Axior Conference Tool UX Case Study

02 · CONTEXT + CHALLENGE + MY ROLE

A meeting interface was only the starting point.

The real product challenge was deciding which capability, control and information deserved attention for a particular person at a particular moment.

THE PROBLEM

Research needs more than conferencing.

Audio, video, chat and sharing are only the substrate. Live research also needs Discussion Guides, Notes, Key Moments, Polls, Stimuli, Whiteboard, Breakouts, transcripts, Backroom collaboration and research-specific permissions.

THE REFRAME

Power without visible complexity.

Every capability adds value. Every visible control competes for attention. The system had to stay familiar while adapting by role, session state and research task.

MY ROLE

Lead Product Designer

I led experience strategy, critical workflows and UI, cross-role consistency, component behavior, engineering clarification, implementation review and design-led UAT—without claiming solo authorship of every screen.

DESIGN PRINCIPLE

“Use a familiar meeting model as the foundation. Adapt capability by role, state and research task.”

03 · PRODUCT CONTEXT + ACCESS

The conference is the live interaction layer of a research study.

Study context, identity, schedule, permissions and research configuration already exist before anyone joins. The conference had to inherit that context rather than behave like a disconnected call.

AXIOR ECOSYSTEMClient · Audience · Expert · Mobile

↓

STUDY CONTEXTIdentity · Schedule · Role · Permissions · Method

↓

CONFERENCE TOOLAccess → Readiness → Waiting → Live Research → Recovery → Takeaway

Original portfolio UI concept showing join, waiting and live research states — Original case-study UI concept: entry confidence, waiting-room state and research activity are treated as one continuous experience.

01Client login + verification

→

02Participant login / guest route

→

03Guest identity → participant

04 · USERS + ROLE COMPLEXITY

The same meeting created different products for different people.

Nine role variants appear across the project evidence. The scalable question was not ‘which screen belongs to each persona?’ but ‘which capability should the shared shell expose under this role and state?’

SHARED SHELLFamiliar meeting grammar

ROLEModerator · Participant · Host · Observer

STATEBefore · Waiting · Live · Recovery

→Right capability at the right moment

Researcher / Moderator

Aditi Mehra

Conduct structured research without letting the tools compete with the participant.

Needs: Guide · Notes · Key Moments · Stimuli · Polls · research collaboration
Pain points: Cognitive overload · tool switching · public/private confusion

Participant

Priya Nair

Join confidently, understand what is happening and concentrate on responding.

Needs: Easy access · device confidence · consent · clear waiting state · recovery
Pain points: Did my mic work? · Did the host see me? · Why am I waiting?

Host / Client

Kavya Sharma

Support the study with the right oversight and operational control.

Needs: Admission · participant state · recording · polls · transcript · private collaboration
Pain points: Role confusion · exposing internal discussion · admin taking over

Expert Participant

Dr. Sameer Khan

Share nuanced expertise naturally with minimal software overhead.

Needs: Stable media · clear AI state · turn protection · simple sharing
Pain points: Interruption after a pause · network loss · AI moving too early

Synthetic representative personas — not claims about specific interviewed participants.

Moderator

“I need to capture that point before we move on.”

System responseOne-action Bookmark / Key Moment capture.

Participant

“Does the host know I’m waiting?”

System responseExplicit lobby feedback: host notified.

Host

“Who is waiting, joined or still invited?”

System responseLifecycle-based participant states + actions.

Expert

“I paused because I’m thinking. I’m not finished.”

System responseVisible AI listening + turn controls.

05 · RESEARCH + COMPETITIVE EVIDENCE

We studied where research work diverges from normal meeting behavior.

Methodologies are evidence. The design decisions are the story.

RESEARCH INPUTS

Secondary UX research · reference/competitive analysis · workflow analysis · stakeholder discussions · product input · UX Research input · engineering discussions · internal validation · implementation review · UAT.

