← back to ranking

AXP · American Express Company

Financial - Credit Services · mkt cap $212.2B · calls: Q1 FY2026 vs Q4 FY2025
48.0 conviction · conf-adj 48

conf 3/10 partial

enthusiasm:27.0 · trend:8 · quantifies:0 · impact:0 · under_radar:0 · credibility:5 · business_impact:8 · disruption:0 · commitment:0 · confirmation:0

Enthusiasm latest 9 / prev 7 (rising)

AXP's AI thesis rose from internal data-platform enablement in Q4 FY2025 to a fuller Q1 FY2026 story spanning agentic commerce, fraud/risk, developer tools, purchase protection, partner distribution, proprietary AI experiences, servicing and developer productivity. Credibility is moderate-high because management ties the thesis to AXP's closed-loop data advantage and gives concrete productivity/process metrics, but direct revenue impact from agentic commerce remains unquantified and explicitly early-stage.

GROUNDED NEXT-FY IMPACT vs CONSENSUS

Grounded on actual base — revenue $80.5B · net income $10.8B · net margin 13.5% · diluted EPS 15.38

These are next-fiscal-year annual uplift estimates, not next-quarter numbers.

Aggregate next-FY est. rev uplift: % · next-FY EPS uplift: % · vs analysts: inline · priced in: high · confidence: 3/10

ClaimFigureArithmeticNext-FY Rev %Next-FY EPS %
AI gives ~30% programmer productivity in coding/testing
productivity · soft		~30%30% productivity benefit * undisclosed engineering/technology payroll base = unsizable. No tech-labor cost or FTE base disclosed in the financial base or any management claim (dev cost is buried in AXP opex: rev 80.464B less operating income 13.80B = 66.67B total expense is mostly interest, rewards, provisions, marketing — not isolable). Pure productivity, so topline assumed ~$0/80.464B = 0.0%; after-tax saving = 30%*(tech labor $)*0.79 / 10.833B NI but tech labor $ is unanchored.0
New data/analytics platform cuts marketing & fraud process time by 90%
cost · soft		90%Process-time %, not a dollar saving; 90% time reduction != 90% cost reduction. No marketing/fraud opex base disclosed in inputs, and it cannot be mapped to a disclosed affected revenue line, so after-tax saving = 90%*(undisclosed base)*0.79 / 10.833B NI is uncomputable. null.
Migrate 100% of data/analytics processes to new cloud platform by 2027
other · soft		100% by 2027Implementation milestone, not a disclosed $ revenue/savings claim. Enables the 90%-time-reduction claim but carries no independent sizable base and no next-FY attributable $ against 80.464B revenue or 10.833B net income. null.
Calls per account into service centers down 25% over last 3 years (digital self-service)
cost · soft		25% (trailing 3 yrs)Cost-saving in spirit but (a) no service-center/operating-cost base disclosed and (b) it is a TRAILING 3-year result already embedded in the 2025 net margin (13.46%) and net income base — not incremental to next FY. Pure cost, topline ~$0/80.464B = 0.0%; after-tax saving = 25%*(undisclosed servicing $)*0.79 = unsizable and already in run-rate.0

Assumptions: Tax rate 21% (after-tax saving = pre-tax*0.79); default incremental net margin = current net margin 13.46% had any revenue claim existed. Phasing: next-FY = FY2026. No revenue/bookings figure attached to any AI claim and no cost base (tech payroll, marketing/fraud opex, servicing cost) disclosed anywhere in inputs, so per the no-invention rule every quantified claim is %-without-obtainable-base = soft. Pure cost/productivity claims are modeled topline ~$0 (no quantified capacity/volume lift); the 25% call-reduction is explicitly trailing and already in the 2025 base. Forward statements ('accelerant', AI-agent commerce, intent-driven authorizations, partner discoverability) are explicitly unquantified by management.

Top line: Nil that can be sized. Every quantified AI claim is an internal efficiency/process metric (coding productivity, process-time, call deflection) or a migration milestone — none is a revenue or bookings figure; pure cost/productivity claims modeled at $0 topline, the two unmappable ones null. The genuine top-line optionality (AI-agent commerce, intent-driven authorizations, making membership assets discoverable on partner AI platforms) is real and well-positioned via the closed-loop network but management calls it 'really hard to quantify at this early stage', so aggregate rev_uplift = null.

