AI Absorption Ledger / JPM

JPMorgan Chase

JPM · Q2 2026 · reported 2026-07-14 · revenue $57.35bn

Assessment

Q2 2026 is the quarter JPMorgan Chase started describing its own AI in operating terms. The CEO put use cases at almost , with that matter across risk, fraud, marketing, hedging, prospecting, note-taking and document reading, and said jobs were cut by or in discrete areas, with no period or area given and most staff redeployed; the CFO volunteered that token expense was a trivial number for the first half, with a meaningful acceleration forecast for the second. None of these comes with a dollar figure, and the 10-Q still mentions AI only in its list of forward-looking factors.

Every sized channel remains the ledger’s own estimate. Internal build investment is the largest spend at , and the model and token bill is small at , both read from other buyers’ readings as shares of revenue excluding the Visa and equity-investment gains (). Knowledge-work productivity is . The operations channel is directional and unsized: its only measures are shares of jobs cut, a count with no period or area. Total noninterest expense rose to on revenue-related pay, wage inflation, front-office hiring and technology investment, and employees still grew , so any AI saving shows as growth avoided, not as a shrinking line; the operations motive stays exploratory.

The CFO’s remarks on loan growth from capital spending came in answer to an analyst’s question about AI, and the CFO declined to tie the money to AI, so lending for the buildout is context, not a channel. Management’s frame runs against a margin story throughout: the CEO says the benefit of AI accrues to customers, which bears on price, not on whether the savings happen.

Sized channels against the income statement, Q2 2026

5 of 9 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

1 new4 expanded4 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$38mn to $809mn sized

new $2.6mn to $38mnexpanded $36mn to $771mn

Incremental total $2.6mn to $809mnpoint $13mn$203mn in 1 channel has no traced baseline
Cost displaced by AI$12mn to $360mn sized

expanded $12mn to $340mnrelabelled $0 to $20mn

Incremental total $12mn to $340mnpoint $61mn
Revenue arriving through AI$0 to $51mn sized

expanded not sizedrelabelled $0 to $51mn

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 AI2 channels · $38mn to $809mn sized · $2.6mn to $809mn incremental

vendor bill

Model and token bill

0% to 0.07% of the quarter’s revenue

Incremental total: counts in full.

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

The CFO volunteered the bill this quarter: token expense was a trivial number for the first half of the year, a meaningful acceleration is forecast for the second half, the full-year contribution is still trivial and partly budgeted, and the raised expense outlook does not rest on it. The firm routes work to the right model and uses open source where it suffices. The size shown is the ledger’s own estimate, , a share of revenue excluding the Visa and equity-investment gains () taken from the ledger’s readings of other buyers’ AI vendor bills; the word trivial is consistent with the range and does not set it.

Evidence: 6 quotes, 3 figures, 2 confounds, 1 from before coverage

What the firm pays outside model providers for tokens, model access and model testing. The CFO calls it token expense; it sits inside the technology, communications and equipment line or professional and outside services, and neither is split.

Why this motive

The CFO calls token expense trivial, budgeted and something the firm is spending time on, says the firm lags the cutting edge of adoption by design, and stresses using cheaper or open-source models where they suffice (claims c11, c13 and c15): a cost being studied for later, the exploratory tell.

Before LLMs: new

The anchor names data, cloud computing and technology vendors and no model or token bill; technology, communications and equipment expense was for FY2024 with no AI share named. A token bill could not exist without LLMs, so the channel is new whatever the anchor says.

“JPMorganChase also depends on its ability to access and use the operational systems of third parties, including its custodians, vendors (such as those that provide data and cloud computing services, and security and technology services) and other market participants”
Filing, risk factors, 10-K periodic report, 2025-02-14

Figures

  • Technology, communications and equipment expense · 2026-CQ2
  • Technology, communications and equipment expense growth, year over year · 2026-CQ2
  • Total net revenue excluding the Visa and equity-investment gains · 2026-CQ2

What else could explain it

  • line composition: The bill sits inside the technology line or professional and outside services, which also carry software, cloud, communications and auto lease depreciation.
  • bundling: Model access may be bought inside cloud or software agreements and never billed as tokens.

