active Erica® users, up 23%”
c39 · Filing, press release, 8-K earnings release, 2026-07-14
“40.1K interactions with CashPro® Chat, supported by Erica® technology”
c40 · Filing, press release, 8-K earnings release, 2026-07-14
“166 167 174 200 Erica® interactions (MM)”
c41 · Filing, presentation, 8-K earnings release, 2026-07-14
“Erica® Interactions (MM)5 2.7 3.0 3.4 3.8 2Q23 2Q24 2Q25 2Q26”
c42 · Filing, presentation, 8-K earnings release, 2026-07-14

By quarter

  • Q1 2026direction only · described, no size · product-defensive
  • Q2 2026direction only · described, no size · product-defensive

Revenue arriving through AI2 channels · $9.7mn to $121mn sized · $0 incremental · 1 not sized

sales force · expensive to verify

Advisor, banker and sales effectiveness from AI tools

Not sized

The CFO credits the wealth franchise’s benefit to advisor productivity, digital engagement and new AI-enabled tools together (claim c9) and the banking teams’ gains to investments in technology and AI (claim c11); the slide counts users and gives no lift, so nothing separates AI’s part (the methodology rule for AI named beside another cause).

described, no sizedisclosure: described· motive: exploratory· before LLMs: expanded

The presentation counts more than wealth professionals with AI meeting tools meant to increase advisor capacity, about corporate and investment bankers using AI for research and presentations, and about payments sales and servicing representatives using AI. The CFO credits the benefit to AI beside advisor productivity, digital engagement and technology investment, and nothing is measured, so the channel is left unsized. The counterparty is mixed: wealth households served by advisors and corporations served by bankers.

Evidence: 8 quotes, 5 figures, 2 confounds, 2 from before coverage

Revenue gained because AI tools help advisors prepare for client meetings and serve more clients, help corporate and investment bankers research and prepare materials, and help sales and servicing staff respond faster: the sales effectiveness the CFO names. The clients are wealth households and corporate clients, so the counterparty is mixed. Unpriced tools on existing advice, banking and payments revenue.

Why this motive

Tools meant to increase advisor capacity and client coverage (claims c31 and c32) with no measured lift, and a benefit credited to AI beside other causes (claim c9): AI named as a cause with no measure of its own, the exploratory tell. The sources do not contradict each other, so unknown does not apply.

Before LLMs: expanded

At the anchor relationship managers and advisors already provided financing solutions and advice, and total revenue was for FY2024 as first reported; the annual report named AI initiatives among the technologies the company must apply. The advice, banking and payments work existed; the LLM tools put into it during coverage (meeting preparation, research and presentation drafting) are new, so the channel is expanded and sized as an increment on the revenue it lifts. No lift has been measured. The size is the change AI made, not the whole line.

“Just like in our consumer business, we have seen good growth in customers seeking the benefits of both our physical and our online capabilities and also the care of our talented relationship managers who provide financing solutions and advice for our clients with global needs.”
CEO, prepared remarks, earnings call, 2024-04-16
“our businesses may be negatively impacted if we, or our third-party providers, do not timely development and apply emerging technologies, like AI and quantum computing, or if our initiatives in these areas are deficient or fail”
Filing, risk factors, 10-K periodic report, 2025-02-25

Figures

  • Wealth professionals equipped with AI meeting tools, a floor (presentation) · as-of 2026-06-30
  • Corporate and investment banking professionals using AI for research and presentations (presentation) · as-of 2026-06-30
  • Global Payments Solutions sales and servicing representatives using AI (presentation) · as-of 2026-06-30
  • Global Wealth & Investment Management total revenue · 2026-CQ2
  • Global Banking total revenue · 2026-CQ2

What else could explain it

  • relabel: Advisor and banker tools that rank leads and prepare materials predate LLMs.
  • mix shift: Wealth and Global Banking revenue grew on market levels, asset management flows and investment banking fees, which the filings explain without AI.