CORE QUESTIONS

What stays familiar? Which research tools need immediate access? Which should be contextual? How do roles alter permissions? What changes when AI becomes the moderator? What happens when network or viewport conditions degrade?

LIMITATION

External-user usability testing had not yet been completed at the recorded project stage. UAT is not presented as usability testing.

EXTERNAL SECONDARY RESEARCH · NOT AXIOR DATA

Virtual meeting cognitive friction

Used to contextualize why low cognitive overhead, clear state and context recovery matter in a live interview.

EXTERNAL SECONDARY RESEARCH · NOT AXIOR DATA

AI adoption in research

Used to frame transparency, turn-taking and the distinction between AI assistant and AI moderator behavior.

EXTERNAL SECONDARY RESEARCH · NOT AXIOR DATA

Accessibility + inclusive communication

Used to reinforce captions, clear status communication and participation as research-validity concerns.

Microsoft TeamsZoomGoogle Meetfamiliar meeting grammar→Differentiate where research changes the job

CONFERENCING SUBSTRATE

Familiar by design

Mic · Video · Participants · Chat · Share · Reactions · Recording · Captions · Transcript · Views · Leave / End

RESEARCH OPERATING LAYER

The product differentiator

Guide · Notes · Bookmarks · Key Moments · Polls · Stimuli · Whiteboard · Breakouts · Backroom · Insights · AI assistance · AI moderation · Takeaways

06 · JOURNEY + PROBLEM REFRAME

Research begins before the first question—and does not end at Leave.

The lifecycle has seven meaningful stages. Readiness, recovery and post-session reflection are part of the research experience, not operational leftovers.

01Access

→

02Eligibility

→

03Readiness

→

04Waiting

→

05Live Research

→

06Recovery / Exit

→

07Takeaway

Too early

Session unavailable; explain when to return.

Ready

Study context, devices, consent, recording, identity.

Waiting

Host notified, admission state, leave option.

Live

Conversation + research tools + evidence + collaboration.

Recovery

Connection, performance, rejoin, leave/end states.

Takeaway

Call quality, learning, biggest takeaway, comments.

PROBLEM REFRAME

Create one familiar live-session model that can safely adapt to role, research activity, moderation type, screen size and system condition—without transferring that complexity to every user.

How might we preserve a natural research conversation while dynamically exposing the right tools, permissions and feedback for the current person and moment?

07 · TAXONOMY + IA + PERMISSIONS

The architecture is partly navigational and partly conditional.

A conventional sitemap could not explain this system. Role, state, platform and permission determine which capabilities appear inside the same meeting grammar.

Access & Readiness

Session Core

Mic · video · chat · participants · share · reactions · recording · language · leave/end

People & Permissions

Waiting · admit · deny · invite · resend · host · observer · translator · guest

Research Workspace

Guide · Notes · Bookmark · Key Moments · Transcript

Research Activities

Poll · Stimuli · Whiteboard · Breakout Rooms

Private Collaboration

Backroom · observer/research-team communication

AI

Assistant · AI Chat · Moderator · states · turn controls

System & Recovery

Network · performance · recording · empty state · support · errors

Post-session

Takeaway · call quality · comments

Role

Session type

Method

Permission

State

Platform

↓Visible controls · tools · participant rights · recording · polls · breakouts · AI · end authority

08 · EXPLORATIONS + CORE INTERACTION MODEL

The key layout decision was how much research UI should remain visible.

Three directions could hold the features. Only one protected the conversation.

A · REJECTED

Everything persistent

Maximum discoverability; minimum breathing room. Participant visibility collapses as tools accumulate.

B · SELECTED

Participant-first + contextual workspace

Active participant/content remains primary. Essential meeting controls persist. Research tools appear when the task needs them.

C · REJECTED

Research operations dashboard

Excellent operational overview, but the research conversation becomes visually secondary.

One meeting model. Different capability envelopes.

ProblemBespoke role interfaces would fragment learning.

DecisionPreserve the shared meeting grammar; adapt capability by role/state.

