Overview

A conversation designer finishes an assistant for a client. It handles the happy path, the edge cases, the tone. Now they need feedback, and the options are grim: publish it to production and hope nothing goes wrong, or record a screen video that stakeholders can watch and never talk to.

VoiceXD supported the full lifecycle of assistant creation, from flow design to publishing, and this was the missing step. I owned the design of Sharable Prototypes, a feature that lets teams share assistants as interactive, branded prototypes without pushing them to production.

My role: Product Designer

Responsibilities: User Research, Product Strategy, Interaction & Visual Design, Prototyping, User Testing

Team: CEO, CTO, and 2 engineers

Timeline: 4 weeks

Where did testing break down?

Before this feature, testing lived entirely in workarounds: publishing assistants just to try them, and sharing screenshots or screen recordings that stripped away the one thing that makes a conversation a conversation: the ability to respond.

The costs showed up downstream: bugs and broken flows surfaced only after publishing, designers hesitated to ship without proper validation, and reviews ran through screen shares instead of real interaction. In my conversations with teams, the feeling underneath was consistent: publishing carried a quiet anxiety, because the first real test of an assistant happened in front of real users.

Success meant:

• Designers validate assistants quickly, without publishing
• Reviewers experience assistants in a realistic, interactive format
• Creators control what is tested, for how long, and how it appears
• The business gains a shorter path from design to publish, and new users experiencing assistants firsthand

Who was I designing for?

The primary users of this feature were the internal product teams building and validating assistants. Early on, I focused on who was involved in testing, what broke down for each of them, and why the existing workarounds weren't scaling. Three user types emerged.

Conversation Designers

They want to validate flows without touching production

They design the assistant logic and flows, and need a fast way to check conversational paths. Publishing to production or maintaining throwaway versions just to test slowed every iteration.

Subject-Matter Reviewers

They want a realistic experience with zero setup

Often non-technical, they review accuracy, tone, and coverage. They need to experience the assistant the way an end user would, without editor access or setup overhead.

Internal Stakeholders

They want clarity and control over what ships

PMs, QA, and leadership evaluate readiness and risk. They need to know what is being tested, trust that it reflects production behavior, and control who can access it and for how long.

Why not just a share link?

The simplest version of this feature is a share button. One click, link copied, done. It matched how designers wanted to move, and I started my explorations there.

When I stress-tested that direction against the three user types, its flaws showed from both ends.

For testers, the link carried no intent. Opening straight into a simulated chat left reviewers guessing: what am I testing, which scenario, what counts as feedback? The quality of feedback depends on how focused the testing experience feels, and a bare link provided no focus at all.

The fix in the other direction failed too. An expressive upfront configuration gave designers full control over scope, interaction modes, and context. It also turned every quick share into a setup task, exactly the overhead this feature existed to remove.

My explorations converged on three rules:

• Designers need control, but only when it adds value
• Testers need clarity, and too many options cause confusion
• The system should carry intent from creator to tester without manual explanation

What did we ship?

Creator Experience

Every testing decision in one place

Prototype configuration lives within the assistant settings, using progressive disclosure so quick shares stay quick and deliberate setups stay possible. It answers the stakeholder's control questions before they're asked.

Users can:

• Choose the interaction mode (chat, voice, and combined)
• Share a specific version of the assistant for testing
• Set access rules, including temporary or persistent links
• Apply brand styling so the experience feels polished and stakeholder-ready
• Select which scenarios or situations are available in the prototype

Tester Experience

Open the link and start talking

The testing page removes every barrier between the reviewer and the conversation. No login, no instructions, no editor. The page carries the designer's intent forward: it communicates clearly that this is a test prototype rather than a live bot, offers a scenario picker when the designer enabled one, supports chat and voice to mirror real usage, and wears the creator's brand styling for context and credibility.

What began as a validation feature became an organic user acquisition channel. More users also meant more activity on the platform, which gave us more to observe and learn from, feeding directly into further improvements. A win on every side.

Did it work?

Moving at startup pace, I tested the experience with 7 early users, including conversation designers and internal stakeholders who regularly validated assistants before launch. Each participant shared a prototype, validated at least one scenario, and compared the experience to their previous testing workflow.

Multiple participants highlighted that testing without publishing helped them iterate faster.

The ability to send a link to clients and have them test on their own devices can be a game changer for our approval process.

Conversation designer, digital agency building assistants for clients

We'll be able to catch edge cases (in the conversation logic) that we would have missed until post-launch.

QA lead, SaaS company with a customer-facing support assistant

What did this teach me?

Early signals beat delayed certainty. Shipping fast meant validating with 7 users instead of a full study. Those sessions surfaced the insights that mattered and kept the feature moving without sacrificing confidence in the direction.