American Express Company (AXP)
Card issuer and network, one pattern. Read October 1, 2026 from the annual report for the period ending December 31, 2025, filed February 6, 2026. Ranked 47 of the 173 Financial Services companies read so far, by the share within reach now.
Within reach now
56%
projected from METR's April 2026 measurement and Epoch's November 2026 index
End of 2028
68%
Employees, annual report
76,800
Leading staffing pattern
Nondepository Credit Intermediation
Share of the working time within reach, long-run pace, four times in five
Now is projected from METR's April 2026 measurement and Epoch's November 2026 index.
- Now56%
- End of 202766%
- End of 202868%
- End of 203068%
What changes for its people by the end of 2028, by its industries' figures
- Stays with people 32%
- Becomes checking a model's work 60%
- Handed over with lighter checks 7%
The filing counts 25,900 U.S. and 50,900 non-U.S. colleagues and nothing by segment, so one credit pattern carries the workforce (c).
The figures on this page come from the company's own annual report and from BLS staffing patterns for its industries, never from a roster. They are estimates with a stated method. Nothing here says what the company will do: within reach is not a forecast of jobs lost or of savings.
What follows from the reading
Share of working time within reach of AI, on the long-run pace at four in five: 56% now, 66% at the end of 2027, 68% at the end of 2028, 68% at the end of 2030. By the patterns, the work is about 68% desk work, 28% people work and 4% body work. These are the site's industry estimates weighted by the shares below: estimates with a stated method and range, not a forecast of what the company will do. By the same patterns, at the end of 2028 about 32% of the working time stays with people (conversation, in-person and hands-on work), 60% becomes checking a model's work and 7% could be handed over with lighter checks, from the industries' own figures, not the company's roles.
Headcount
76,800 employees as of December 31, 2025 (stated). The filing counts approximately 25,900 colleagues in the United States and 50,900 outside it.
The filing: “As of December 31, 2025, we employed approximately 76,800 people, whom we refer to as colleagues, with approximately 25,900 colleagues in the United States and approximately 50,900 colleagues outside the United States.”
The workforce as industry patterns
| Pattern | Share | Reason, from the filing |
|---|---|---|
| Nondepository Credit Intermediation | 100% | By description (c): card issuing, lending and the payments network across all four segments, with no count by segment or kind of work. |
What the filing says about the workforce: Approximately 76,800 colleagues, 25,900 in the United States and 50,900 outside; 91 percent of survey participants would recommend the company as a place to work.
Companies read with the same leading pattern
Other companies whose largest staffing pattern is also Nondepository Credit Intermediation, largest first, each read from its own annual report. Within reach now and at the end of 2028, long-run pace, four in five. Now is projected from METR's April 2026 measurement and Epoch's November 2026 index.
- Rocket Companies, Inc. (RKT, 23,500 employees)52% now, 62% by 2028
- FirstCash Holdings, Inc (FCFS, 22,000 employees)55% now, 68% by 2028
- Synchrony Financial (SYF, 20,000 employees)56% now, 68% by 2028
- OneMain Holdings, Inc. (OMF, 9,300 employees)56% now, 68% by 2028
- UWM Holdings Corporation (UWMC, 9,100 employees)55% now, 69% by 2028
- EZCORP, Inc. (EZPW, 8,500 employees)56% now, 68% by 2028
- Rithm Capital Corp. (RITM, 7,240 employees)54% now, 66% by 2028
- Bread Financial Holdings, Inc. (BFH, 6,000 employees)56% now, 68% by 2028
What the filing says about AI
36 sentences in the filing name AI. The filing names AI in 36 sentences: generative AI and agentic commerce explored for its payments platform, AI agents executing transactions on behalf of users, and AI among the areas of growing regulation.
- “We also continue to explore ways to deploy new and developing technologies to enhance our payments platform and customer experience, such as uses for generative artificial intelligence (AI) and the integration of our products and services in agentic commerce.”
- “The development of agentic commerce solutions, in which autonomous or semi-autonomous AI agents initiate and execute transactions on behalf of users, has accelerated as generative AI technologies have advanced and become more popular.”
- “Privacy, Data Protection, Data Management, AI, Resiliency, Information Security and Cybersecurity Regulatory and legislative activity in the areas of privacy, data protection, data management, AI, resiliency, information security and cybersecurity continues to increase worldwide.”
- “Our regulators are increasingly focused on ensuring that our privacy, data protection, data management, AI, resiliency, information security and cybersecurity-related policies and procedures are adequate to inform customers of our data collection, use, sharing, retention and/or security practices, to provide them with choices, if required, about how we use and share their information, and to ap...”
- “In addition, regulators and legislators have heightened their focus on the use of AI and machine learning (ML) through the application of existing laws and regulations as well as by adopting new laws and regulations, such as the EU AI Act and AI legislation in several U.S. states (e.g., in California, Colorado and Utah).”
