How to Be an Innovator in the Age of Agentic AI: The American Express Approach

American Express shares its experience implementing agentic AI. Four key approaches: start with real business problems, experiment quickly, embed into processes, monitor risks. The company is already applying them to increase efficiency.

AI-processed from ZDNet AI; edited by Hamidun News
How to Be an Innovator in the Age of Agentic AI: The American Express Approach
Source: ZDNet AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceZDNet AI ↗
Editor-in-Chief
Key metric #1Agent operates in real-time
Key metric #2Selects tools and data sources on its own
Key metric #3Learns from historical data

Autonomous AI agents are becoming reality in the corporate world. American Express shows how global companies can stay ahead in the age of agentic AI: strategy is needed, not spontaneous experiments.

Why Amex Is Looking at Agentic AI

Traditional AI models require constant human intervention. Agentic AI is different: the agent makes decisions on its own, performs actions in a chain (reads data → thinks → acts), and does this without constant prompts. For a payment company, this means automating complex processes: transaction verification, risk management, customer service.

  • Agent operates in real-time
  • Selects tools and data sources on its own
  • Learns from historical data
  • Requires fewer human interventions

Four Ways from American Express

According to Amex's top executive, successful innovators in the company do the following:

First—start with pain, not technology. Don't look for a complex use case just to check a box. Choose a real problem that costs the company dearly. For Amex, this was automating fraud protection and payment verification.

Second—experiment in small iterations. Don't build for years. Sprints of 2-4 weeks, fast feedback, quick deployment on real data (in a controlled environment). Amex tests agents on a subset of transactions before scaling.

Third—embed into existing processes. Agentic AI shouldn't replace everything at once. Good agentic AI complements existing workflows: it helps people make decisions faster, rather than excluding them entirely.

Fourth—monitor risks from day one. An autonomous agent can make a mistake that humans won't notice. You need monitoring, logging of all decisions, quick rollback, and an escalation chain.

Results at Amex

The company did not disclose specific figures, but mentions a noticeable reduction in transaction processing time and fraud protection errors. The key is culture: the company began to see agentic AI not as a panacea, but as a tool for specific tasks.

What This Means

For large fintech companies and banks, agentic AI is not futurism, but a tool for competitive advantage. Amex has shown that the winner is not the one who launches the most complex agent first, but the one who quickly connects AI to real revenue and risks. Others must move in that direction too.

Story

Key Questions Answered

Why Amex Is Looking at Agentic AI?

Traditional AI models require constant human intervention. Agentic AI is different: the agent makes decisions on its own, performs actions in a chain (reads data → thinks → acts), and does this without constant prompts. For a payment company, this means automating complex processes: transaction verification, risk management, customer service. Agent operates in real-time Selects tools and data sources on its own Learns from historical data Requires fewer human interventions

What are the key details on Four Ways from American Express?

According to Amex's top executive, successful innovators in the company do the following: First—start with pain, not technology. Don't look for a complex use case just to check a box. Choose a real problem that costs the company dearly. For Amex, this was automating fraud protection and payment verification. Second—experiment in small iterations. Don't build for years. Sprints of 2-4 weeks, fast feedback, quick deployment on real data (in a controlled environment). Amex tests agents on a subset of transactions before scaling. Third—embed into existing processes. Agentic AI

What are the key details on Results at Amex?

The company did not disclose specific figures, but mentions a noticeable reduction in transaction processing time and fraud protection errors. The key is culture: the company began to see agentic AI not as a panacea, but as a tool for specific tasks.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Autonomous AI Agents

Delegate Deep Research, Coding & Workflows to an AI Agent

Deploy autonomous multi-step execution with NomadNet Agent (agent.nomadnet.ai) — automated web research, data synthesis, code generation, and complex task orchestration.

Launch NomadNet Agent →NomadNet AI Platform
What do you think?
Loading comments…