When BB Biotech needed their IR (Investor Relations) website rebuilt on a less common CMS (Content Management System), the team faced a hard deadline tied to a financial report publication. At that point, the question was not whether AI-assisted development was worth trying. It was whether the project was even deliverable without it.
BB Biotech’s IR website: what needed to change and why
BB Biotech is a listed investment company focused exclusively on the biotech sector. Their website serves a demanding audience: investors, analysts, and financial professionals who expect fast, reliable access to relevant information. No patience for buried content. No tolerance for confusing navigation.
what. was brought in to rebuild their digital presence from the ground up. Before any design work began, the team ran a full IR and website analysis.
The findings were clear. Content wasn’t aligned with the core audience, navigation had too many layers, and important information was genuinely hard to find. The structure had grown organically in ways that no longer served users in practice.
The brief: fix all of that, redesign the visual identity, and go live in time for BB Biotech’s financial report publication in February. Total runway: around 2.5 months.
Building in Neos CMS with a tight deadline and thin documentation
BB Biotech’s website runs on Neos CMS, embedded within a broader infrastructure connected to other group websites. That meant working inside an existing technical environment rather than starting fresh.
Neos is a capable system, but it has a smaller community than WordPress or similar platforms, which means thinner documentation, fewer forum answers, and slower UI development due to its older architecture.
This is where most agencies would’ve struggled. And honestly, without a deliberate approach to AI-assisted web development, what. would have too.
The combination of an unfamiliar CMS, limited community resources, and a fixed deadline is exactly the scenario where traditional development approaches start to crack. You can’t just Google your way through gaps that don’t have good answers online.
How AI-assisted development actually worked in practice
what. used Claude Code directly inside the Neos codebase – not as a general-purpose chatbot, but as a structured development tool. The AI analyzed existing project code and available documentation to generate working components, translations, and related code that matched how the project was built.
The team at what. created project-specific guidance for the AI early on. This meant the outputs followed the project’s conventions from the start, rather than producing generic code that needed heavy correction afterwards.
The speed gains showed up in specific, concrete ways:
- Less time lost to documentation gaps. Instead of hours searching through sparse Neos resources and community forums, developers could close knowledge gaps quickly and keep moving.
- Faster component delivery. Repetitive tasks like scaffolding components and setting up translations moved significantly faster.
- Design to implementation in minutes. UI components went from Figma designs to working Neos implementations far faster than traditional development would allow.
For a CMS ecosystem as small as Neos, this wasn’t a marginal efficiency gain. It was what made the timeline achievable.
AI doesn’t just help you code faster – it helps you stay unblocked when the usual resources aren’t there.
Where AI-assisted development has real limits
Being honest about limits is more useful than overselling the technology.
AI didn’t always handle Neos-specific edge cases correctly on the first attempt. Translation logic and caching behaviour in particular needed human review and correction before they were production-ready. Outputs still needed validation to make sure they matched project standards throughout.
The lead developer at what. retained full ownership of the most complex implementation decisions. AI acted as an accelerator for the surrounding work, not a replacement for engineering judgment on the hard problems.
To run better from day one, what. would now:
- Document key project patterns and conventions before development begins
- Build project-specific AI guidance earlier in the process
- Define reusable workflows upfront that both developers and AI can follow consistently
The earlier that groundwork is done, the more accurate and consistent the AI outputs are throughout the project.
What this project changed about how what. works
A few things shifted permanently after this project.
The most significant: AI-assisted web development is now a standard part of how what. scopes and plans projects, especially when the CMS or technical environment sits outside the mainstream. It’s not an experiment or a shortcut. It’s a legitimate method for closing knowledge gaps fast and maintaining pace under real constraints.
The project also challenged the idea that you need to know a CMS well in order to achieve good results. Neos was unfamiliar territory. The project delivered on time. That opens up a wider range of technically viable options for clients whose infrastructure doesn’t fit the most common stacks.
And one process learning that applies well beyond this project: invest in project-specific AI guidance early. The difference between AI that produces consistent, on-convention outputs from week one and AI that requires constant correction is almost entirely down to how well the project context was established upfront.
What’s still evolving
The website launched on time for the financial report publication. The immediate goals were met.
That said, a relaunch is a starting point, not an endpoint. The simplified architecture and cleaner content foundation make it much easier to build on over time, whether that means content updates, further refinements to the IR user journey, or expanding the visual language as the brand evolves.
If you’re planning a website relaunch, whether that means restructuring your content architecture, redesigning for a specific audience, or building on a less common CMS, what.’s website services cover the full process from strategy through to go-live. AI-assisted development is part of how what. keeps ambitious timelines realistic.
Read also: What we learned about content, design, and SEO during the Rudolf Hirt AG website relaunch – a different project, but a lot of overlapping lessons about what actually matters when you’re rebuilding a site from scratch.