Bottom line: Directionally positive but unsizable. 30% programmer productivity, 90% marketing/fraud process-time cuts, and 25% call deflection all push toward lower unit costs and operating leverage, but no dollar base is disclosed for any of them, so after-tax savings (saving_$*0.79 / 10.833B NI) cannot be computed (eps_uplift = null). Note the 25% call reduction is a trailing-3-year result already reflected in the 13.46% net margin — not incremental to FY2026.

[impact n/m (all claims soft/unanchored)] Consensus already embeds strong, steadily-compounding growth (EPS 15.42 -> 17.62 -> ~20.14, ~+14.3% y/y; net income 10.737B -> 12.286B = +14.4%; revenue ~79.4B -> ~86.4B). Because management disclosed no anchored AI revenue or savings dollars, the calculable adopter-side AI uplift is null and cannot be shown to exceed that trajectory; these efficiencies plausibly help sustain the margin/EPS curve rather than beat it. The agentic-commerce upside is the part NOT in the numbers, but it is unanchored, so it cannot lower priced_in on evidence.

MODEL CONSENSUS (impact)

partial

Both agree: all AI claims adopter-side, soft, unsizable (no dollar base disclosed); aggregates null. Reconciled topline-zero treatment and verdicts conservatively.

Conflicts reconciled
FieldOpus 4.8GPT-5.5
Rev uplift %––
EPS uplift %––
Priced inmediumhigh
vs analystsinlineunclear
Confidence34
Top lineEffectively nil that can be sized. Every quantified AI claim is an internal efficiency/process metric (coding productivity, process-time, call deflection) or a migration milestone — none is a revenue or bookings figure. The genuine top-line optionality (AI-agent commerce, 'intent-driven authorizations', making membership assets discoverable on partner AI platforms) is real and strategically well-positioned via the closed-loop network, but management states it is 'really hard to quantify at this early stage,' so rev_uplift = null, not zero-by-evidence.No hard AI revenue uplift is quantified. The coding/testing and service-center claims are operational, so modeled topline is $0; marketing/fraud and migration claims lack an affected revenue base, so revenue uplift is null rather than invented.
Bottom lineDirectionally positive but unsizable. 30% programmer productivity, 90% marketing/fraud process-time cuts, and 25% call deflection all push toward lower unit costs and better operating leverage, but no dollar base is disclosed for any of them, so after-tax savings cannot be computed (eps_uplift = null). Note the 25% call reduction is a trailing-3-year result already reflected in the 13.46% net margin — not incremental to FY2026.All quantified claims are adopter-side, but each lacks a disclosed cost base. EPS math cannot be completed: savings would be saving_$ * 79% / $10.833B net income, but saving_$ is undisclosed for programmer productivity, marketing/fraud processes, migration, and service-center call reduction.
ReasoningConsensus already embeds strong, steadily-compounding EPS growth (15.42 -> 17.62 -> 20.14 = +14.3% then +14.3% y/y) and rising revenue (79.4B -> 86.4B). These efficiency gains plausibly help sustain that margin/EPS trajectory, so they are largely consistent with — not above — what consensus assumes; no AI math points clearly above the curve (all adopter figures are null/soft). Management itself says AI is 'an accelerant' but unquantified, so the agentic-commerce upside is the part NOT in numbers — but it's unanchored, so it cannot move priced_in to 'low' on evidence. Net: medium.Consensus already models net income rising from $10.737B in 2025 to $12.286B in 2026, a $1.549B increase, or 14.4%; EPS rises from $15.417 to $17.617, up 14.3%. Because management disclosed no anchored AI revenue or savings dollars, the calculable adopter-side AI uplift is null and cannot be shown to exceed the consensus trajectory.

Rows highlighted where the two models disagreed.

