Case study 03 · AI · Telco and contact centres · SaaS
Rewind AI recordings app
Product design for a recording and playback app. Find any customer interaction, listen without losing your place, and follow a call across every participant and transfer.
- Role
- Product strategist and owner
- Built for
- Supervisors and quality teams
- Sector
- Telco and contact centres
- Platform
- SaaS, cloud-based system
- Year
- 2022
Project Overview
Contact centres record thousands of conversations, and the recordings only matter if someone can find the right one and understand it quickly. This app is the place where supervisors, quality analysts and compliance teams do that. They search the list, play a call, and read exactly who said what.
The hard part is that a call is rarely just a customer and one agent. It can include two agents, three parties, a blind transfer or a consult transfer. The design makes every one of those readable in the same screens.
Platform
- Desktop
- Web app
- Voice and digital interactions
The app is used on wide screens during review sessions, often for hours. Dense tables and precise playback controls mattered more than decoration.
Challenge
Reviewers need to move fast through a long list and still be able to go deep on a single call. A list that is too thin hides the details they filter by. A playback screen that is too simple hides who was on the call and when.
Find a call in a few steps, from a list that can hold thousands of rows.
Play a recording without leaving the list.
Show every participant on one timeline, including agents added by a transfer.
Keep the list clean with archive, and make archive safe to undo.
Strategy & Approach
I designed the list as the home of the product and let everything else support it. Three decisions shaped the work.
- Table first, player docked
The list stays on screen while a recording plays. A docked player holds the controls, so a reviewer can skim the next call while the first one finishes.
- One lane per participant
The detail screen puts every participant on a shared timeline. You see who spoke, when they spoke, and where voices overlapped.
- Link the legs
Transferred calls leave more than one recording. Icons, nested rows and View Recording links tie the legs together so none of them is lost.
My Contribution
I owned the design from the first flow diagram to the clickable prototype: the structure of the list, the filter and archive model, the playback screen, and the way different call types appear in the participant timeline.
- Flow design for the transfer journey
- Table, filter and player components
- Participant timeline for each call type
- Interactive Figma prototype
Scope
The prototype covers the transfer interaction flow, starting from the Interactions list.
- Interactions
- Transfer interaction
User Research/Journey
- User Interviews
I ran 14 one on one interviews: 6 contact centre supervisors, 4 quality analysts, 2 compliance officers and 2 system administrators. Each session lasted about 40 minutes and started with the person showing me how they find and review a call today.
Finding the right call takes too many filters and too many clicks, so people keep their own notes of call times.
When a call was transferred, reviewers had to guess that a second recording existed and hunt for it.
Quality analysts listen to many calls in a row. Any control that takes them out of the list slows the whole session.
User Personas
Three personas came out of the interviews. The first two shaped most of the list and playback decisions.
- Fatima Noor, quality analyst
Reviews 25 to 40 calls a day against a scorecard. Needs fast filters, a stable list and playback she can control without the mouse leaving the table. “I lose my place every time a call opens in a new screen.”
- Usman Tariq, contact centre supervisor
Listens to calls to coach his team. Needs to know who was on a call and who handled which part. “When a call was transferred I never know if I heard the whole story.”
- Zainab Khalid, compliance officer
Pulls specific calls for audits and keeps a clean working list. Needs reliable archive and exact dates. “I need to find a call from three weeks ago and be sure I am looking at the right one.”
Research Workflow
- Interviews and call reviews
- Task and flow mapping
- Wireframes
- Prototype and walkthrough
Interviews and recorded review sessions showed how each group works. I mapped the work into four tasks: find, listen, understand and keep. Wireframes compared a table against cards for the list. The Figma prototype was walked through with reviewers, and each walkthrough fed the next round of changes.
User Roles & Priorities
Quality analyst: speed through many calls. This shows up as a docked player and persistent list.
Supervisor: understand who handled what. This shows up as participant lanes and linked transfer legs.
Compliance officer: precision and a tidy record. This shows up as filters with exact ranges, archive with undo, and share and download actions.
Defining The Architecture
The structure is flat on purpose. The list is the root. Filters open as a panel, the player docks to the bottom, and the detail screen holds everything a single call needs.
Reviewer Journey Map
The journey follows one review from finding the call to filing it away. Each stage has a need and a design answer.
Flow: Transfer interaction
A transfer icon next to an agent name shows that the call moved between agents. Clicking the name expands the row in place and shows the other leg, with its own date, agent, channel, direction, contact and timings.
Both legs stay in the list. Nothing is hidden behind a separate screen.
- Blind and consult transfers
The prototype separates two kinds of transfer. In a blind transfer Agent A passes the customer to Agent B without speaking first. In a consult transfer Agent A holds the customer, calls Agent B, then hands over. Both leave linked recordings.
Interactive prototype
A working feel of the recordings app. Press play for a guided walkthrough, or open any recording yourself. Filter the list, play a call, jump between chapters from the AI summary, and watch every participant on one timeline.
Usability Testing Based On Heuristic Evaluation
I reviewed the flows against standard usability heuristics. This is where each one shows up in the design.
- Visibility of system status
- Recognition over recall
- User control and freedom
- Consistency
- Error prevention
- Flexibility and efficiency
- Minimalist design
Visibility of system status comes from the playhead, the time readout and the applied filter chips. Recognition over recall comes from channel icons and participant names in every lane. Control and freedom comes from Unarchive and Reset. Error prevention comes from the disabled Apply button and the tooltips on row actions.
- A/B Five Minute Testing
I tested with 7 of the interviewees across two rounds. Round one was a five second test of the list. Round two was a click test with tasks such as “find the call from last Tuesday with Savannah” and “find who took the second half of this transferred call”.
- Conclusion & Learnings
Testers found the list quickly, so the column set stayed. People missed the transfer icon at first, so I added a tooltip and the nested row. Colour in the three party lanes was the clearest change. Testers said they could tell who spoke without reading names.
Some key takeaways from this project are
Keep the list on screen while listening. It made review sessions feel faster than any control I added.
Show every participant. A single waveform hides the story of a transferred call.
Make undo visible. Archive felt safe once Unarchive was one click away.
Define call types early. Writing down blind and consult transfers with engineering saved weeks of rework.
Next time I would test long recordings sooner, because a 60 minute call stresses the timeline in ways a 17 minute one does not.
Thank you for your time!