Quotes

“is the question of token expense. That is something that we're spending a bunch of time on, I think as probably pretty much everyone in corporate America is. Just for the avoidance of doubt, it is a trivial number for the first half of the year. We are forecasting some meaningful acceleration in that number for the second half of the year.”
c11 · CFO, qa, earnings call, 2026-07-14
“Still, nonetheless, the full year contribution of that is still trivial, and obviously we had budgeted some of that, so it's not in any way a meaningful driver of the current outlook or to the revision of the outlook.”
c12 · CFO, qa, earnings call, 2026-07-14
“We're, in a sense, like a representation of the economy as a whole, that we're probably lagging a little bit some of the cutting edge adoption and usage as we should given who we are as a company.”
c13 · CFO, qa, earnings call, 2026-07-14
“I think the good news is that we've done a lot of really high quality thinking on this, and a lot of the infrastructure that we've built over the last couple of years is going to position us to be quite sophisticated about using the right models for the right purpose.”
c14 · CFO, qa, earnings call, 2026-07-14
“As you know, the tools are quite good at doing that, and you really don't need the latest cutting edge incredibly expensive model to summarize an analyst report. The idea is use the right model for the right purpose, be smart about open source where appropriate, and ensure that you're getting value out of it ultimately.”
c15 · CFO, qa, earnings call, 2026-07-14
“higher investments in technology across the LOBs and Corporate and marketing in CCB,”
c27 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026described · our inference · exploratory · $2.5mn to $37mn
  • Q2 2026bounded · our inference · exploratory · $2.6mn to $38mn

engineering

Internal investment in AI capability

0.06% to 1.3% 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

The CEO said the firm spends quite a bit of money on AI, the whole company is working on it, and there are almost use cases, of them important; asked about returns on AI and other investments, the CEO called the spending a continuation of past years. The CFO said infrastructure built over the last couple of years lets the firm route work to the right models. Neither gave a figure. The technology line rose , which the 10-Q puts down to investment in technology across the businesses. The size shown is the ledger’s own estimate, , a share of revenue excluding the Visa and equity-investment gains taken from the ledger’s readings of other buyers’ build investment.

Evidence: 9 quotes, 8 figures, 3 confounds, 4 from before coverage

Internal money spent building AI into the firm: technologists building use cases and the platforms the CFO says were built over the last couple of years to route work to the right models. The lines it sits in are compensation (technology employees) and technology, communications and equipment. Separate from the outside model bill.

Why this motive

The CEO says the firm spends quite a bit on AI while its return remains to be seen and cannot be projected (claims c2, c9 and c8): investing for later without a measured return, the exploratory tell.

Before LLMs: expanded

At the anchor the firm already built machine learning and AI models across businesses and functions, and explained rising expense by investment in technology; technology, communications and equipment expense was and compensation for FY2024, with headcount growth mainly in front office and technology. No AI share of either line was given then or since, so no quarter of the activity before AI can be traced. The size is the whole of an activity that existed before.

“The Firm uses models and other analytical and judgment-based estimations, including those based upon machine learning or artificial intelligence techniques, across various businesses and functions.”
Filing, mdna, 10-K periodic report, 2025-02-14
“higher investments in technology in the businesses, as well as marketing, predominantly in CCB,”
Filing, mdna, 10-K periodic report, 2025-02-14
“The increase was primarily attributable to growth in the number of front office and technology employees.”
Filing, business, 10-K periodic report, 2025-02-14
“Expenses of $8.8 billion were up 9% year-over-year, largely driven by field compensation and continued growth in technology and marketing.”
CFO, prepared remarks, earnings call, 2024-04-12

Figures

  • AI use cases in the firm, per the CEO (almost 1,000) · as-of 2026-07-14
  • AI use cases the CEO calls really important · as-of 2026-07-14
  • Total net revenue excluding the Visa and equity-investment gains · 2026-CQ2
  • Total net revenue excluding the Visa and equity-investment gains, growth year over year · 2026-CQ2
  • Compensation expense · 2026-CQ2
  • Compensation expense growth, year over year · 2026-CQ2
  • Employees at quarter end · as-of 2026-06-30
  • Employees, change over the year · 2026-CQ2

Reported line it is matched to

Technology, communications and equipment expense rose against growth of in total net revenue excluding the Visa and equity-investment gains, more than a year earlier; the 10-Q attributes it to higher investment in technology across the businesses and Corporate.