QUANTIFICATIONS
programmer productivity from AI in coding and testing: about 30% benefit (Q1 FY2026/current, bottomline)
“As far as technology goes, we've been very fortunate with some of the results we're seeing from an AI perspective and that we're getting about 30% benefit with our programmers from a coding perspective and testing perspective.”
time reduction for marketing and fraud processes from new data and analytics platform: 90% (Q4 FY2025/current, both)
“The new platform, is built on the public cloud, is already reducing the time for key processes in marketing and fraud by 90% and we expect to migrate 100% of our data and analytics processes to the new platform by 2027.”
data and analytics process migration to new platform: 100% (by 2027, both)
“The new platform, is built on the public cloud, is already reducing the time for key processes in marketing and fraud by 90% and we expect to migrate 100% of our data and analytics processes to the new platform by 2027.”
service center calls per account reduction from digital self-service: 25% (over the last three years, bottomline)
“Over the last three years, for example, the number of calls per account coming into our service centers has dropped by 25%.”
PAST (realized)
CURRENT (now)
FORWARD (guidance)
TRACK RECORD — PROMISE vs DELIVERY

55/100 track record   too-early  6 calls reviewed

American Express's only quantified AI commitments — full data/analytics platform migration by 2027 and the 2026 agentic-commerce/commercial roadmap — were introduced in just the last two calls and their deadlines have not arrived, so there is no completed prior AI target to confirm or refute and the track record is genuinely too-early to judge.

Migrate 100% of data and analytics processes to the new third-generation (public-cloud, GenAI/agentic-enabling) platform by 2027 — promised Q4 FY2025
too-early Still in progress as of Q1 FY2026; the 2027 deadline has not arrived, so not yet judgeable.
New data/analytics platform 'already reducing the time for key processes in marketing and fraud by 90%' — promised Q4 FY2025
too-early Stated as an already-achieved result rather than a forward target; reaffirmed via continued AI/fraud investment in Q1 FY2026 but not independently re-quantified.
Roll out 8 new/enhanced commercial products plus AI-powered capabilities (ACE agentic-commerce kit, agent purchase protection) in the U.S. during 2026 — promised Q1 FY2026
too-early Announced as a 2026 roadmap mid-year; timeframe still open, no delivery data yet.
PRICED-IN (REFINED)
HIGH (already in)

Est. revisions rising  ·  Fwd P/E 20.2  ·  EV/Sales 2.7x

AI claim maps to Global Merchant and Network Services, Global Consumer Services Group, Global Commercial Services

Analyst ratings show upward migration from January to June, with more buys and fewer holds/strong sells, and price targets are also rising with last-month average above last-quarter and last-year averages. Forward estimates already bake in steady revenue and EPS growth through 2027, while a 20.2x forward P/E and 2.7x EV/Sales are rich for a mature credit-services company. AI benefits would most plausibly flow through network/fraud/data services and card servicing efficiency, so the combination of rising revisions plus a premium valuation points to AI upside already being reflected.
COVERAGE — ENTHUSIASM TRAJECTORY + CATALYSTS
2Q4 FY20242Q1 FY20252Q2 FY20253Q3 FY20258Q4 FY20259Q1 FY2026

AI enthusiasm across 6 calls — trend ↗ rising

AI moved from absent to specific data, GenAI, agentic commerce, fraud, personalization, automation, and customer-facing growth initiatives.

RECENT AI CATALYSTS & NEWS
BUSINESS IMPACT - QUALITATIVE MATERIALITY

7/10 qualitative impact   material  medium-term · mixed evidence

Where AI matters: agentic commerce, fraud/risk, servicing and developer productivity

AXP has credible AI use cases tied to its closed-loop network: agentic-commerce card integration, purchase protection, intent-driven authorization, fraud controls, servicing tools, and internal productivity. The upside is material because it can reinforce transaction share and risk economics, but revenue impact is still early and unsized, with hard evidence mostly limited to process/productivity metrics.

Caveats: Agentic-commerce monetization remains unquantified and adoption timing is uncertain; AI platforms could intermediate customer intent and weaken direct engagement; Fraud and authorization complexity may rise in autonomous-agent transactions; Productivity metrics are process-based, not disclosed as dollar savings or EPS uplift

AI DISRUPTION / CANNIBALIZATION RISK  two-sided · 3/10

AI does not automate away card issuing, credit, rewards, or merchant acceptance, but agentic shopping could shift customer interface and payment-choice influence toward AI platforms. AXP's brand, affluent card base, network data, and fraud/trust role make the core durable, though routing and engagement could face pressure.

