The revolution of working methods in the AI ​​era! Anthropic team internal information: How to reconstruct the old working methods with AI?

Recently, Anthropic released an internal research report that details how 10 different departments in the company use Claude Code in their daily work.

These real-life cases not only demonstrate the powerful potential of AI tools, but more importantly, they reveal a revolution in the way we work that is taking place.

Marketing Team: From 2 hours to 15 minutes

The most impressive story is from the growth marketing team.

As a non-technical team, they faced a seemingly impossible task: creating and optimizing advertising content at scale across multiple platforms.

The traditional creative ad creation process is as follows: marketers manually create ad copy, adjust titles and descriptions one by one, and then test the effects on different platforms.

This process is not only time-consuming, but also difficult to do A/B testing on a large scale.

But now, they have built an "ad-generating bot" using Claude Code.

The system analyzes tables containing hundreds of existing ads and their performance data and automatically identifies those that are underperforming.

Then two specialized "sub-helpers" are called:

One is responsible for creating the title (must be within 30 characters), and the other is responsible for the description copy (90 character limit).

Within minutes, the system can generate hundreds of new ad variations that meet your requirements.

"It used to take two hours to create advertising copy, but now it only takes 15 minutes," said a team member.

More importantly, this scale of creative production allows them to conduct more tests and discover more effective marketing strategies.

They also developed a Figma plugin that can batch generate up to 100 ad image variations.

What used to take hours of copying and pasting can now be done in half a second.

Designer’s new identity: I can also write code

How do people with non-technical backgrounds cross the technical threshold?

In the traditional product development process, there is a gap between designers and engineers.

Designers create visual drafts and then need to explain the design intent to engineers in detail. It takes multiple rounds of communication to achieve the final effect.

This process is not only inefficient, but also often leads to a "gap between ideal and reality". Now, the design team directly uses Claude Code to realize their design ideas.

A designer shared this experience: when he needed to adjust the font, color or spacing of the interface, he no longer needed to write a long design specification, but could directly describe the requirements in Claude Code, and even paste a screenshot of the interface so that AI could understand and implement the modification.

Designers are not only tweaking surface visual elements, they are also drilling down to the logical level of the application to make changes.

This change resulted in two very different user experiences:

For developers, this is an "enhanced workflow" that makes their work more efficient; for non-technical users, this is a new experience of "Oh my god, I'm a developer." A complex project that originally required a week of coordination can now be completed in two 30-minute conference calls .

Designers now have Figma and Claude Code open at the same time 80% of the time, which has become their main working tools.

Data team: Let machines understand business language

In the past, when finance staff needed specific data, they had to turn to the technology team for help.

Now, they can describe their needs in plain text: "query this dashboard, get information, run these queries, generate Excel output", and Claude Code will understand and execute the entire workflow.

This "natural language programming" method allows employees without any programming experience to complete complex data processing tasks independently.

This not only frees up the technical team's time, but also gives the business team unprecedented data autonomy.

On the technical level, the data team also uses Claude Code to solve complex infrastructure problems .

One time, their Kubernetes cluster failed and could not schedule new containers. Traditional diagnosis methods require experienced network experts, but they chose a new way: sending a screenshot of the error dashboard to Claude Code.

The AI ​​walked them through Google Cloud’s interface, menu by menu, until they finally spotted a warning about Pod IP address exhaustion .

Claude Code not only diagnosed the problem, but also provided the exact commands to create a new IP pool, bypassing the need for a network expert.

Legal team: AI can also be humane

The most unexpected case came from the legal team. This department, seemingly the least related to technology, discovered some of the most heartwarming application scenarios of Claude Code.

One legal team member shared a personal story about how he used Claude Code to create a communication assistance app in just one hour for a family member who had difficulty communicating verbally for medical reasons.

Using speech recognition technology, customized solutions are provided for people with special needs.

The legal team's working method is also very inspiring: they adopt a "two-step method" - first brainstorming and planning in Claude.ai, and then moving to Claude Code for specific implementation. They also use a lot of screenshots to show the desired interface effects, allowing AI to understand the requirements through visual feedback.

Change is happening

These cases show a common trend:

AI tools are blurring the lines between technical and non-technical jobs.

Marketers are building complex automation systems, designers are writing code, and legal professionals are developing applications.

This is not a simple tool upgrade, but a fundamental change in the way we work.

More importantly, this change has unleashed people's creativity.

A marketer’s creative idea can be instantly turned into a fully functional tool, and a designer’s idea can be directly converted into running code.

Claude’s team’s experience also revealed some best practices for using AI tools: treat AI as a collaborative partner rather than a one-time tool, adopt an iterative working approach, make good use of visual feedback, and establish a checkpoint mechanism for rollback at any time .

Another particularly important point is to overcome the fear of sharing "imperfect" works. The legal team particularly emphasized that you should not be afraid to show "toy" projects or unfinished work, because these prototypes can often inspire others to see possibilities that have not been considered before.

This internal report from Anthropic is not just a product usage guide, but more like a window that allows us to see the future of working in the AI ​​era.

Change has already begun, and it is happening faster and deeper than we imagined.

Perhaps it's time to rethink our definition of "work" itself.