You are building AI into your product.
Who is designing the trust layer?
Chatbots, decision engines, recommendation systems, and conversational AI are powerful. They are also the AI systems most likely to fail publicly, damage user trust, or trigger regulatory action when the human layer has not been designed. We make sure it has been.
Building the AI is the easy part.
Your product works. Your users do not trust it.
The model is performing. The integration is live. The feature is shipped. But users are abandoning it, support tickets are climbing, and the product is not delivering the value it was supposed to. This is the most common failure mode in AI product development and it has nothing to do with the model.
The trust layer was never designed.
Every AI product makes decisions that users need to understand, challenge, and recover from. When those capabilities are missing, users lose confidence and stop engaging. In regulated sectors, missing them is not just a UX problem. It is a legal one. The EU AI Act, GDPR, and sector-specific regulations all place obligations on AI systems that interact with users. Those obligations must be designed into the product, not added afterwards.
Designing AI products people can trust
Trust Layer Integration
We design and build the Trust Layer into your AI product. This means the interface decisions, interaction patterns, and communication structures that make your AI explainable, recoverable, accessible, and subject to human control. We work alongside your product and engineering teams from design through to deployment.
AI Product Trust Audit
We assess your existing AI product against the four Trust Layer dimensions and EU AI Act requirements. You receive a structured gap analysis with specific, prioritised recommendations and an honest assessment of your current compliance and reputational exposure.
Pre-Launch Trust Review
Before you ship, we review your AI product's interface, interaction design, and user communication against Trust Layer standards and relevant regulatory requirements. You go to market knowing the human layer has been properly designed.
Four things every AI product must get right.
Explainability
Users must be able to understand what your AI decided and why. For a chatbot, this means knowing what it can and cannot do. For a recommendation engine, it means understanding why a result was shown. For a decision system, it means having a clear, accurate explanation of every consequential output.
EU AI Act: Article 13
Human control
Users must be able to override, escalate, or exit any AI interaction. A chatbot that traps users in a loop is not just a bad experience. It is a compliance failure. Every AI product needs clear, accessible pathways to human control at every decision point.
EU AI Act: Article 14
Recoverability
When your AI makes a mistake, users must be able to recover without losing data, time, or confidence. Error states, fallback pathways, and recovery flows are not edge cases. They are the moments that determine whether a user trusts your product long-term.
EU AI Act: Article 15
Accessibility
If your AI product is used by the public, accessibility is not optional. Users with disabilities, low digital literacy, or limited language proficiency must be able to use and understand the system. We design to BITV and WCAG 2.1 AA standards as a baseline.
BITV / WCAG 2.1 AA
The most common AI product. The most common trust failures.
Chatbots and conversational AI interfaces are the AI product most organisations build first and most frequently. They are also the AI system with the highest rate of trust failure. Users encounter chatbots that cannot explain their limits, cannot escalate to a human, cannot handle unexpected inputs gracefully, and cannot be held accountable for the advice they give.
In consumer contexts, this is a brand problem. In regulated sectors, it is a legal one. A financial services chatbot that gives investment guidance without adequate disclaimers. A healthcare chatbot that handles sensitive disclosures without a clear human escalation pathway. A government chatbot that gives incorrect information about entitlements without an audit trail.
We design conversational AI products that are trustworthy by design: clear about what they can and cannot do, consistent in how they behave, accessible to all users, and always within reach of a human when it matters.
Teams building AI products that interact with real users.
Product managers and heads of product
Responsible for AI features and products that go in front of customers or end users.
CTOs and engineering leads
Building AI into products and needing the human layer specified clearly enough to implement.
Design and UX teams
Responsible for the interaction design of AI products and needing a framework that makes trust designable.
Legal and compliance teams
Managing EU AI Act and sector-specific obligations for customer-facing AI systems.
Innovation and venture teams
Taking AI product ideas to market and needing trust and compliance built in from the start, not retrofitted after launch.
We have built AI products from the inside.
Most AI product consulting focuses on the model, the integration, or the business case. We focus on the human layer: the interaction design, the communication structure, the governance mechanisms, and the user experience of being on the receiving end of an AI decision.
We developed the Trust by Design framework through research across 14 AI implementation teams in the German public sector. We have led the design of Germany's first BITV-certified government application, used by over a million people. We have seen what happens when the human layer is missing and we know exactly what it takes to get it right.
We work with your product, engineering, legal, and design teams to make the Trust Layer a concrete, implemented part of your product, not a risk register entry or a compliance document that no one reads.
Workplace AI
Deploying AI internally? Agentic workflows and internal platforms need governance too.
Workplace AI →Let's look at what your AI product needs.
A 30-minute discovery call is enough to assess your current product against the Trust Layer dimensions and identify the gaps before they become problems. No pitch. No obligation. No invoice.