Content

Blogs, images, LinkedIn posts and more with AI

Generating content with AI: from creation to publication, all in one app.

Responsibility
Architecture for AI and application
Technology
  • Next.js
  • React
  • TypeScript
  • Node.js
  • PostgreSQL
  • Claude
  • AWS Bedrock
  • Azure
  • WordPress

Context

A post for a company blog rarely comes together in one place. The text is written in one program, the cover image is built in a second, the graphic in a third. Approval runs by mail, and in the end someone enters everything into the website by hand. Every switch costs time, and nobody sees at a glance which post currently stands where.

One path, one app

Everything happens in one place, from the first idea to publication. The post only leaves the app once it is finished:

  1. Settle the topic and the goal, develop the outline in conversation with the AI, back facts up with sources
  2. Write the text and revise it in the editor
  3. Create the cover image and the graphics, each as a draft first, then checked
  4. Approval by the person in charge
  5. Transfer to the website: the images first, then the post with the finished images in the right place

The post arrives in WordPress finished. There is no step in which someone copies, uploads or reformats anything. All images are created in the app as well: cover image, graphics in the text, architecture diagrams, each generated with AI and checked in the same interface. No image program on the side, no folder full of exports.

Beyond the blog, the app carries further content: LinkedIn posts, single images, graphics and drafts for interfaces. Each has its own flow, all share the same interface.

Posts can be shared with colleagues and edited together. Whoever is invited sees the same state, comments, changes or approves.

How it is built

At the center stands a conversation with the AI that runs in phases: topic, outline, text, image idea. Every phase has its own aids and its own goal. The author stays in the conversation, the app leads through the phases and remembers every intermediate state.

Component Technology Decision
Interface Next.js, React One workspace for text, images and approval, no switching between programs
AI access Anthropic API, AWS Bedrock Two providers behind one door; if one fails, the other takes over
Images own flow with a check Cover image, graphics and diagrams are created with AI in the app; no image enters the post unchecked
Sign-in Entra ID Sign-in with the company account, no user accounts of our own
Storage PostgreSQL, Azure Blob Texts and intermediate states in the database, images in file storage
Publication WordPress interface Images first, then the post with the finished images in the right place

Learnings

  • AI helps in every step, not only with the text. Outline, sources, images, diagrams: everywhere it takes the first swing, the human decides and refines.
  • More content in less time. What used to take days across several programs now comes together in one app in a day. More posts are published, not fewer.
  • Better quality, because the check is built in. Every text and every image is checked before it is approved. Before, that depended on who happened to have time.