2026-CQ2: 2025-CQ2: 2026-CQ2: 2025-CQ2:

What else could explain it

  • line composition: Compensation carries revenue-related pay and the technology line carries auto lease depreciation; neither is split by purpose.
  • transformation program: The filing explains the technology increase as investment across the businesses, a standing program the CEO calls a continuation of past years.
  • one time item: Total net revenue this quarter includes gains related to Visa shares and certain equity investments, which flatter the comparison with the technology line.

Quotes

“I think we've mentioned in the past, we spend quite a bit of money on it. We have a lot of MVPs that we know we have. The whole company's working this at this point.”
c2 · CEO, qa, earnings call, 2026-07-14
“I think there's almost 1,000 use cases today, though we'd say that the really important ones are 50 across risk, fraud, marketing, hedging, prospecting, note-taking, idea generation, document reading. It's kind of just starting.”
c3 · CEO, qa, earnings call, 2026-07-14
“AI will have its gives and takes. We can't project.”
c8 · CEO, qa, earnings call, 2026-07-14
“AI still remains to be seen because the other thing I think about AI, which is a little bit different than everybody else, is you don't uniquely benefit from AI. The ultimate beneficiary of AI will be our customers.”
c9 · CEO, qa, earnings call, 2026-07-14
“I would just say it's a complete continuation of what we've been doing for years. You shouldn't really expect any change.”
c10 · CEO, qa, earnings call, 2026-07-14
“We're, in a sense, like a representation of the economy as a whole, that we're probably lagging a little bit some of the cutting edge adoption and usage as we should given who we are as a company.”
c13 · CFO, qa, earnings call, 2026-07-14
“I think the good news is that we've done a lot of really high quality thinking on this, and a lot of the infrastructure that we've built over the last couple of years is going to position us to be quite sophisticated about using the right models for the right purpose.”
c14 · CFO, qa, earnings call, 2026-07-14
“In the end, either we're going to have a lot more capacity or we're going to have a lot more efficiency or both, or we're going to have better revenue outcomes, or we're going to compete more effectively, and we just need to be disciplined about how we handle that. That's a body of work that's happening right now.”
c16 · CFO, qa, earnings call, 2026-07-14
“higher investments in technology across the LOBs and Corporate and marketing in CCB,”
c27 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026described · our inference · exploratory · $34mn to $743mn
  • Q2 2026described · our inference · exploratory · $36mn to $771mn

Cost displaced by AI3 channels · $12mn to $360mn sized · $12mn to $340mn incremental · 1 not sized

customer support · expensive to verify

Operations and call-center work displaced by AI

Not sized

The only measure is a share of jobs cut by count, with no period or area (claim c4), and the headcount in operating and call centers comes from an answer on succession (claim c7). A count sizes nothing (the methodology rule for counts), and the CEO places the cuts in a long run of automation beside AI, so the channel is left unsized.

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

The CEO said AI will bring large efficiency in parts of the company and that jobs were cut by or in discrete areas, with most of those people offered jobs elsewhere; no period or area was given and the cuts were not tied to operations. In a separate answer, on succession, the CEO put of employees in operating centers and call centers. Firm employees rose , and the 10-Q attributes the compensation increase of to revenue-related pay, wage inflation and front-office hiring. Management never names operations or call centers as an AI area; the channel is the ledger’s reading of where the job cuts are most likely. The job-cut shares are counts, so the channel is directional and unsized; the former estimate is no longer used.

Evidence: 10 quotes, 10 figures, 4 confounds, 3 from before coverage

Payroll in operating centers and call centers (servicing, transaction processing, account management, customer support) that AI tools displace or avoid. Management never names operations or call centers as an AI area: the CEO says jobs were cut in discrete areas without naming them, and in a separate answer puts a large share of the workforce in operating and call centers. The channel is the ledger’s reading of where such cuts are most likely; the line it would show in is compensation expense.