Quotes

“34 of those cases are fully implemented, and we see new capabilities coming on every week. These tools are designed to help our customer relationship managers prepare more thoroughly for the client meetings. Our bankers automate the research and presentation materials. Our developers code more efficiently, and all our teammates improve productivity, consistency, and client service while creating significant opportunities ahead of us.”
c3 · CEO, prepared remarks, earnings call, 2026-07-14
“Franchise continued to benefit from strong advisor productivity, growing digital engagement, and new AI-enabled tools that help advisors prepare for client conversations, identify opportunities, and deliver more personalized advice at scale.”
c9 · CFO, prepared remarks, earnings call, 2026-07-14
“Client engagement remained broad-based with activity across capital markets, strategic transactions, liquidity management, we continued our program of growth investments, including technology modernization, digital infrastructure, and AI-related initiatives. We're also using AI-enabled tools to help bankers accelerate their research, prepare materials, and identify relevant client opportunities more efficiently.”
c10 · CFO, prepared remarks, earnings call, 2026-07-14
“Investments in technology and AI are helping teams deliver insights faster, operate more efficiently, and further strengthen our competitive position.”
c11 · CFO, prepared remarks, earnings call, 2026-07-14
“Remember, we've been at this for a little while now, but you can begin to see now in these general purpose productivity tools or the tools that are aimed at specific functions like the bankers or the wealth professionals or the software developers.”
c16 · CFO, qa, earnings call, 2026-07-14
“>20K wealth professionals equipped with AI to assist with meeting-related activities to increase advisor capacity to serve additional clients”
c31 · Filing, presentation, 8-K earnings release, 2026-07-14
“~4K Corporate & Investment Banking professionals using AI for research and presentation preparation to enhance client coverage and growth opportunities”
c32 · Filing, presentation, 8-K earnings release, 2026-07-14
“~2.5K Global Payments Solutions sales and servicing representatives using AI to expedite response time to client inquiries”
c33 · Filing, presentation, 8-K earnings release, 2026-07-14

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026described · described, no size · exploratory

product revenue

Capital raising and lending for the AI buildout

0.03% to 0.38% of the quarter’s revenue

Incremental total: counts at zero.

our inferencedisclosure: described· motive: exploratory· before LLMs: relabelled

A new channel, registered on the CFO’s statement that the AI theme helps revenue because the company leads in capital raising and financing for the AI investment and infrastructure build, and that it has helped commercial loan growth, which is broader than AI. Investment banking fees rose to , which the 10-Q explains by issuance and advisory fees; debt and equity issuance, the capital raising the CFO names, rose ; Business Lending revenue fell to from on lower rates. The size shown is the ledger’s own estimate, , a share of the issuance fee rise only: lending is left unsized because the CFO says its growth is broader than the AI theme. It is the buildout’s money passing through a bank, not absorption by the clients.

Evidence: 7 quotes, 11 figures, 2 confounds

Investment banking and lending revenue from the AI capital buildout. The CFO says the company leads in capital raising and financing for that investment and that the AI theme has helped commercial loan growth, without saying how much. Lending and underwriting to builders predate LLMs; this is the buildout’s money moving through a bank, not absorption by the borrower.

Why this motive

The CFO says the AI theme helps investment banking, markets and commercial loan growth without separating it (claims c14 and c22), and the 10-Q explains the fee and lending lines without AI. AI named as a cause with no measure of its own is the exploratory tell; the sources do not contradict each other, so unknown does not apply.

Before LLMs: relabelled

Underwriting and lending to companies building capacity predate LLMs: investment banking fees were for FY2024, and the anchor does not mention data centers or AI borrowers. Management gives no measure of the AI part, so the tie-break reads it as existing financing described with AI named.

Figures

  • Investment banking fees, total corporation · 2026-CQ2
  • Investment banking fees, total corporation, prior year · 2025-CQ2
  • Investment banking fees against the prior year · 2026-CQ2
  • Investment banking fees growth, year over year · 2026-CQ2
  • Debt issuance fees, total corporation (gross, before self-led deals) · 2026-CQ2
  • Debt issuance fees, total corporation, prior year · 2025-CQ2
  • Equity issuance fees, total corporation (gross, before self-led deals) · 2026-CQ2
  • Equity issuance fees, total corporation, prior year · 2025-CQ2
  • Debt and equity issuance fees against the prior year · 2026-CQ2
  • Global Banking Business Lending revenue · 2026-CQ2
  • Global Banking Business Lending revenue, prior year · 2025-CQ2

What else could explain it

  • mix shift: Investment banking fees and lending revenue move with issuance, advisory activity, rates and balances across all clients, of which AI-buildout clients are an undisclosed part.
  • other: The CFO says loan growth is broader than AI and gives no AI share; the company discloses no data center or AI loan balance.

