Project documentation

Reflection & AI report.

Our reflection on the project, how AI tools were used, what we learned, and how we worked as a team.

Report length (sections 1 to 4): 0words Target 500 to 600

01Project reflection

We chose a real business, Alaradi Dental and Cosmetic Center in Salmabad, because the brief rewards real-world relevance and because the clinic’s existing website left out basics such as its phone number and opening hours. The clinic’s owner gave us permission to use its name, logo and photographs. Our goal was a calm, clear site that makes it easy to find a treatment, a doctor, the opening hours and the way to the clinic.

The project moved through clear stages: research, an agreed seven-page structure, a first Home page, a redesign, the remaining pages, testing, and deployment to Cloudflare Pages. The hardest moment was rejecting our first Home page. It worked, but it looked like a generic template, so we kept every feature and rebuilt the visual language around real clinic photography, the navy and gold of the logo, and uneven editorial layouts. Later we added a full Arabic version with a right-to-left layout and a dark theme.

02AI utilisation report

AI was used in the development process only; the website itself has no AI features. We first used Figma, with the HTML to Figma community file, to explore how the pages could look. We then discussed our ideas with ChatGPT, which helped us turn them into clear instructions. Claude Code, an AI coding assistant made by Anthropic, built the HTML, CSS and JavaScript from those instructions and helped with research, the Arabic translation, testing and debugging. Finally, the open-source rembg library removed the backgrounds from the four doctor photos on our own computer.

We did not accept AI output without checking it. We approved the plan before any code was written, resolved the conflicting opening hours ourselves, rejected the first design, and reviewed each stage before the next began. We also set firm rules: no invented clinic facts, prices, qualifications or testimonials, and no changes to the doctors’ faces. The prompts, outputs and changes are recorded in the AI documentation table below.

03Lessons learned

Planning first saved time. Collecting the clinic’s real information before coding showed us that sources can disagree: three sources gave different opening hours, and we had to find out which one was correct. Working with real content also taught us to ask for permission, to credit images, and to leave out anything we could not verify.

On the technical side, we learned to build responsively with relative units. Font sizes and spacing use rem with clamp(), and breakpoints use em, so the layout adapts from 320px phones to very wide screens without horizontal scrolling. Arabic support taught us to use logical CSS properties instead of left and right, and that details such as phone numbers need special care in right-to-left text. We also learned that accessibility is part of the design: labelled form fields, clear error messages, keyboard support, colour contrast and reduced motion. Finally, automated tests at eleven screen widths caught problems we would have missed by eye, and reminded us that AI-written code still has to be tested.

04Team collaboration experience

Team members
Yousif Alaali, 202501330
Mohammed Ali, 202404221
Adam Ali, 202405967
Yousif Alhamad, 202500649

We developed the project as a team of four. We communicated throughout the design and development process, and ideas and feedback were discussed before anything was implemented. The website then went through repeated rounds of testing and refinement, with problems raised and fixed along the way. We reviewed the documentation pages and the final submission together before handing in.

05AI documentation table

The same table appears on the Project overview page, as the brief requires. Prompts are quoted from our saved chat history; long prompts are shortened to their key sentence.

AI tools used during the project
AI toolPurposePrompt usedOutput generatedModifications made
Figma (HTML to Figma community file)Design exploration and prototypingNo text prompt: this was visual design work in the Figma file listed in the references.Visual layouts of the pages for the team to review.The team discussed the layouts and carried the agreed ideas into the written instructions. The design was refined further during the build.
ChatGPT (OpenAI)Refining prompts and instructionsExact wording not saved. In summary: we described our design requirements and ideas and asked for a clear, structured instruction for Claude Code.Structured prompts, such as the instructions quoted in the rows below.We read each prompt, checked it against the brief and our own decisions, and adjusted it before using it.
ChatGPT (OpenAI)Checking progress against the briefNow does every point in the project is successfully agreed with the project requirements what does claude mean by the final QA resultA point-by-point comparison of the project with the requirements, and a plain-language explanation of the final testing step.Used as a second opinion only. The checks themselves were run on the website, and we confirmed the remaining details, such as the submission date, ourselves.
Claude Code (Anthropic)Planning and requirement analysisDo NOT start building the website yet. First, carefully read the instructor's Markdown document and understand exactly what the assignment requires.A planning report: a checklist of the brief, a proposed seven-page structure, facts collected from the clinic’s website, Google Maps and Instagram, and eight questions for us.We approved the structure, chose the Google Maps opening hours after the sources disagreed, removed the consultation price, and decided not to use Bootstrap.
Claude Code (Anthropic)Home page, shared styles and the registration formPlanning is approved. You can start coding now.A first Home page with the registration form, its validation rules, and the shared CSS and JavaScript.We rejected the first design as too basic and asked for a redesign. The second version kept the features and changed the layout, photography and shapes.
Claude Code (Anthropic)Remaining pages and the booking demonstrationHome page is approved. Continue implementing the rest of the project now.The Services, Doctors and Visit us pages, the three documentation pages, and the demo patient portal.We supplied the second member’s ID, replaced the map placeholder with the clinic’s Google Maps listing, and kept the booking demonstration simple.
Claude Code (Anthropic)Testing and debuggingDo the final polish and bug-fixing pass across the entire project before deployment.An automated test script covering seven pages at eleven screen widths, and a list of bugs with fixes.Fixes kept after review: the mobile menu was trapped inside the header, hidden elements were still showing, photo reveals did not start, and doctor cards were hidden on phones.
Claude Code (Anthropic)Arabic version, right-to-left layout and dark modeI now want you to make the following final enhancements, fully test everything, and then update the existing Cloudflare deployment.An Arabic dictionary of about 800 phrases, right-to-left styles, a dark theme, and two header buttons.Corrected after testing: phone numbers appeared reversed in Arabic, colour codes displayed wrongly, one label was unreadable in dark mode, and the booking demo crashed at the time step.
rembg with the BiRefNet portrait model (open source, runs on our computer)Image preparation: doctor portraitsNo text prompt: the model receives a photo and returns the person without the background. Our instruction for this task was: Keep each doctor's real identity, face and appearance completely intact.Cut-outs of the four doctors from their original photos.The first model removed one doctor’s white coat, so a portrait model was used for all four. The cut-outs were placed on a blurred photo of the clinic’s own treatment room. Faces and bodies were not changed.
Claude Code (Anthropic)Project documentation and the reflection draftUse only information supported by the actual project history.A compliance check against the brief, these AI tables, the reference list, and a draft of reflection sections 1 to 3.Limited to facts in the project record. The team supplied the member details and the points for the collaboration section.