Why this motive

The CEO expects large efficiency in parts of the company and says the work is just starting (claims c1 and c3); the job reductions (claim c4) have no stated period or area, most of those staff were offered other jobs, the CEO places them in a long run of automation, and firm headcount rose . With no displaced line visibly shrinking, the tells conflict and the less durable motive stands: exploratory, carried from Q1.

Before LLMs: expanded

At the anchor transaction processing, account management and customer support, by branch, digital and telephone, were existing work inside compensation ( for FY2024, employees at year end), and the annual report already said automation and AI may displace part of the workforce. 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.

“In addition, advances in technology, such as automation and artificial intelligence, may lead to workforce displacement.”
Filing, risk factors, 10-K periodic report, 2025-02-14
“providing services to clients and customers, including transaction processing, lending services, account management and customer support, or”
Filing, risk factors, 10-K periodic report, 2025-02-14
“Consumer & Community Banking offers products and services to consumers and small businesses through bank branches, ATMs, digital (including mobile and online) and telephone banking.”
Filing, mdna, 10-K periodic report, 2025-02-14

Figures

  • Job reduction in discrete areas, lower figure (period and areas not stated) · as-of 2026-07-14
  • Job reduction in discrete areas, higher figure (period and areas not stated) · as-of 2026-07-14
  • People in operating centers and call centers, per the CEO in an answer on succession · as-of 2026-07-14
  • Employees worldwide, per the CEO in an answer on succession · as-of 2026-07-14
  • Operating and call-center staff as a share of employees, per the CEO · as-of 2026-07-14
  • Consumer & Community Banking compensation per employee relative to the firm · 2026-CQ2
  • Compensation expense · 2026-CQ2
  • Compensation expense growth, year over year · 2026-CQ2
  • Employees at quarter end · as-of 2026-06-30
  • Employees, change over the year · 2026-CQ2

What else could explain it

  • operating leverage: Operations cost growing slower than volumes is ordinary scale at a bank of this size.
  • line composition: Compensation holds front-office and revenue-related pay, which rose for reasons unrelated to AI.
  • transformation program: The CEO places the job reductions in a long run of automation with mainframes, APIs and other tools, so part of what is credited to AI may be that program.
  • other: The CEO expects the same tools to reach smaller competitors through processors and fintechs and the benefit to pass to customers as price; that changes who keeps the saving, not whether it happens.

Quotes

“It's not a sensitive topic at all. We are going to use AI to do a better job for our clients. That's our job. We fully expect it'll have huge efficiency in certain parts of the company. We analyze it all the time.”
c1 · CEO, qa, earnings call, 2026-07-14
“We are preparing to make sure we can retrain our people. We have had discrete areas where we did reduce jobs by 30% or 40%, and most of those people offered jobs elsewhere.”
c4 · CEO, qa, earnings call, 2026-07-14
“You want it across the whole company, not weighted to investment banking or trading or just big CEOs, but also weighted to the fact that we've got 300,000 employees around the world. In our branches, we have 50,000 top-notch people. In our operating centers and our call centers, we have 150,000 people.”
c7 · CEO, qa, earnings call, 2026-07-14
“I also think that over time, remember, this will be offered to smaller competitors too, through Fiserv and FIS and other fintech companies.”
c5 · CEO, qa, earnings call, 2026-07-14
“I do also want to point out, maybe you could be ahead of other people, kind of, but what always happens is the benefit accrues to the customer, not to JPMorgan in this case.”
c6 · CEO, qa, earnings call, 2026-07-14
“AI will have its gives and takes. We can't project.”
c8 · CEO, qa, earnings call, 2026-07-14
“AI still remains to be seen because the other thing I think about AI, which is a little bit different than everybody else, is you don't uniquely benefit from AI. The ultimate beneficiary of AI will be our customers.”
c9 · CEO, qa, earnings call, 2026-07-14
“In the end, either we're going to have a lot more capacity or we're going to have a lot more efficiency or both, or we're going to have better revenue outcomes, or we're going to compete more effectively, and we just need to be disciplined about how we handle that. That's a body of work that's happening right now.”
c16 · CFO, qa, earnings call, 2026-07-14
“growth in the number of employees, primarily front office employees.”
c25 · Filing, mdna, 10-Q periodic report, 2026-08-06
“the impact of wage inflation, and”
c26 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026not mentioned · inscrutable · exploratory
  • Q2 2026direction only · described, no size · exploratory

back office · cheap to verify

Knowledge-work productivity from AI tools

0.02% to 0.59% of the quarter’s revenue

Incremental total: counts in full.

