Not a description of how I work. A direct look at it.
Four independent working examples — an orientation audit excerpt, an information architecture blueprint, an orientation journey, and an AI agent briefing example. Each is based on Bettina's own independent projects, not a paid client engagement, and is shown here as an illustrative working method rather than a finished deliverable.
← Back to For OrganizationsOrientation Audit — Excerpt
Independent working example · Gesundheitsnavigator (Health Navigator) concept page
This is an excerpt from an independent audit conducted on one of Bettina's own projects, used here as a working example of her analysis method — not a paid client engagement or a complete audit.
Subject: the Gesundheitsnavigator concept page — a German-language page introducing an orientation tool for women navigating complex health decisions. Method: a structural and content review against the same criteria applied across Praxis Liebenswert's own portfolio.
The page clearly explains the idea behind the Gesundheitsnavigator: helping women orient themselves in health information and decisions without diagnosing or directing. The concept is well-articulated and the target audience is specific.
The page describes what the tool is for, but a first-time visitor cannot yet experience the orientation logic itself — there is no interactive, state-based flow live on this page. The gap sits between the promise ("this will help you find your next step") and the immediate experience (a description, not a demonstrated next step).
The text itself is well-structured. But it asks the visitor to hold an abstract concept in mind — "a navigator that adapts to my situation" — without a concrete example. For someone already dealing with a stressful health situation, an abstract promise adds a small but real interpretive burden.
No visible, low-commitment first action exists between "read about the concept" and "get in touch." There is no sample question, no single worked scenario, no visible entry point into the underlying logic.
- Add one concrete example: a single situation → one sample orientation output, visible without any interaction required.
- Convert one section of prose into a visible decision-tree fragment, so the mechanism is shown, not just described.
- Add a low-commitment next step above the general contact call to action.
Add one worked example. Convert one paragraph into visible decision logic.
Build the interactive, state-based demo behind the concept.
Add micro-copy clarifying the current status of the tool.
Full visual redesign of the concept page before a working demo exists.
Axes: Impact on visitor orientation (vertical) · Implementation effort (horizontal)
Two changes — one worked example and one visible decision fragment — would close most of the orientation gap without any new development.
See the same audit method applied to a real, public health page: IVF Patient-Page Communication Audit → (independent, unaffiliated — not commissioned by the institution referenced)
Information Architecture Blueprint — Excerpt
Independent working example · System Compass (Systemkompass)
A simplified view of the layered logic behind System Compass — an independent, privacy-first orientation tool for states of inner overwhelm. Shown here as architecture, not as finished screens.
Entry & Framing
A calm framing step: what this space is — and, just as important, what it is not (no diagnosis, no data storage, no therapy claim).
State Identification
Two inputs held separately rather than merged into one category:
- Current nervous-system state (e.g. overwhelmed, shut down, wired, calm)
- Life phase / situational context (e.g. fertility, pregnancy, postpartum, general)
Reflective Content
State and context together select from a library of grounding content and journaling prompts — written for that specific combination, not just for the general topic.
Local Processing
All input is processed in the browser only. Nothing is transmitted or stored — an architectural decision, not a privacy-policy footnote, and one that shapes what the tool can and cannot do (no history, no account, no cross-session personalization).
Branching Next Step
Three honest exit paths, not one generic "contact us":
Stay and continue reflecting
Routed to Quiet Rooms for audio-based regulation
Plain-language pointer to professional help — always visible, never hidden behind a menu
Decision points are kept separate instead of collapsed. A tool that merges "what state are you in" and "what is your situation" into a single question loses information a person could have given clearly.
Orientation Journey — Before / After
Independent working example · Illustrative framework, not a specific patient case
An illustrative comparison — not a medical journey, but an orientation journey: what changes when a structured orientation system exists between an event and a decision.
- Event — a diagnosis, referral, or difficult result
- Searches broadly for information
- Finds volume, not orientation — many pages, no clear priority
- Cognitive load rises — harder to judge what matters
- Uncertainty about the next concrete step
- Decision delayed, or made from fear rather than clarity
- Event — a diagnosis, referral, or difficult result
- Enters a structured orientation point (guide, navigator, or content architecture)
- Situation is identified, not just the topic
- Relevant information surfaces — the rest is deliberately withheld, not hidden
- One clear next step is presented, with its own boundaries stated
- Decision made with orientation, at the person's own pace
The difference is not more content. It's a system that decides, deliberately, what to show — and what to leave out.
AI Agent Briefing — From Vague Question to Verified Result
Independent working example · A real project brief from this website's own development
This is a real, only lightly abridged example from this site's own build process — not a hypothetical case. It shows what has to be resolved before an AI coding agent can be trusted with a change, and how the result was checked afterward.
A business question, not a technical one: "My site doesn't show what I can actually build. Should I publish a fixed price for it?" No file names, no scope, no mention of what should change on the page.
- Which exact capability to price, and at what boundary — a small, self-contained tool, not a full platform with backend, accounts or a database
- A defensible price grounded in comparable market cost, not in hours spent — hours-based pricing would have undervalued the work
- Which existing page and which exact section the new content belongs in, without disturbing what was already live
- Which markup patterns and CSS classes already existed and had to be reused rather than reinvented
- A second, older FAQ answer that had quietly become inaccurate and needed correcting in two places at once — the visible text and the structured data behind it — without anyone having flagged it
- What explicitly should not be built, so the result wouldn't expand into a new page, a new product line, or changed pricing elsewhere on the site
### Neuer Service-Block (nach Service 05, vor bestehendem Abschluss der Sektion) <span class="fo-service-num">06</span> <h3 class="fo-service-title">Lightweight Orientation Tool (Fixed Scope)</h3> ... <dt>Typical size</dt> <dd>Fixed scope, starting at €1,800. Does not include backend development, user accounts/login, payment processing or database-driven systems — these fall outside this fixed-scope format and would be quoted separately.</dd> ### Nicht tun - Keine neue Seite oder Unterseite anlegen - Keine weiteren Services über den einen hinzufügen - Keine Änderung an Preis oder Scope der bestehenden 5 Services
- Rendered page checked block by block against the brief — the service card, the FAQ text, the form dropdown — via direct screenshots, not by trusting the agent's own summary
- Visible FAQ text and the structured data (JSON-LD) version compared for exact wording match, since the two can silently drift apart
- The new dropdown option checked live inside the actual form, not assumed from a code diff
- Structured data re-validated with Google's Rich Results Test before publishing — a single schema error can silently change how AI search systems read the whole page
The skill was never "using Claude Code." It was knowing exactly what still had to be decided before an agent could be trusted with the decision — and refusing to call it finished until every piece had been checked, not assumed.
See how this applies to your organization
These are working examples from independent projects. A project inquiry usually starts with a short look at your own content or patient journey — not a generic proposal.