From briefing to publication: a real workflow with Claude Code

A real-world workflow

A lot of people ask how Claude Code is actually used in practice, beyond the theory. In this article we share the exact process we follow at Divergente.Digital to create content with this tool.

The step-by-step process

1. We define the briefing

Everything starts with a markdown file inside the project. We include: topic, target keywords, audience, tone, and expected length. This file is the context that Claude Code will use as its foundation.

2. Claude Code reads the context

We give it access to the briefing along with the client's existing blog content. This is critical: Claude doesn't generate in a vacuum, it generates on the basis of what already exists. This way it maintains style coherence and avoids repetition.

3. It generates the first draft

The draft is created directly as an MDX file in the correct project folder. No copying and pasting from a chat window. The file already has the correct frontmatter, an SEO-optimized heading structure, and content ready for review.

4. We review and edit

The human team reviews the content. We request specific adjustments if needed: "make the tone more approachable," "add a practical example in section 3," "shorten the conclusion." Claude Code applies the changes directly to the file.

5. We optimize for SEO

With the content approved, we ask Claude Code to generate meta tags, image alt texts, and review the heading structure. Everything optimized for the keywords we defined in the briefing.

6. We publish

A simple commit and deploy. What used to take half a day now gets done in under an hour.

Tips for replicating this workflow

Start with the repetitive stuff

Don't try to revolutionize everything at once. Identify the tasks that consume the most time and are the most predictable. Weekly reports, social media posts, meta descriptions: those are ideal candidates.

Create context files

The magic of Claude Code lies in context. Build documents with:

• Style guides for each client
• Detailed buyer personas
• Keyword lists and topic clusters
• Examples of content that performed well

The more context you provide, the better the results.

Iterate instead of expecting perfection

The first output is rarely the final one. Use Claude Code as a collaborator: request changes, adjust the tone, refine the structure. The iterative process is where the best results come from.

Don't skip human review

Always review generated content to make sure that:

• Data and statistics are accurate
• The tone is right for the brand
• There are no claims you can't back up
• The content delivers real value to the audience

Conclusion: automate to create better

Automation with Claude Code isn't about churning out generic content at scale. It's about freeing up time to think better, to design more creative strategies, and to give each client the attention they deserve.

At Divergente.Digital we help agencies and businesses integrate AI into their marketing processes. If you want to know how we can help you, [get in touch](/#contacto).