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

The CEO listed note-taking, idea generation and document reading among the important use cases, and the CFO described staff summarizing analyst reports with cheaper models. Nothing was measured, and compensation rose on revenue-related pay, wage inflation and front-office hiring. The size shown is the ledger’s own estimate, .

Evidence: 10 quotes, 2 figures, 2 confounds, 2 from before coverage

Front-office and corporate-function time saved by AI tools for note-taking, document reading, summarizing and idea generation, which may show up as fewer hires or more output per head inside compensation expense. Separate from operations and call-center work.

Why this motive

The knowledge-work uses the CEO lists (note-taking, document reading, idea generation; claim c3) come with no displaced line, and the CFO puts capacity ahead of efficiency among possible outcomes (claim c16); the job cuts are not tied to these roles. The exploratory tell is the closer one.

Before LLMs: expanded

The work is the existing front and middle office inside compensation, for FY2024; the annual report already offered AI training to employees and expected automation and AI to displace some roles. No productivity effect was measured at the anchor. The size is the change AI made, not the whole line.

“In addition, advances in technology, such as automation and artificial intelligence, may lead to workforce displacement.”
Filing, risk factors, 10-K periodic report, 2025-02-14
“In addition, the Firm offers extensive voluntary training programs and educational resources to all employees covering a broad variety of topics such as leadership and management, artificial intelligence, data literacy and operational and professional skills.”
Filing, business, 10-K periodic report, 2025-02-14

Figures

  • Compensation expense · 2026-CQ2
  • Compensation expense growth, year over year · 2026-CQ2

What else could explain it

  • line composition: Compensation holds revenue-related pay that moves with markets and banking results.
  • other: Management says the benefit passes to clients through competition, so a productivity gain may show as price rather than cost.

Quotes

“It's not a sensitive topic at all. We are going to use AI to do a better job for our clients. That's our job. We fully expect it'll have huge efficiency in certain parts of the company. We analyze it all the time.”
c1 · CEO, qa, earnings call, 2026-07-14
“I think there's almost 1,000 use cases today, though we'd say that the really important ones are 50 across risk, fraud, marketing, hedging, prospecting, note-taking, idea generation, document reading. It's kind of just starting.”
c3 · CEO, qa, earnings call, 2026-07-14
“We are preparing to make sure we can retrain our people. We have had discrete areas where we did reduce jobs by 30% or 40%, and most of those people offered jobs elsewhere.”
c4 · CEO, qa, earnings call, 2026-07-14
“I also think that over time, remember, this will be offered to smaller competitors too, through Fiserv and FIS and other fintech companies.”
c5 · CEO, qa, earnings call, 2026-07-14
“I do also want to point out, maybe you could be ahead of other people, kind of, but what always happens is the benefit accrues to the customer, not to JPMorgan in this case.”
c6 · CEO, qa, earnings call, 2026-07-14
“AI still remains to be seen because the other thing I think about AI, which is a little bit different than everybody else, is you don't uniquely benefit from AI. The ultimate beneficiary of AI will be our customers.”
c9 · CEO, qa, earnings call, 2026-07-14
“As you know, the tools are quite good at doing that, and you really don't need the latest cutting edge incredibly expensive model to summarize an analyst report. The idea is use the right model for the right purpose, be smart about open source where appropriate, and ensure that you're getting value out of it ultimately.”
c15 · CFO, qa, earnings call, 2026-07-14
“In the end, either we're going to have a lot more capacity or we're going to have a lot more efficiency or both, or we're going to have better revenue outcomes, or we're going to compete more effectively, and we just need to be disciplined about how we handle that. That's a body of work that's happening right now.”
c16 · CFO, qa, earnings call, 2026-07-14
“growth in the number of employees, primarily front office employees.”
c25 · Filing, mdna, 10-Q periodic report, 2026-08-06
“the impact of wage inflation, and”
c26 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026not mentioned · inscrutable · exploratory
  • Q2 2026described · our inference · exploratory · $12mn to $340mn

back office · expensive to verify

Fraud and risk losses reduced by AI models

0% to 0.04% of the quarter’s revenue

Incremental total: counts at zero.

