Meetingful AI
Turning long meetings into transcripts, summaries and insights people can act on.

Meetingful's AI could do a lot, and that was the problem: too much capability risked an overwhelming product. I led the UI/UX design for Phase 1, from research to final UI, working closely with the client and developers, to make AI meeting insights quick to find and easy to act on. The result: AI-generated meeting insights helped users save 60% of their time.
The challenge
Meetingful AI transcribes, summarises and analyses meetings. The technology was powerful, but people don't open a meeting tool to admire the AI. They want to know what was decided and what they need to do next.
The challenge was balancing advanced AI functionality with a product busy people could understand at a glance, across very different users and meeting types.
My role
I was the Lead UI/UX Designer, responsible for the experience from first research to final UI. I led a design team of two UI/UX designers and worked closely with the client throughout the two-month design phase.
- 01Planned and ran user research and competitive analysis
- 02Defined the information architecture and user flows
- 03Led wireframing, the design system and visual design
- 04Guided the design team through reviews and critique
- 05Partnered with developers through implementation
How I approached it
I followed a user-centred process, so every major decision traced back to something we learned from users.
Interviews and competitive analysis
Personas, affinity mapping, priorities
Information architecture and flows
Wireframes, design system, UI
Usability testing and iteration
What users told us
I interviewed 12 people who regularly run or attend meetings, to understand how they capture and use meeting outcomes today. Three insights shaped the product:
- 01People rarely reread full transcripts. They want the decisions and next steps, not a wall of text.
- 02Transcripts lose the emotion of a meeting. Text alone can’t show how a discussion actually felt, or whether people left satisfied.
- 03Juggling multiple tools is a headache. People recorded in one tool, took notes in another and wrote reports in a third.
Competitive analysis
I mapped Meetingful against Otter AI, Fireflies, Read AI and Noota. The common gaps were long transcripts that are hard to scan and limited support for multiple formats. Those gaps became our opportunity: lead with insights, and let people take results anywhere.

From insights to priorities
I created two personas from the research. John, a marketing manager in Toronto, loses hours turning meetings into reports. Emily, a business executive at a tech startup, struggles to track action items across tools. They gave the client and the design team a shared picture of who we were designing for.


I then ran affinity mapping on feedback from two groups: everyday users and admins. Users asked for calendar integration, simple onboarding and clear meeting lists. Admins wanted satisfaction insights, flagged follow-ups, audio playback to check transcript errors, and folders that organise themselves. These clusters became the product priorities.

Key decisions
These were the calls that shaped the product most.
Summary first, transcript second
We could open each meeting on its full transcript, the way most competitors did, or on what the meeting produced. Research showed people rarely reread transcripts, so I made the home screen a Latest Meeting Recap: a short summary, key topics and counts of actions, opinions and ideas. Each meeting page then breaks the detail into tabs for takeaways, key topics, highlights, pulse and Ask Meetingful AI. Trade-off: people who want the raw record take an extra step, so search and Ask Meetingful AI stay one click away on every screen.
Making AI insights trustworthy
AI summaries are only useful if people believe them. Rather than presenting insights as final answers, I tied every highlight to a speaker and a timestamp and labelled it as an opinion, idea or action item, so users can see exactly where it came from. To capture what transcripts miss, Meeting Pulse shows each participant’s sentiment across the meeting. Trade-off: more information on screen, so I kept these as separate tabs and filters instead of showing everything at once.
Structure and flows
I structured the product around four areas: Home for the latest meeting recap and quick actions, All Meetings for recorded, analysed and upcoming meetings, the Multi Meeting Wizard for cumulative analysis and comparing meetings, and Ask Meetingful AI, a chatbot for questions across meetings. This kept single-meeting tasks simple while giving power users cross-meeting analysis in one place.

I mapped flows for the main scenarios. The meeting report flow, for example, lets users view, edit, share or download a report as PDF or DOCX, which answered the need for multiple formats from our competitive analysis.

Design system and UI
I started with low-fidelity wireframes to agree on structure with the client before investing in visual design, then refined them through several rounds of feedback.
I built a design system so the design team could work in parallel without the product drifting. It uses Inter for clarity in tables, numbers and small text, an Oxford Blue and Periwinkle palette, a shared icon set, and a component library including meeting cards, upload cards, notifications, pills and the meeting timeline used across highlights and pulse.
Built on Material UI's proven, accessible component library and customised to the product's brand, so designers and developers worked from the same component foundation and could build faster.
Colour, typography, spacing, radius and elevation defined once as tokens and reused across every component.
Tokens set up as Figma variables, so one change updates every component and screen consistently.
Documented components, states and spacing rules, giving designers and developers one shared reference.
The visual design focused on making dense information easy to scan: clear hierarchy, generous spacing, and summaries placed before details.


The final product
Every screen follows the same idea: lead with what the meeting produced, and keep the detail one step away.






Designed for John and Emily
Every Phase 1 feature traces back to a pain point from research. Here is how the design answers each persona.
- Inefficient meeting documentationAI summary in the Latest Meeting Recap
- Hard to extract insights from long meetingsKey topics and the Meeting Highlights timeline
- Time-consuming manual reportsOne-click Generate Report, as PDF or DOCX
- Difficulty tracking minutesEvery highlight tied to a timestamp
- Lack of actionable insightsAction items identified and labelled automatically
- Poorly documented discussionsSummaries, takeaways and highlights for every meeting
- Unproductive meetingsMeeting Pulse and Meeting Comparison to spot patterns
- Juggling multiple toolsRecording, insights, reports and Ask Meetingful AI in one place
Phase 1 focused on the core value: turning recorded meetings into insights people can act on. These needs, mostly from Emily, are planned for the next phase:
- Real-time insights during live meetings
- Customisable meeting templates
- Action item tracking with owners and status
- Integrations with Google and Outlook calendars and tools like Slack and Trello
Testing and iteration
We ran task-based usability tests on the home dashboard, using click heatmaps to see where people went first. The results confirmed the summary-first layout worked: 94% of tasks were completed successfully, 89.7% of clicks landed accurately, 97.3% of testers followed the expected path, and the average tester completed navigation in about 25 seconds.

The outcome
Phase 1 of Meetingful AI launched with an interface that makes AI meeting insights quick to find and easy to act on, in a product that brings recording, insights, reports and questions into one place.
What I learned
The biggest lesson was that with AI products, trust matters as much as usability. Our testing showed people could find their way around easily, but whether they acted on an insight depended on believing it. Tying every highlight to a speaker and timestamp did more for confidence than any visual polish.
If I did it again, I’d bring real meeting recordings into testing earlier and extend the tests beyond the home dashboard to deeper screens like Meeting Highlights and Meeting Pulse. Those are where the AI’s output is most visible, and where its mistakes would hurt most.
Leading two UI/UX designers on a two-month timeline also taught me to settle the design system early. Once the components were in place, the team could move in parallel without the product drifting.