Building Orena’s agentic content engine

Building Orena’s blog meant figuring out how to keep a useful content operation running alongside everything else involved in building the company.

Orena helps people understand concerns about memory and thinking through at-home cognitive assessment. Content has a natural role in that experience. Someone researching cognitive testing may want to understand what an assessment involves, how to interpret the terminology, or what questions to bring to a clinician. I wanted Orena to be useful at that stage, before someone was ready to buy anything.

I also wanted to build an organic acquisition channel that we could develop over time. That required a repeatable process for choosing topics, creating useful articles, getting them onto the site, and deciding what to improve next.

I built an agent-assisted content workflow around that process. The publishing queue now records 148 published articles, and the refresh workflow records 45 completed tasks. Together, they give Orena a substantial content library and a structured way to keep working on it.

The workflow starts with a topic map and a publication queue. Broad guides connect to more specific questions, so individual articles have a defined role within the site. That gives each piece a purpose and helps keep the content from becoming a collection of disconnected posts.

The agent works from that shared context: the topic, related content, editorial instructions, and publication state. It prepares articles and supporting assets, works through build and validation steps, and packages changes for review. The work lives alongside the website in version control, where changes can be inspected and traced.

That is what makes the agentic part useful to me. The AI has a defined job, access to the context needed to do it, and a sequence of actions to work through. The queue records where the work stands. I can return to the process and see what has happened without reconstructing it from a chat history.

The publishing layer handles the repeatable details around an article: page generation, metadata, internal navigation, and the files that help search engines discover the content. Important article content is available in the page’s HTML. I wanted the production process to account for those details consistently, rather than leaving them as a separate checklist every time.

One of the most useful additions has been the refresh queue.

Search Console data gives us observations about how existing pages appear in search. Those observations can become specific tasks: revisit a title, clarify an answer, improve the relationship between pages, or investigate a piece of content that needs more data.

For example, an article explaining cognitive test scores had search impressions but relatively few clicks. Its refresh task focused on making the title more closely reflect the question people were asking. The record preserved the reason for the change and the wording before and after it.

That makes the work concrete. There is a page, an observation, a proposed change, and a record of what was done. We can then look for evidence of whether the change helped, keeping the action separate from its eventual outcome.

It also changes how I think about the content library. Existing articles remain part of the growth work. They can be revisited as we learn more about search behavior, update the product, or find a clearer way to explain something.

Orena’s subject matter makes editorial judgment especially important. A technically valid article can still overstate a claim, misinterpret a source, or leave a worried reader with the wrong impression.

I built the editorial instructions around plain language, credible sourcing, and clear boundaries between educational content and diagnosis. Automated checks help enforce structure and catch technical problems. Source relevance, factual accuracy, and the meaning a reader takes away still require scrutiny. A review label or a passing build cannot establish those things on its own.

The same thinking shapes the SEO and answer-engine optimization work. I’ve focused on clear questions, direct explanations, useful supporting material, and accessible pages. The objective is to make Orena’s content helpful to readers and suitable for discovery through search and AI-assisted answers.

The immediate value is an organized content operation with work that can be carried forward. Topic decisions have a home. Editorial expectations are reusable. Publishing tasks are connected to the site. Refresh work has a rationale and a recorded status.

As a founder, that gives me a clearer view of what needs attention and a more consistent starting point for the next piece of work. The decisions made for one article can inform the next without requiring the whole process to be rebuilt.

I used AI-assisted development to build and refine the workflow, while owning the growth objective, product decisions, and editorial standards. Much of the product work was deciding how the pieces should connect and where a person still needed to take responsibility.

The long-term goal is to turn that operating capability into sustained organic acquisition. The work now is to keep improving the content, measure how people discover and use it, and connect that evidence to the next set of decisions. Orena has a substantial foundation to work from, and a process for continuing to improve it.