Trade-offMore conditional logic below the surface.

Keep research tools contextual.

ProblemPermanent Guide, Notes, participants, chat and activities compete with the person.

DecisionThree layers: primary workspace → persistent essentials → contextual research.

Trade-offSome tools become one reveal away.

Make AI conversational state visible.

ProblemA pause can mean thinking, finished or wanting to continue.

DecisionExpose Thinking/Listening/Waiting and explicit turn controls.

Trade-offLess ‘invisible magic’; more participant agency.

Preserve the conversation when conditions degrade.

ProblemNetwork failure can damage rapport and lose an answer.

DecisionReduce incoming video intentionally, explain why, recover visibly.

Trade-offTemporary visual fidelity loss protects research continuity.

09 · RESEARCH WORKSPACE

Guide, Notes and Key Moments had to work while the researcher was listening.

Evidence capture and study structure belong beside the conversation—not in a separate tool the moderator has to mentally re-enter.

Original desktop research-conference interface concept — Original UI concept created for this portfolio: participant-first meeting canvas with Discussion Guide, Notes, Highlights and research context beside the live conversation.

DISCUSSION GUIDE

A live moderation instrument—not a document viewer.

Topic/question structures · estimated time · Mark Done · Completed · Discussion Points · compact/expanded widths · empty state.

EVIDENCE CAPTURE

Capture now. Interpret later.

Timestamped Notes · Bookmark · Key Moment · comments/replies · mentions · edit/delete · linked timecodes.

PUBLIC VS PRIVATE

Chat and Backroom have different consequences.

Participant-facing communication remains distinct from researcher/observer coordination to avoid accidental exposure or research contamination.

PARTICIPANT STATE

Presence is a workflow.

Waiting · Joined · Invited with Admit, Deny, Admit All, Invite and Resend close to the relevant lifecycle state.

10 · RESEARCH ACTIVITIES

Activities stay inside the research conversation.

Polls, Stimuli, Whiteboard and Breakouts are not feature cards. Each is an end-to-end workflow with its own state model.

Original product UI showcase with waiting room, poll creation and live poll results — Original portfolio UI concept used to explain the poll lifecycle and waiting-room workflow in a redesigned visual system.

01Choose

→

02Preview

→

03Launch

→

04Respond

→

05Analyse

STIMULI + WHITEBOARD

The study already knows the material.

Screen/window sharing, study stimuli, documents and blank Whiteboards reduce the need to search a desktop mid-session. Whiteboard tools include move, pen, shape and text.

BREAKOUT ROOMS

Complexity stays hidden until the moderator needs it.

Configure → assign → review → start → manage → announce → close/return, including automatic/manual assignment and timing validation.

11 · AI — ASSISTANT VS MODERATOR

AI creates two different interaction contracts.

Treating ‘AI’ as one feature hides the hardest design work. A supportive assistant and an AI controlling the next research turn are fundamentally different relationships.

Original AI-moderated research interview interface concept — Original AI interview concept: visible Listening / Thinking / Follow-up states, turn protection and research context.

MODE 01

Assistant inside a human-led session

Human owns the conversation. AI is invoked for summaries, notes or questions and does not control conversational pacing.

MODE 02

AI as interviewer / moderator

AI controls the next research turn, so state, consent and turn completion need to become explicit.

Thinking…→Listening…→Take your time→Let me finish talking→Done Talking

In research, participant agency and response integrity matter more than making AI feel magical.

RESPONSIBLE AI

AI-derived behavioral or emotion-style indicators require validation around accuracy, consent, bias, culture, privacy and interpretation. AI output should support researcher judgment—not silently replace it.

12 · RESPONSIVE + THEMES + ACCESSIBILITY

Responsive did not mean squeezing the desktop interface.

We preserved the same mental model, not the same simultaneous density.

Original mobile research-conference interface concepts — Original mobile concepts for live meeting, research tools, polls and notes—designed around priority rather than desktop parity.

