Source: Analytics Insight (2026). Anthropic’s Claude Becomes Smarter at Handling Workplace Tasks .
https://www.analyticsinsight.net/news/anthropics-claude-becomes-smarter-at-handling-workplace-tasks
When I first came across an article about Anthropic, I felt genuinely excited.
The idea of AI agents itself is not new. I have long believed that most of daily tasks in organizations and in private life consists of small, repetitive tasks built on patterns. Once the context of each step can be processed, there is little to no reason for humans to stay involved in routine execution.
What felt different this time was that AI agents were no longer just a concept or an internal tool for advanced teams, but something that is openly available for everyone.
In this article, I want to explore the use of AI agents in the context of marketing automation.
Not whether it is possible—because it clearly is—but what the real hurdles are when trying to implement it.
I draw this perspective from my experience as a CRM marketing professional at a large Japanese company with over 10 million users, where I utilized marketing automation platforms such as Salesforce Marketing Cloud and Adobe Journey Optimizer.
TL;DR
Anthropic is not just building a powerful LLM.
From a CRM and marketing automation perspective, its real value lies in enabling safe, operationally realistic AI agents.
We have had some human error problems in the past even when using marketing automation tools, so automating more is a good solution to achieve safer and more reliable system.
Source: The AI Cheatcode (2026). AI Cheatcode — Master Claude AI.
https://aicheatcode.substack.com/p/ai-cheatcode-master-claude-ai
In Marketing automation platforms, the challenge is not whether AI can help, but how to introduce it without increasing operational risk and receiving consent from cautious companies.
Why AI Agents Are Hard to Deploy in Marketing Automation
In theory, marketing automation looks like a perfect use case for AI agents.
Most tasks follow patterns:
- segment users
- trigger journeys
- monitor performance
- adjust parameters
- document changes
However, in real production environments, there are three major hurdles.
1. Operations Are Fragile
In SFMC, a small mistake can have a large impact:
- wrong segmentation
- broken automation
- unexpected volume spikes
- compliance or brand risks
Because of this, fully autonomous AI is rarely acceptable. What teams actually need is decision support, not blind execution.
2. Systems Become Too Complex to Understand
As operations scale:
- automations multiply,
- Data Extensions grow uncontrollably,
- dependencies become unclear,
- and knowledge becomes tribal.
New team members often struggle not because they lack skill, but because they cannot see how things are connected. Even senior members spend significant time manually tracing flows and reviewing setups.
3. Documentation Is Always Behind
Documentation is critical in marketing automation—but also the first thing to fall behind. When it does, systems turn into black boxes, reviews become slow, and human error increases.
Where AI Agents Actually Add Value
The most realistic value of AI agents in marketing automation is not “optimizing everything automatically.”
It is reducing cognitive load.
Practically, this means:
- summarizing complex automations and data flows
- explaining dependencies between Data Extensions and activities
- highlighting suspicious changes or anomalies
- generating draft documentation from existing setups and logs
- proposing improvement ideas that humans can review and approve
In other words, AI agents work best as operators’ assistants, not replacements.
Source: Anthropic (2024). Building effective agents.
https://www.anthropic.com/engineering/building-effective-agents
Why Anthropic’s Approach Fits This Reality
Anthropic’s focus on safety and control aligns well with how marketing automation is actually run.
- It assumes AI will operate under constraints
- It fits naturally into human approval workflows
- It is well-suited for reading and structuring long, messy operational context
This makes it easier to introduce AI incrementally—without forcing teams to choose between efficiency and safety.
My Take
AI agents will absolutely become part of marketing automation. In the future, it can definetely replace marketing automation tools however because most of the system is precisice and different, further development is necessary.
I am excited and hoping to be a part of making it come true.