PeakPrivacy

01 — The Challenge

AI was easy to access, but hard to adopt safely

When we started working on PeakPrivacy, the market was already full of AI tools. The problem was not access. The problem was security. Swiss companies wanted to use AI, but they needed to know what happened to their data.

They wanted to understand where their data goes, whether it stays on Swiss servers, who can access it, and which models are safe enough for sensitive work. They also needed to make sure that employees could not accidentally share confidential information in a regular AI chat.

So we designed PeakPrivacy to give teams the AI tools they need for everyday work, while keeping data inside a controlled environment with clear rules for how sensitive information is handled.

02 — Discovery & Strategy

From fragmented prototype to a clear product strategy

The product started as disconnected building blocks. Discovery came from customer calls, internal workshops, sales feedback, onboarding discussions, and prototype reviews with the team. My task was to turn these inputs into one coherent product system with clear logic, consistent design language, and a connected user journey.

Discovery showed that the main challenge was not only the interface, but adoption. Companies could introduce an AI platform, but employees would not automatically use it in daily work. They needed clear use cases, training, internal ownership, and a simple way to understand which path was right for them.

This shaped my product strategy work: I mapped the journey from awareness to onboarding and connected it with three customer readiness paths. The result was a clearer system that helped teams understand where they were, what level of support they needed, and whether the right next step was the AI Readiness Pilot, Growth Plan, or Sovereign Plan.

03 — Core Product

Designing the AI workspace

When teams started using PeakPrivacy, one thing became clear. AI work did not stay inside the chat. People copied answers into Slack, shared summaries with colleagues, discussed outputs in meetings, and tried to turn good prompts into repeatable team workflows.

That changed how we thought about the product. PeakPrivacy could not remain just a chat interface, so we started building around the way teams actually used AI: shared assistants, reusable workflows, internal knowledge, meeting summaries, and context that could be reused across the organisation.

Some ideas did not work. We originally built group video calls with AI transcription directly inside the app, thinking teams might want to run meetings in PeakPrivacy. But users kept using Slack, Teams, and Google Meet. The problem was not the call itself. The real need came after the meeting: capturing what was said, turning it into useful summaries, and making the outcome easier to reuse. So we removed the calling feature and built meeting bots that join external calls instead. That became Meetings, now one of the most-used parts of the product.

04 — Trust & Privacy

Making privacy visible in the product

Privacy was one of the main concerns for Swiss companies that wanted to introduce AI. In conversations with CEOs, CTOs, and internal teams, the same questions kept coming up: where does the data go, who processes it, and can employees safely use AI for sensitive work?

I designed the model selector so this information was visible before the user sent anything to AI. Each model showed a clear security level and processing location — Switzerland, the EU, or outside Europe — written simply enough for any employee to understand, not only technical or legal teams.

This was supported by onboarding and training, so teams understood how to choose the right model for the sensitivity of their task and avoid using AI in the wrong context.

The goal was simple: make privacy understandable at the moment of use, not only in legal text or sales conversations.

05 — Knowledge & Assistants

Turning company knowledge into reusable AI workflows

AI becomes much more useful when it can work with a company's own context, not only general prompts. I designed the knowledge structure around Core Context and Knowledge Bases: baseline company knowledge was available across the workspace, while teams could organise specific documents and sources for different use cases.

This knowledge could be used directly in chat or connected to Constructors — reusable AI assistants with predefined instructions, model settings, privacy level, and data sources. This allowed teams to create assistants for specific roles, departments, or workflows without starting from scratch every time.

Permissions were built into the system, so Knowledge Bases and Constructors could be shared with the whole company, a team, or selected employees only. Instead of rebuilding the same context in every chat, teams could create reusable knowledge systems once and apply them across chats, assistants, and workflows.

06 — Design System

Unifying the product design system

PeakPrivacy started with several disconnected surfaces: an AI chat interface, a dashboard, and a marketing website that came from different prototypes and tools. My task was to bring them into one coherent product language — visually, structurally, and functionally.

I built the first shared UI kit and design system, then used it to unify layouts, components, states, and interaction patterns across the product. This was not only a visual clean-up: I also simplified product flows, reduced unnecessary steps, and made the relationship between dashboard, workspace, billing, usage, and chat easier to understand.

Over time, the system evolved through two major iterations and supported the move toward a more unified workspace. Beyond standard UI components, it included AI-specific patterns such as model labels, security status indicators, message states, chat actions, and skeleton loaders.

07 — Website & Customer Journey

Designing the website around the buyer journey

Before designing the website, I mapped the customer journey: how different users discover PeakPrivacy, what questions they have, and which path should lead them to the right offer — Team, Growth, Sovereign, AI Readiness Pilot, Core Context Build, or training programs.

The website was not always the first touchpoint. Users could come from marketing campaigns, referrals, sales conversations, or direct recommendations. My task was to make the website work as a clear storytelling layer inside that journey: explain the product, address trust and security concerns, and guide different buyer types toward the next step.

I based the structure, copy, and page flow on customer personas and buyer objections. Swiss companies did not only want AI features; they needed to understand where data is stored, how privacy works, what support they get, and which offer fits their readiness level.

This shaped the homepage, pricing, product pages, security content, model pages, academy, feature catalogue, and contact flow — connecting product positioning, trust-building, and commercial paths into one clear experience.

Supporting brand touchpoints

I also created marketing and sales materials in the same design language: LinkedIn visuals, campaign assets, posters, advertising banners, client presentations, onboarding materials, and sales-support assets. This kept the product story consistent across advertising, social media, the website, and sales conversations.

Homepage promo video

I led the production of the homepage promo video: developed the concept, wrote the script, found the director, collected references, and guided the creative direction to final delivery. The goal was to explain PeakPrivacy in three minutes directly in the hero section — showing the product, key features, and the support behind adoption, so potential clients could quickly understand whether it was relevant for them.

08 — Outcome

From prototype to live AI product with paying clients

Over 2+ years, PeakPrivacy grew from an early prototype into a live product used by Swiss companies across enterprise and mid-market segments.

09 — Reflection

What I learned

Working on PeakPrivacy made one thing clear: people hesitate when they do not understand what the AI is doing. They want to know which model they are using, where their data goes, and what will happen next. If that is unclear, the product feels risky.

AI-assisted prototyping helped us move faster. We could turn an idea into a rough prototype, show it to the team, and quickly see if it made sense or not.

• Unclear logic creates hesitation. If users did not understand a label, flow, or model choice, they started to treat it as a risk and avoid using it.
• AI access does not mean AI adoption. Giving people the tool was not enough. Teams needed onboarding, training, and internal AI ambassadors who could help others understand where AI fits into their daily work. Without that support, many users would try the product once and then go back to their old habits.
• Some features looked useful in theory, but did not match how users worked. If a feature creates friction or asks people to change habits without a strong reason, better UI will not fix it.