01Context
Marketing software is powerful but intimidating. Our users, small businesses and marketers, often know what they want (more opens, more sales) but not how the tool gets them there. An AI assistant promised to close that gap. But an assistant is only useful if people trust it enough to actually follow its advice.
As Product Design & Experience Lead I own Uma's experience end to end: how it sounds, how it guides, how it appears in the product, and how a brand-new user meets it for the first time.
02The challenge
Embedding an AI assistant into an existing SaaS raises hard UX questions well beyond the model itself:
- Trust from second one. Users are skeptical of AI. The first interaction has to prove Uma is helpful, not gimmicky.
- Personality that fits the brand. Warm and encouraging, never robotic, more like a capable colleague.
- Guidance without hijacking. Uma should offer to help, not take over. Users stay in control.
- Onboarding that shows, not tells. A first-run experience that gets users to a real win fast.
03Approach
I ran extensive UX prototype iteration on Uma inside the product, testing how the assistant should appear, speak and behave in context rather than in a vacuum, partnering closely with Tech and CS along the way.
- Conversation & personality design. Defining Uma's tone, boundaries and phrasing so every reply feels human, clear and on-brand.
- Embedded interaction patterns. Where Uma lives, how it's summoned, and how it hands control back, so it augments the workflow instead of interrupting it.
- First-run onboarding. A guided moment that introduces Uma by delivering a genuine, useful result: the fastest way to turn skepticism into trust.
- Rapid prototyping. Iterating on real mockups of the product to pressure-test flows before a line of production code.
An AI feature succeeds or fails on trust, and trust is a UX problem long before it's a model problem.
04Key decisions
- Lead with a win, not a tour. Onboarding centers on getting the user to one concrete outcome with Uma, so value is felt before it's explained.
- Proactive, never pushy. Uma suggests next steps and always leaves the decision, and the undo, with the user.
- Plain, encouraging language. Advice framed as "here's what I'd try," with the reasoning visible so users can judge it themselves.
- Consistency as a system. Uma's voice and behaviour documented so the whole team ships one coherent assistant, not many.
05Where it stands
Uma is live work: the assistant and its onboarding continue to evolve through iteration and testing. It's the clearest expression of how I lead today: owning a product-defining experience, shaping an emerging design discipline (AI UX), and keeping the human on the other side of the screen at the center of every decision.
Detailed metrics and screens available on request under NDA.