OPTIONS / MARKET STRUCTURE

option liquidity: good

ATM IV
—
TYPICAL BID-ASK
—
OPEN INTEREST
—

proxy inputs — dollar-ADV $1.1B · beta 1.084 · px $310.97

source: proxy (no options chain on FMP)
FMP /stable/ exposes no options-chain endpoint on this key, so ATM IV, bid-ask spread and open interest are unavailable. Liquidity below is a PROXY from dollar-ADV, beta and price level (a stand-in for option depth), not measured option-market data.

CONFIRMATION — INSIDERS · 13F · LANGUAGE
Undercutting — insiders selling, institutions trimming, management language 6/10 measured.
INSIDERS selling 13 open-market sell(s) vs 0 buy(s) — net distribution
INSTITUTIONS (13F) trimming as of 2026-03-31: 184 new / 345 closed positions; 1251 increased / 1218 reduced; institutional ownership -0.24pp; -155 net 13F holders
MGMT LANGUAGE 6/10 measured Real AI launches and 2026 rollout language, but framed with early-stage, well-positioned, and long-term-growth qualifiers.
commit “Earlier this month, we introduced the Amex Agentic Commerce Experiences or ACE Developer Kit”
commit “We have more AI-powered products and capabilities under development that will roll out this year”
hedge “while it's still early days, we are embarking on a new era of commerce”
VERBATIM AI QUOTES
“As I said in my recent annual letter to shareholders, while it's still early days, we are embarking on a new era of commerce, where AI-powered agents can make autonomous decisions on behalf of consumers and businesses.”
— Stephen Squeri, Q1 FY2026
“Given our closed-loop network that provides an end-to-end view of transactions, and supported by the investments we've been making in our technology and risk capabilities, we are well positioned to deliver intent-driven authorizations, enhanced fraud protection and strong security features to help protect our card members and merchants.”
— Stephen Squeri, Q1 FY2026
“Earlier this month, we introduced the Amex Agentic Commerce Experiences or ACE Developer Kit, which will enable the integration of American Express cards into AI-powered transactions with trust and control.”
— Stephen Squeri, Q1 FY2026
“Along with the kit, we announced Amex Agent purchase protection, an industry-first commitment to back our card members by protecting registered agent purchases.”
— Stephen Squeri, Q1 FY2026
“We have more AI-powered products and capabilities under development that will roll out this year to help transform and grow our business.”
— Stephen Squeri, Q1 FY2026
“This includes upcoming announcements with leading AI companies to make our membership assets discoverable and actionable on their platforms and building proprietary AI-powered experiences across our own platforms.”
— Stephen Squeri, Q1 FY2026
“In summary, our business continues to perform at a high level, exhibiting continued momentum from executing our proven strategy and making meaningful progress on the strategic use of AI to drive long-term growth and efficiencies.”
— Stephen Squeri, Q1 FY2026
“As far as technology goes, we've been very fortunate with some of the results we're seeing from an AI perspective and that we're getting about 30% benefit with our programmers from a coding perspective and testing perspective.”
— Stephen Squeri, Q1 FY2026
“But what that has done is allowed us to get to more stuff.”
— Stephen Squeri, Q1 FY2026
“Not as many people want to talk on the phones and plus we're making the people that are answering the phone is more efficient because they have AI tools at their disposal, whether that's for travel or whether that's for card servicing.”
— Stephen Squeri, Q1 FY2026
“I mean, look, I mean, I think from my perspective here is that in an agenetic world, data is king.”
— Stephen Squeri, Q1 FY2026
“Data is a king from a service perspective, an identification perspective, a fraud perspective, a credit perspective, data is king.”
— Stephen Squeri, Q1 FY2026
“And so as we released our ACE developer kit, one of the things that we did were a developer kit is we said, look, to control the transaction to understand what's going on, what we want to do is have the agentic declare intent, and we want to match that intent with what was actually purchased.”
— Stephen Squeri, Q1 FY2026
“And so I think it's going to make our fraud and our risk capabilities and our ability to detect fraud and our ability to back our card members even better than we would in a brick-and-mortar world or in a traditional e-commerce world, which is why we came out with Amex agent protection.”
— Stephen Squeri, Q1 FY2026
“And I think anybody that talks to you about large language models will basically say to you, the model is as good as the data that it has.”
— Stephen Squeri, Q1 FY2026
“For example, we're rolling out our new third-generation data and analytics platform.”
— Stephen Squeri, Q4 FY2025
“Which will enable greater personalization in marketing, improve servicing experiences, augment our industry-leading fraud capabilities, and enable new GenAI and AgenTic use cases.”
— Stephen Squeri, Q4 FY2025
“The new platform, is built on the public cloud, is already reducing the time for key processes in marketing and fraud by 90% and we expect to migrate 100% of our data and analytics processes to the new platform by 2027.”