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

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

The CEO listed risk and fraud among the most important AI use cases. Operating losses, primarily fraud losses, were against a year earlier, with no cause given in the 10-Q. The size shown is the ledger’s own estimate, , a share of that fall.

Evidence: 1 quote, 3 figures, 2 confounds, 4 from before coverage

Fraud and scam losses avoided because AI models detect fraudulent transactions and risk earlier. The line it would show in is operating losses, which the firm says are primarily fraud losses on deposit accounts and cards, inside other expense.

Why this motive

Carried from Q1 on the same tells: fraud and risk are among the most important use cases (claim c3) and the line that holds fraud losses fell again, though the filing does not attribute the fall.

Before LLMs: relabelled

At the anchor the firm already used machine learning and AI models across functions and ran data-driven fraud detection; operating losses, primarily fraud losses, were for FY2024. The covered calls credit AI with reducing fraud and give no measured change, so the tie-break reads the channel as existing modelling under a new name.

“The increasing sophistication of artificial intelligence technologies poses a greater risk of identity fraud, as malicious actors may exploit artificial intelligence to create convincing false identities or manipulate verification processes. This challenge necessitates ongoing enhancements to client verification systems and security protocols to prevent unauthorized access and protect sensitive client information.”
Filing, risk factors, 10-K periodic report, 2025-02-14
“The Firm uses models and other analytical and judgment-based estimations, including those based upon machine learning or artificial intelligence techniques, across various businesses and functions.”
Filing, mdna, 10-K periodic report, 2025-02-14
“monitoring and detecting fraudulent transactions and cyber threats”
Filing, risk factors, 10-K periodic report, 2025-02-14
“Operating losses: Primarily refer to fraud losses associated with customer deposit accounts, credit and debit cards; exclude legal expense.”
Filing, notes, 10-K periodic report, 2025-02-14

Figures

  • Operating losses (primarily fraud losses) · 2026-CQ2
  • Operating losses, prior year · 2025-CQ2
  • Fall in operating losses against the prior year (negative: a rise) · 2026-CQ2

Reported line it is matched to

Operating losses fell by year over year, the direction the claim predicts; the filing does not attribute the fall.

2026-CQ2: 2025-CQ2:

What else could explain it

  • relabel: Fraud detection models predate LLMs; the AI language may describe the same systems.
  • other: Operating losses move with card and deposit volumes, recoveries and the timing of charge recognition.

Quotes

“I think there's almost 1,000 use cases today, though we'd say that the really important ones are 50 across risk, fraud, marketing, hedging, prospecting, note-taking, idea generation, document reading. It's kind of just starting.”
c3 · CEO, qa, earnings call, 2026-07-14

By quarter

  • Q1 2026described · our inference · efficiency · $0 to $60mn
  • Q2 2026described · our inference · efficiency · $0 to $20mn

Revenue arriving through AI3 channels · $0 to $51mn sized · $0 incremental · 2 not sized

marketing · cheap to verify

Prospecting, marketing and offers improved by AI

0% to 0.09% of the quarter’s revenue

Incremental total: counts at zero.

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

The CEO listed marketing and prospecting among the most important AI use cases; nothing is measured, and the 10-Q credits consumer marketing spend without naming AI. The size shown is the ledger’s own estimate, .

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

Revenue gained because AI applied to client data improves prospecting, marketing and the offers shown to customers (travel and offers in what the firm calls Connected Commerce). Mostly the consumer bank; prospecting also reaches business clients.

Why this motive

Marketing and prospecting are named as important use cases (claim c3) with no measured lift. AI named as the cause with no measure and no line moving is exploratory, as in Q1.