Original desktop research session concept — Desktop reference for the same mental model: live participants remain primary while Guide and Notes stay adjacent.

1Active research content

2Participant visibility

3Essential controls

4Role-specific tools

5Secondary actions

13 · RESILIENCE — ERRORS, NETWORK, RECOVERY

A live product is not defined only by its happy path.

The system needed to explain what happened, why it happened and what the user can do next.

PERFORMANCE ISSUE

Protect the conversation before the video.

Detect degradation → recommend turning off incoming video → pause incoming video → explain protection → restore quality → offer video return.

NETWORK LOSS

Recovery should be explicit.

Lost connection → trying to reconnect → maintain meeting identity/context → resume rather than silently reset.

No Discussion GuideExplain absenceReturn to meeting without dead-end

No stimuliExplain study has no materialContinue / choose other share option

Muted micExplain others cannot hearUnmute

Recording stateStarted / stopped feedbackMaintain visible status

Leave / EndExplain consequenceCancel / leave / end for all

Remove participantConfirm destructive actionCancel / confirm

NETWORK QUALITYConnection degraded

Reduce incoming video to protect the conversation.

Reduce video

RECOVERYReconnecting…

Meeting identity and context are preserved.

RESTOREDConnection stabilised

Video can return without resetting the session.

Resume video

DESTRUCTIVE ACTIONEnd session for everyone?

Explain consequence before the irreversible action.

CancelEnd for all

14 · ENGINEERING + UAT + ITERATION

The design was not finished when Figma was finished.

My responsibility extended through component behavior, engineering clarification, implementation review, UAT, fixes and retesting.

01Design intent

→

02Component behavior

→

03Engineering walkthrough

→

04Build

→

05Implementation review

→

06UAT

→

07Fix + retest

→

08Release readiness

ITERATION 01

Guide density

Multiple widths and structures balanced orientation against participant visibility.

ITERATION 02

AI turn-taking

Explicit turn controls clarified conversational ownership after silence.

ITERATION 03

Network degradation

Visible quality protection replaced silent failure.

ITERATION 04

Role density

Participant and Observer capability narrowed relative to Host / Co-host.

BEHAVIORAL DESIGN SYSTEM

Scalability meant reusable behavior, not just reusable visuals.

Participant tiles · meeting controls · notifications · drawers · menus · Guide · Notes · polls · Breakout Rooms · recording · AI state · network state · settings and empty states had to remain coherent across Role × Permission × Session Type × Platform × Theme × Language × Viewport × Network Condition × AI/Human Moderation.

≈100

UAT issues reported as identified / resolved

Visual fidelity · behavior · permissions · controls · research tools · alerts · mobile · network · AI · empty states · recovery.

Approximate delivery metric — not “100 usability problems.”

15 · OUTCOMES + CONTRIBUTION

Separate delivery evidence from product-impact evidence.

The project demonstrates breadth, system coherence and implementation maturity. It does not yet prove post-launch behavior or business KPIs.

DELIVERED

A 0→1 research-conferencing system.

5 primary platforms · Human + AI moderation · multiple role variants · access/lobby · Guide · Notes · Key Moments · Chat/Backroom · Polls · Stimuli · Whiteboard · Breakouts · transcription · responsive/mobile · theme states · recovery · takeaways · 80+ screens/states (approx.).

INTERNALLY IMPROVED

Clearer system behavior.

Role separation · workflow integration · explicit AI/meeting feedback · resilient edge states · implementation consistency · responsive behavior.

STILL UNPROVEN EXTERNALLY

What requires future evidence.

Task success · participant confidence · moderator cognitive load · research-tool discoverability · AI-state comprehension · mobile performance · representative accessibility testing · post-launch adoption · business KPIs.

I LED

Experience strategy · core architecture · critical flows/UI · interaction direction · role consistency · research integration · team reviews · component behavior · engineering clarification · implementation review · design-led UAT · iteration.