Its annual reports, year by year
Sentences naming AI in each annual report, counted by one rule, beside the median report in the Index that year. Each count links its filing. Every company, by year and sector
- Fiscal 20216median 0
- Fiscal 20226median 0
- Fiscal 202316median 1
- Fiscal 202426median 4
- Fiscal 202536median 10
Headcount by filing year
As reported in each year's annual report, with the filing. A year marked not read had no figure the reading could take from the report's text.
| Period | Employees | Filed | Filing |
|---|---|---|---|
| 2025-12-31 | 76,800 | 2026-02-06 | 10-K |
| 2024-12-31 | 75,100 | 2025-02-07 | 10-K |
| 2023-12-31 | 74,600 | 2024-02-09 | 10-K |
| 2022-12-31 | 77,300 | 2023-02-10 | 10-K |
| 2021-12-31 | 64,000 | 2022-02-11 | 10-K |
| 2020-12-31 | 63,700 | 2021-02-12 | 10-K |
| 2019-12-31 | 64,500 | 2020-02-13 | 10-K |
| 2018-12-31 | 59,000 | 2019-02-13 | 10-K |
| 2017-12-31 | 55,000 | 2018-02-16 | 10-K |
| 2016-12-31 | 56,400 | 2017-02-17 | 10-K |
| 2015-12-31 | 54,800 | 2016-02-19 | 10-K |
| 2014-12-31 | 54,000 | 2015-02-24 | 10-K |
| 2013-12-31 | 62,800 | 2014-02-25 | 10-K |
| 2012-12-31 | 63,500 | 2013-02-22 | 10-K |
| 2011-12-31 | 62,500 | 2012-02-24 | 10-K |
| 2010-12-31 | 61,000 | 2011-02-28 | 10-K |
| 2009-12-31 | 58,300 | 2010-02-26 | 10-K |
| 2008-12-31 | 66,000 | 2009-02-27 | 10-K |
| 2007-12-31 | 67,700 | 2008-02-28 | 10-K |
| 2006-12-31 | 65,400 | 2007-03-01 | 10-K |
| 2005-12-31 | 65,800 | 2006-03-06 | 10-K |
| 2004-12-31 | 77,500 | 2005-03-10 | 10-K |
| 2003-12-31 | 78,200 | 2004-03-12 | 10-K |
| 2002-12-31 | 75,500 | 2003-03-27 | 10-K |
| 2001-12-31 | 84,400 | 2002-04-01 | 10-K405 |
| 2000-12-31 | 89,000 | 2001-03-30 | 10-K |
| 1999-12-31 | 88,400 | 2000-03-29 | 10-K |
| 1998-12-31 | 85,000 | 1999-03-30 | 10-K |
| 1997-12-31 | 73,620 | 1998-03-31 | 10-K405 |
| 1996-12-31 | 72,300 | 1997-03-28 | 10-K405 |
| 1995-12-31 | 70,347 | 1996-03-29 | 10-K |
| 1994-12-31 | 72,412 | 1995-03-31 | 10-K |
| 1993-12-31 | 64,493 | 1994-03-31 | 10-K |
Limits of this reading
- The filing counts no people by segment or kind of work; the single pattern rests on its description (c).
- Two-thirds of colleagues are abroad; the pattern applies abroad because the work is the same in kind (rule 7).
Source: the company's annual report on EDGAR, accession 0000004962-26-000080, the filing. Read by Claude Fable 5.1 under rubric version 1, by the method. A company can write to info@stratussc.com to have its reading checked against its filing.
The report on American Express Company, $29
This page as a document to keep, with what it does not show: the share within reach at every date on three paces and two reliabilities, a typical workforce of this shape at the stated headcount role by role and task by task (the industries' staffing, never the company's roster), where to start, what changes for the people and what they would learn, every sentence in the filing that names AI, the headcount by year, the companies read with the same leading pattern, the method and a citation line. To print or save as a PDF, open again in your browser for a year.
See the report, $29See the sample first: Walmart's report is free
Questions about this company
- How much of American Express Company's working time is within reach of AI?
- An estimated 56% now and 68% by the end of 2028, on METR's long-run pace at four in five, from the workforce American Express Company's own annual report describes and BLS staffing patterns for its industries, never from a roster. It is an estimate with a stated method. It is not a forecast of jobs lost or of savings, and it says nothing about what the company will do.
- How many people does American Express Company employ?
- 76,800 as of December 31, 2025, as its annual report for the period ending December 31, 2025 states. About 34% in the U.S.
- What kinds of work make up American Express Company's workforce?
- Read from its annual report as a blend of 1 industry staffing pattern, the largest Nondepository Credit Intermediation; by those patterns the work is about 68% desk work, 28% people work and 4% body work. Each share has a one-sentence reason on this page that can be checked against the filing.
- What does American Express Company's annual report say about AI?
- 36 sentences in the filing name AI. Its annual reports for fiscal 2021 to 2025 name it in 6, 6, 16, 26 and 36 sentences, year by year. They are counted and quoted on this page, never judged: a count says how much a filing talks about AI, not what the company does with it.
- How is this estimated?
- Claude Fable 5.1 read the annual report under the Index's written rubric (version 1) on October 1, 2026: the stated headcount, the share in the U.S. when stated, and the workforce as a blend of industry staffing patterns. The estimates are the site's industry estimates weighted by those shares, from O*NET tasks with estimated lengths and METR's measurements of how long a task AI models can finish. A company can write to info@stratussc.com to have its reading checked against its filing.
Embed these figures on your site
A small card with the share of working time within reach now and at the end of 2027, 2028 and 2030, the date and a link back. Free under CC BY 4.0; the card carries the credit.