Before LLMs: relabelled

At the anchor marketing was a growing investment ( for FY2024) and models built on machine learning or AI informed business decisions; card services already offered travel. No lift attributed to AI has been measured then or since.

“The Firm uses models and other analytical and judgment-based estimations, including those based upon machine learning or artificial intelligence techniques, across various businesses and functions.”
Filing, mdna, 10-K periodic report, 2025-02-14
“higher investments in technology in the businesses, as well as marketing, predominantly in CCB,”
Filing, mdna, 10-K periodic report, 2025-02-14
“Expenses of $8.8 billion were up 9% year-over-year, largely driven by field compensation and continued growth in technology and marketing.”
CFO, prepared remarks, earnings call, 2024-04-12

Figures

  • Consumer & Community Banking total net revenue · 2026-CQ2

What else could explain it

  • relabel: Model-driven marketing and prospecting predate LLMs.
  • mix shift: Consumer revenue growth came from card balances, auto leases and wealth management fees, which the release explains without AI.

Quotes

“I think there's almost 1,000 use cases today, though we'd say that the really important ones are 50 across risk, fraud, marketing, hedging, prospecting, note-taking, idea generation, document reading. It's kind of just starting.”
c3 · CEO, qa, earnings call, 2026-07-14
“higher investments in technology across the LOBs and Corporate and marketing in CCB,”
c27 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026described · our inference · exploratory · $0 to $49mn
  • Q2 2026described · our inference · exploratory · $0 to $51mn

product revenue · expensive to verify

AI cash-management tool for consumers

Not sized

The CEO says the tool is still a test case and that tests are still to come (claims c17 and c18); a product in test earns nothing in the quarter and is never sized at zero (the methodology rule for money that has not started).

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

Asked about the smart cash tool, the CEO said it is still a test case for a narrow segment of accounts, that tests will come out and that something may be seen this year. The answer did not mention AI; the CFO named AI in the tool in Q1. The tool is still in test, so the channel is left unsized; the former pilot-scale estimate is no longer used.

Evidence: 2 quotes, 1 figure, 1 confound, 2 from before coverage

A tool, described by the CFO as having AI in it, that helps a narrow segment of clients with investments move money between checking, deposits and investments, aimed at a larger share of their investment wallet. Not live in Q1 2026 and a test case in Q2 2026; unpriced, so it is a feature on existing deposit and investment products.

Why this motive

The CEO calls it a test case from which the firm expects to learn (claims c17 and c18): the exploratory tell, in management’s words.

Before LLMs: expanded

At the anchor the consumer bank already offered deposit, investment and cash management products through digital channels; total net revenue was for FY2024. The tool would change how much of a client’s cash and investments sits with the firm, not create the products. The size is the change AI made, not the whole line.

“Consumer & Community Banking offers products and services to consumers and small businesses through bank branches, ATMs, digital (including mobile and online) and telephone banking.”
Filing, mdna, 10-K periodic report, 2025-02-14
“Banking & Wealth Management offers deposit, investment and lending products, cash management, payments and services.”
Filing, mdna, 10-K periodic report, 2025-02-14

Figures

  • Consumer & Community Banking total net revenue · 2026-CQ2

What else could explain it

  • other: The tool may move client cash out of low-rate deposits into investments, so revenue gained on investments may come with deposit margin lost.

Quotes

“This is still a test case. Banks, people are in a different position, and if you actually look at accounts, these don't relate to every account. They relate to a narrow segment of accounts and where you're competing for their investment business and their deposit business.”
c17 · CEO, qa, earnings call, 2026-07-14
“What you're going to see is certain tests coming out, and then you'll find out about what we can do, what we can't do, and we're going to learn a lot by doing some of that. We think it could be good for customers and good for us.”
c18 · CEO, qa, earnings call, 2026-07-14

By quarter

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

product revenue

Lending and financing for the AI buildout

Not sized

Withdrawn from the ledger as not an AI channel: a lending or financing door is a channel only where the company itself ties the money to AI, not where AI enters only through an analyst’s question. AI entered through the analyst’s question on capital spending, and the CFO declined to tie the loan growth to AI (claims c19 and c20, now context).