— Stephen Squeri, Q4 FY2025
“As a result, we're driving more revenue-generating engagement via the apps we're creating operating efficiencies from digital self-service.”
— Stephen Squeri, Q4 FY2025
“Over the last three years, for example, the number of calls per account coming into our service centers has dropped by 25%.”
— Stephen Squeri, Q4 FY2025
“Including the integration of Center's expense management solution including those already in market, such as our travel counselor assist tool, our dining companion experience, as well as the deployment of GenAI tools to nearly all our colleagues worldwide.”
— Stephen Squeri, Q4 FY2025
“I think what's really exciting for us is to be able to take large language models that are out there and take our data and insert that in and really come up with great card member offers, great card member insights, be able to create archetypes of various cardholders, be able then to treat cardholders in a and target cardholders in a much more effective way.”
— Stephen Squeri, Q4 FY2025
“But we're already seeing some benefits of that in some of our card member marketing, which again, leads to some of the reduction in overall cost.”
— Stephen Squeri, Q4 FY2025
“But this is a business that you know, not only you have to invest in value propositions, but you really have to invest in a light way in the technology behind it because, ultimately, it's technology that drives those value propositions, and it's a technology that drives the appropriate engagement with your cardholders.”
— Stephen Squeri, Q4 FY2025
ANALYST QUESTIONS ON AI
Q (Q1 FY2026, Ryan Nash): And you talked about increased marketing and tech spend. Can you maybe just talk about what that's offsetting in terms of where was performance tracking better than expected? And where are you using that to offset?
A: As far as technology goes, we've been very fortunate with some of the results we're seeing from an AI perspective and that we're getting about 30% benefit with our programmers from a coding perspective and testing perspective. But what that has done is allowed us to get to more stuff.
Q (Q1 FY2026, Jeffrey Adelson): But just given the market focus on AI jobs-related displacement. Just curious if you're seeing any sort of impact in the customer base today? Or just if you have any sort of views on what that trend might look like for you over the next few years?
A: Yes, we're not seeing any impact at all on this at all. And maybe I'll just make a couple of comments. I think technological change over the years, no matter what it has been, whether it's been the Internet, the cell phone, what have you and even eliminating the word processor and going to desktop PCs has always brought a plethora of new jobs, number one; and number two, has fuel GDP. Now will AI lose some jobs? Yes, it would.
Q (Q1 FY2026, Darrin Peller): Steve, you recently launched your Agentic Commerce Experience developer kit, I know you wrote about it at length in your letter. So just given our checks are indicating in general across AI and agentic, if there's been more fraud on some of these transactions. It's early days, but you're still seeing some of the questions on that. And then just structural questions around networks in an increasingly agentic world. Just I'd love to hear how you would think about through all your closed loop data [indiscernible] these transactions?
A: I think from my perspective here is that in an agenetic world, data is king. Data is a king from a service perspective, an identification perspective, a fraud perspective, a credit perspective, data is king. And when you look at our -- when you look at our business model, we have the card member, we have the network and we have the merchant. And we have a free flow of information and it's a perfect information as you're going to get in this model.
Q (Q1 FY2026, Cristopher Kennedy): Steve, in your letter, you kind of mentioned how new technology can accelerate growth at American Express. Is there a way to frame kind of the opportunity today with data in agentic relative to prior innovations, such as e-commerce or mobile payments?
A: Yes. I think it's a little bit too early. And I think the company is so big at this particular point, as I said just before, I think it was so early stages. I think if you would ask me that question when e-commerce first started, I would have probably given you the same answer.
Q (Q4 FY2025, Christophe Kennedy): You've given a lot of great engagement metrics. And you do have the new data analytics platform on the horizon. Can you just talk about that journey and the tools that you'll have to drive more card member engagement as you get into AI, etcetera?
A: I think what's really exciting for us is to be able to take large language models that are out there and take our data and insert that in and really come up with great card member offers, great card member insights, be able to create archetypes of various cardholders, be able then to treat cardholders in a and target cardholders in a much more effective way.