The caller's whole story, before hello.
Agent Desk is the workspace contact-center agents use all day: voice, video and every chat channel, an AI copilot at their side, and live numbers for the people who run the floor.
Every second the agent searches, the caller waits.
Contact-center agents handle one conversation after another, often for banks and utilities where customers are already stressed. Before the redesign, the answer to a simple question could sit in a different system from the call itself.
The goal: give the agent everything for the current conversation in one workspace, and give supervisors a live view of the floor, so nobody has to ask "who is free?".
We also integrated an AI chatbot into the same platform. It answers routine questions around the clock, in chat and in the mobile app, and when it can't help, it hands the conversation to an agent with a short summary, so the customer never has to repeat their story.
Too many tabs
Customer data, call controls and notes lived in separate tools, so agents switched windows while the caller waited.
Scripts hidden in documents
The right procedure existed, but finding it mid-call meant searching through long documents.
Supervisors saw problems late
Queue spikes showed up in reports after the fact, not while there was still time to react.
Measured against the best in the world.
We mapped what the global leaders, Genesys Cloud and Zendesk, offer a modern contact center and grouped it into six capability areas. Then we compared flows one by one: taking a call, verifying a caller, handing over from a bot and running the floor at peak time.
Zendesk's own guide to the best contact center software lists what the category now expects: a unified agent workspace, AI agents, live transcription with real-time sentiment, automated callbacks, call deflection, AI-driven quality assurance and real-time dashboards. We used it as a checklist.
We sat next to agents and supervisors during their busiest hours, listened to recorded calls and read the chats where the bot gave up. The map below shows the bar the leaders set, and how Agent Desk answers it for regional banks.
Routing
Skills-based and predictive routing, inbound and outbound, IVR.
- ACD
- Predictive routing
- IVR
AI automation
Real-time agent assistance, virtual agents that act in context, AI-powered knowledge.
- Agent copilot
- Virtual agent
- Knowledge
Digital engagement
Live chat, messaging apps, SMS, email, video and co-browse in one desktop.
- Video
- Co-browse
Journey management
Journey analytics and orchestration across interactions and events.
- Journey analytics
- Predictive engagement
Workforce optimization
Forecasting and scheduling, quality monitoring, gamified performance.
- Forecasting
- Scheduling
- Gamification
Reporting & analytics
Real-time dashboards, speech and text analytics, quality assurance.
- Real-time
- Speech analytics
- QA
The first 30 seconds decide the call
Agents who knew who was calling and why before saying hello finished calls calmer and faster.
Bots fail at the handover
Customers didn't mind a bot. They minded explaining everything again to a person afterwards.
Supervisors need minutes, not reports
A queue spike is only useful information while there is still time to move people.
Agent Desk: a calm voice in a loud room.
Agent Desk is the platform's own brand, used by agents all day and by supervisors on the big screen. The mark is a headset whose microphone ends in a spark: the person on the call, and the AI that listens along.
One confident indigo carries every primary action. Violet belongs only to AI, so people always know where a suggestion came from, and mint and coral are kept for good news and things that need you now.
Colour in use
A dark, quiet base is easier on the eyes over an eight-hour shift. Colour appears only when it means something, so a coral alert is impossible to miss.
Typography
Voice
Seven workspaces, one platform.
Agents get a workspace for the live conversation, one inbox for every channel and the customer's whole journey. Supervisors get the floor, CX and AI impact, the schedule and the analytics. One dark interface built for eight-hour shifts, one set of components, one AI copilot across all of it.
Face to face, with an expert at your side.
Calls can turn into video with one click when showing is faster than telling. The transcript and sentiment update live, and Copilot listens along: it suggests the answer, finds the procedure and drafts the summary before the call ends.
- Video with live captions and recording
- Transcript with key moments highlighted
- Sentiment curve across the call
- Suggested answer, steps and auto summary
Every channel, one conversation.
WhatsApp, SMS, email and web chat land in one list. The virtual agent solves routine questions on its own and hands over the rest with the intent, what was already tried and the customer's mood. Agents can co-browse or switch to video without losing the thread.
- All channels and callbacks in one list
- AI summary: intent, what happened, mood
- Co-browse and video one click away
- Predicted satisfaction before replying
See the friction before the customer mentions it.
Every touchpoint across app, web, email, chatbot and calls sits on one visual timeline. Friction is marked in red, and predictive engagement shows how likely the customer is to come back if the problem isn't solved now.
- Journey across five channels
- Friction points marked automatically
- Recontact risk, live
- Next best action, ready to start
See the spike before it happens.
Live numbers with trends, today's interactions against the forecast with the predicted peak marked, service level against target, and alerts that come with an action: join a call, whisper to an agent, move people to the queue.
- KPIs with live sparklines
- Forecast vs actual, peak predicted
- Agent states and channel mix
- Alerts with one-click actions
Measure what customers feel, not just what agents do.
NPS, satisfaction and first contact resolution sit on top, with repeat contacts as the honest check. Below: how conversations shift from calls to messages, how callbacks replace waiting on hold, how often agents accept Copilot, and who actually resolved each conversation.
- NPS, CSAT, first contact resolution, repeat contacts
- Volume by channel: calls, messages, chat, email
- Callbacks completed, scheduled and missed
- Copilot suggestions, acceptance and time saved
The right people, at the right hour.
Forecast demand against scheduled staff, hour by hour, with the gaps in red. Shifts show what every agent is doing now and next, and a light, friendly leaderboard turns quality and satisfaction into something the team actually looks at.
- Forecast vs staffing with gaps marked
- Shifts by channel, break and coaching
- Shift swaps approved in place
- Leaderboard with points and badges
Why people call, and how it went.
Speech and text analytics turn every conversation into topics and trends. A heatmap shows when customers are least happy, and AI scores every interaction against the quality checklist, so coaching starts from facts, not samples.
- Topic trends across all channels
- Sentiment heatmap by day and hour
- Every interaction scored, not a sample
- Customer words, straight from transcripts
Dark, calm and readable for eight hours.
Agents look at this screen all day, so the interface is dark and quiet, and colour is reserved for meaning. Every chart, status and alert uses the same three signals, and AI always looks the same, so people know what came from a person and what came from the copilot.
Signals
Three colours, one meaning each, on every screen.
AI marker
People always know what the copilot wrote.
Charts
Solid is what happened, dashed is the forecast.
Motion
If it moves, it is happening right now.
What changed.
Agents stopped hunting for information and started talking to people. Routine questions are answered by the AI chatbot at any hour, and every chat it hands over arrives with a summary.
The product has been live since 2021 and keeps evolving, with new channels like video added on the same components.
The handover is the product
A bot is judged by what happens when it gives up. A good summary made the whole channel feel smarter.
Context beats speed
Agents worked faster once they stopped looking for things, not once they were asked to hurry.
Copilot that acts
From suggesting the next step to doing it, with the agent's approval, inside the same panel.