How Headspace Built an AI Companion that Fosters Trust and Transparency | Figma Blog (opens in new tab)

Headspace built Ebb as an AI companion for reflection between therapy sessions—not as a replacement for human care. Because mental-health AI carries significant safety and trust risks, the team prioritized clinical grounding, transparency, user autonomy, and careful cross-functional design. Early alignment workshops, iterative prototyping, and explicit interface guidelines helped shape Ebb into a friendly but clearly non-human companion.

Defining Ebb’s Role

• Headspace created Ebb in response to growing use of general-purpose AI for emotional support.
• The companion was intended to:
  • Support reflective practices between therapy sessions.
  • Help people who may be unable to access or afford therapy.
  • Complement, rather than replace, human care.
• The team first clarified the complete user experience and business objectives through FigJam workshops.
• This early alignment helped address internal concerns and ambiguity around using AI in mental health.

Building a Non-Human, Approachable Identity

• The team explored names including Odom, Ibo, and Scribe before choosing Ebb.
• The name suggests the fluidity and changing nature of emotions.
• Designers avoided a gendered human name to reduce stereotypes that associate caregiving with women.
• Ebb was designed as a friendly, human-adjacent entity without a specific gender.
• Brand, product, illustration, animation, and copy teams used FigJam “playgrounds” to explore how Ebb could look and sound.

Iterative Collaboration and Prototyping

• Headspace followed a “build-to-learn” approach, bringing brand and product teams together early.
• The teams tested approximately six brand identities before selecting a direction.
• Figma allowed designers to apply different visual treatments directly to product screens.
• Sharing prototypes in one workspace kept teams connected and enabled rapid iteration.
• The team deliberately stress-tested ideas before rejecting them.

Designing for Trust and Safety

• Ebb was trained with input from clinical psychologists, providing a strong scientific foundation.
• Designers focused on reducing adoption barriers by helping users understand:
  • That Ebb is an AI system.
  • How it can support them.
  • How their information and conversations are handled.
• The interface was designed to make users feel safe expressing themselves.
• Users retain agency to exit and delete conversations at any time.
• A central principle was that AI should never be invisible: members should always know whether they are interacting with AI or a human.
• The team’s broader guidelines emphasized differentiating AI from human-delivered care, reinforcing privacy and safety, supporting member choice, and creating a reflective environment.

Headspace’s approach suggests that mental-health AI should be designed transparently and collaboratively, with safety and user control treated as foundational product requirements rather than features added later.