Quotes

“Overall, the U.S. economy has proved more durable than expected, supported by the strong consumer, ongoing AI-driven investments across the board, and easing energy costs, though inflation and tighter monetary policy remain key risks.”
c4 · CEO, prepared remarks, earnings call, 2026-07-14
“Activity remains healthy across lending, payments, investment banking, markets, and wealth management, including technology, digital infrastructure, and AI-related opportunities.”
c12 · CFO, prepared remarks, earnings call, 2026-07-14
“AI plays a role, I think, in two ways. The first one is on the revenue side. There's obviously a big AI theme going on in the world. We're leading in investment banking and Global Markets around capital raising, financing that massive capital investment and infrastructure build around the world. That's helping us there.”
c14 · CFO, qa, earnings call, 2026-07-14
“On the commercial loan growth, yes, it's broader than just an AI theme. The AI theme has helped because there's so much in the way of capital investment going on globally now.”
c22 · CFO, qa, earnings call, 2026-07-14
“As we look at underlying deals, we're always looking at the credit, the capability and the earnings power of the underlying tenants, so to speak, that are driving the revenue to the build-out. That we can keep care on.”
c23 · CEO, qa, earnings call, 2026-07-14
“Investment banking fees increased $710 million for the three-month period due to higher debt issuance, advisory and equity issuance fees.”
c35 · Filing, mdna, 10-Q periodic report, 2026-07-31
“Business Lending revenue decreased $141 million for the three months ended June 30, 2026 compared to the same period a year ago primarily driven by the impact of lower interest rates, partially offset by the benefit of higher average loan balances.”
c36 · Filing, mdna, 10-Q periodic report, 2026-07-31

Cost imposed, or revenue lost, by others’ AI1 channel · 1 not sized

other

Cyber defense against AI-enabled attacks

Not sized

No source this quarter mentions AI-enabled attacks or the cost of defending against them.

inscrutabledisclosure: not mentioned· motive: imposed· before LLMs: relabelled

Cyber risk from AI did not come up on the call, and the release and the 10-Q add nothing beyond forward-looking factors. The cost presumably continues; the Q1 ballpark rested on the CEO’s words that AI creates cybersecurity issues, which nothing this quarter restates.

Evidence: 0 quotes, 1 from before coverage

Security spending the company carries because attackers use AI, which the CEO says creates cybersecurity issues the company has invested heavily against. The cost sits in compensation (the cybersecurity staff) and information processing and communications and is not split.

Why this motive

Carried from Q1: a toll the company did not choose.

Before LLMs: relabelled

The FY2024 annual report already listed AI, including generative AI, among the technologies attackers may use to enhance their tactics; information processing and communications expense was for FY2024. No line is shown to move because of AI, so the tie-break reads it as the existing cyber budget described with AI named.

“from emerging technologies, such as AI (including machine learning and generative AI) and quantum computing, which may be used to enhance the tactics, techniques and procedures described above and facilitate new cyber threats”
Filing, risk factors, 10-K periodic report, 2025-02-25

By quarter

  • Q1 2026described · our inference · imposed · $1.9mn to $32mn
  • Q2 2026not mentioned · inscrutable · imposed

Reported lines, year-over-year growth

Revenue +19.3%

Q2 2026. Growing slower than revenue: total noninterest expense (+8.4%), compensation and benefits (+6.3%), information processing and communications (+5.8%), occupancy and equipment (+4.2%), professional fees (−10.5%). A displaced cost shows up as a line that stays under the dashed revenue line. These are the audited lines, as first reported; nothing here is attributed to AI by the filing.

Total noninterest expenseCompensation and benefitsInformation processing and communicationsOccupancy and equipmentMarketingProfessional feesRevenue
-20%-10%0%10%20%30%40%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026MarketingRevenueTotal noninterest expenseCompensation and benefitsInformation processing and communicationsOccupancy and equipmentProfessional fees
Reported values and filings
LineQ1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026
Total revenue, net of interest expense$27.37bn$26.46bn$28.09bn$31.18bn$30.27bn$31.56bn
Total noninterest expense$17.77bn$17.18bn$17.34bn$17.44bn$18.53bn$18.63bn
Compensation and benefits$10.89bn$10.33bn$10.52bn$10.60bn$11.33bn$10.99bn
Information processing and communications$1.89bn$1.82bn$1.83bn$1.91bn$2.02bn$1.92bn
Occupancy and equipment$1.86bn$1.84bn$1.87bn$1.88bn$1.90bn$1.91bn
Marketing$506mn$563mn$572mn$563mn$533mn$736mn
Professional fees$652mn$640mn$606mn$682mn$583mn$573mn