WE DELIVERED WITH

Senior UX Designer · junior designers · Product · stakeholders · UX Research · web/mobile engineers · QA/UAT collaborators.

16 · REFLECTION + NEXT VALIDATION

I started by asking where to place the features. I ended by asking what deserved attention.

A live system is fundamentally an attention-management problem.

The Participant needs simplicity. The Moderator needs research power. The Host needs control. The Observer needs visibility without interference. The AI needs to reveal its state. The network needs to communicate degradation. Mobile needs to hide things without making them undiscoverable.

The strongest interaction model was not the one that exposed the most capability. It exposed the right capability at the right time.

Participant onboardingGuest/login · recording · consent · devices · too-early · waiting · rejoin

Moderator attentionGuide · Notes · Bookmark · Key Moments · Poll · Stimuli without losing focus

Public / privateCan users clearly distinguish Chat from Backroom?

AI turn-takingThinking · Listening · Waiting · Let me speak · finish talking · Done

Network recoveryDo users understand why video was reduced and how it returns?

MobileDoes hierarchy survive small screens, large groups, research tools and sharing?

AccessibilityKeyboard · screen reader · captions · focus · status announcements · non-color meaning

AI InsightsConsent · accuracy · bias · interpretability · appropriate researcher use

AXIOR CONFERENCE TOOL

Protect the conversation. Design everything around it.

I led the design of a zero-to-one research-conferencing system that had to adapt across roles, research activities, human and AI moderation, mobile and desktop contexts, and real-world failure states. Complexity belongs in the system—not in the conversation.

Designing a Research-First Conference Tool

The product story, explained while it moves.

A meeting interface was only the starting point.

Research needs more than conferencing.

Power without visible complexity.

Lead Product Designer

The conference is the live interaction layer of a research study.

The same meeting created different products for different people.

Aditi Mehra

Priya Nair

Kavya Sharma

Dr. Sameer Khan

“I need to capture that point before we move on.”

“Does the host know I’m waiting?”

“Who is waiting, joined or still invited?”

“I paused because I’m thinking. I’m not finished.”

We studied where research work diverges from normal meeting behavior.

Virtual meeting cognitive friction

AI adoption in research

Accessibility + inclusive communication

Familiar by design

The product differentiator

Research begins before the first question—and does not end at Leave.

Create one familiar live-session model that can safely adapt to role, research activity, moderation type, screen size and system condition—without transferring that complexity to every user.

The architecture is partly navigational and partly conditional.

Access & Readiness

Session Core

People & Permissions

Research Workspace

Research Activities

Private Collaboration

AI

System & Recovery

Post-session

The key layout decision was how much research UI should remain visible.

Everything persistent

Participant-first + contextual workspace

Research operations dashboard

One meeting model. Different capability envelopes.

Keep research tools contextual.

Make AI conversational state visible.

Preserve the conversation when conditions degrade.

Guide, Notes and Key Moments had to work while the researcher was listening.

A live moderation instrument—not a document viewer.

Capture now. Interpret later.

Chat and Backroom have different consequences.

Presence is a workflow.

Activities stay inside the research conversation.

The study already knows the material.

Complexity stays hidden until the moderator needs it.

AI creates two different interaction contracts.

Assistant inside a human-led session

AI as interviewer / moderator

Responsive did not mean squeezing the desktop interface.

A live product is not defined only by its happy path.

Protect the conversation before the video.

Recovery should be explicit.

The design was not finished when Figma was finished.

Guide density

AI turn-taking

Network degradation

Role density

Scalability meant reusable behavior, not just reusable visuals.

Separate delivery evidence from product-impact evidence.

A 0→1 research-conferencing system.

Clearer system behavior.

What requires future evidence.

I started by asking where to place the features. I ended by asking what deserved attention.

Protect the conversation. Design everything around it.

Continue exploring complex product systems.

Unified Participant Experience

Prism Design System

Before the UI, the system had to make sense.

If the work resonates, let’s talk.