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

The CFO’s remarks on capital spending and loan growth, the CEO’s data center remarks and the release’s line on AI-driven capital investment are kept as context. None of them ties the firm’s own revenue to AI, so the channel is read silent and unsized.

Evidence: 0 quotes, 1 from before coverage

Withdrawn from the ledger on 2026-10-06 as not an AI channel, under the rule that a lending or financing door is a channel only where the company itself ties the money to AI: AI entered only through an analyst’s question, and the CFO declined to tie the loan growth to AI. Registered in Q2 2026 as lending revenue from the AI capital buildout; the passages are kept as context.

Why this motive

No source ties the firm’s revenue to AI, so no motive is read from this quarter; exploratory is kept from the channel’s registration.

Before LLMs: relabelled

Wholesale lending was already growing at the anchor, and the Commercial & Investment Bank earned of revenue in FY2024; 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 lending described with AI named.

“higher wholesale loans in CIB, and”
Filing, mdna, 10-K periodic report, 2025-02-14

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 boilerplate risk factors. The cost presumably continues; the Q1 ballpark rested on the CEO’s words that AI has made the risk worse, and nothing this quarter restates it.

Evidence: 0 quotes, 4 from before coverage

Security spending the firm carries because attackers use AI, which the CEO says has made cyber risk worse and created additional vulnerabilities. The cost sits in technology and compensation lines and is not split.

Why this motive

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

Before LLMs: relabelled

The FY2024 annual report already listed generative AI used by malicious actors among the reasons breach risk could rise, and called cybersecurity a continuing investment; technology, communications and equipment 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.

“advances in artificial intelligence, such as the use of machine learning, generative artificial intelligence and quantum computing by malicious actors to develop more advanced social engineering attacks, including targeted phishing attacks”
Filing, risk factors, 10-K periodic report, 2025-02-14
“The increasing sophistication of artificial intelligence technologies poses a greater risk of identity fraud, as malicious actors may exploit artificial intelligence to create convincing false identities or manipulate verification processes. This challenge necessitates ongoing enhancements to client verification systems and security protocols to prevent unauthorized access and protect sensitive client information.”
Filing, risk factors, 10-K periodic report, 2025-02-14
“The dynamic nature of the cyber threat landscape, including the pace of innovation and increased threat of novel attack methods, necessitates ongoing investment in, as well as enhancement and adaptation of, cybersecurity controls”
Filing, risk factors, 10-K periodic report, 2025-02-14
“monitoring and detecting fraudulent transactions and cyber threats”
Filing, risk factors, 10-K periodic report, 2025-02-14

By quarter

  • Q1 2026described · our inference · imposed · $3.0mn to $68mn
  • Q2 2026not mentioned · inscrutable · imposed

Reported lines, year-over-year growth

Revenue +27.7%

Q2 2026. Growing slower than revenue: total noninterest expense (+14.9%), compensation expense (+10.6%), technology, communications and equipment expense (+14.9%), occupancy expense (+17.2%). 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 expenseTechnology, communications and equipment expenseProfessional and outside servicesMarketingOccupancy expenseRevenue
-20%-10%0%10%20%30%40%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026MarketingProfessional and outside servicesRevenueOccupancy expenseTechnology, communications and equipment expenseTotal noninterest expenseCompensation expense
Reported values and filings
LineQ1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026
Total net revenue$45.31bn$44.91bn$46.43bn$45.80bn$49.84bn$57.35bn
Total noninterest expense$23.60bn$23.78bn$24.28bn$23.98bn$26.85bn$27.32bn
Compensation expense$14.09bn$13.71bn$13.57bn$13.12bn$15.34bn$15.16bn
Technology, communications and equipment expense$2.58bn$2.70bn$2.84bn$2.91bn$3.02bn$3.11bn
Professional and outside services$2.84bn$3.01bn$3.17bn$3.34bn$3.48bn$3.85bn
Marketing$1.30bn$1.28bn$1.48bn$1.47bn$1.60bn$1.67bn
Occupancy expense$1.30bn$1.26bn$1.42bn$1.48bn$1.45bn$1.48bn