RMT Engineering Logo
Contact centre operations

Agent Desk

The AI enabled Agent desk of OptiML, extends one desk for every conversation the team handles. By the time the Voice, WhatsApp and Email arrive in the same agent queue, the customer's entire history is already displayed on their screens, and the wrap-up is drafted by the time the call ends.

  • Voice and chat, one queue
  • Every suggested reply carries its citation
  • Wrap-up drafted before the agent types
We reply within one business day. No newsletter.
1 Workspace for voice and chat
360° Customer timeline
Auto Wrap-up and disposition
Cited Every suggested reply

Response provided in the Live Call

An agent works with the Agent Desk throughout the call in which Voice calls, WhatsApp threads and Emails arrive in one queue. The Client history from every channel is displayed on the agents screen, a copilot reads along and drafts replies with applicable policies. Entire Call work is drafted before the agent reaches acts on the call.

Agent Workspace

Agent Screen is displayed with Call controls on one side, the customer's whole conversation history on the other and a copilot that has already found the answer while your agent is still reading the question makes the agent fully geared upto respond to any Client query effectively.

How it fits together

The path a conversation takes through Agent Desk

What an AI agent hands to a human agent: the transcript, a summary, the extracted entities, the sentiment trajectory and the reason for handover, with the conversation then assigned on skill, priority, geography and availability AI agent Voice or chat Handover bundle Full transcript, turn by turn Conversation summary Extracted entities and order references Sentiment trajectory Reason for handover Human agent Agent Desk Assigned on skill, priority, geography and availability

The bundle is assembled by the platform, not retyped by the customer. Every handover is recorded on the conversation, so the reason it left the AI agent is auditable afterwards.

By the numbers

Few Facts - at a glance

  • 1

    workspace for voice and chat

  • 360°

    customer timeline

  • Auto

    wrap-up and disposition

The process

How It Works

The four stages, end to end

A conversation reaches the desk the same way whether the customer dialled in or an AI agent decided this one needed a person. Same four steps either way. This is what happens between the ring and the closed record.

  1. The conversation arrives

    Assignment runs on skill, priority, geography and availability. The AI agent helps in call transcription, prepares the summary, with extracted entities, Creates the sentiment trajectory and the stated reason for handover arrive with it. The Client does not need to repeat any part of the conversation.

  2. Context loads first

    The customer 360 panel opens on one timeline carrying every earlier contact on every channel. The CRM account, the order history and the open tickets are read live from the systems that own them, not from a copy that went stale overnight. Nobody asks the customer to repeat what they said on WhatsApp yesterday.

  3. The copilot reads along

    Turn by turn it tracks intent and sentiment, surfaces the article that matches what was just said, and drafts a reply with the clause behind it and a groundedness score beside it. Your agent inserts it, edits it or ignores it.

  4. Wrap-up is already written

    The wrap-up comes back drafted from what was actually said, with the policy clause that was applied cited in the note and the order references already pulled out. Your agent corrects whatever is wrong and confirms. The record closes.

Features

Every capability you need in one module

1. The answer is on screen before your agent finishes reading

As the conversation runs, the copilot suggests a reply grounded in your own knowledge base and shows the clause it was drawn from. It surfaces the matching article beside the transcript rather than sending anyone off to search for it. It proposes a next best action and tracks intent and sentiment turn by turn. Your agent decides. The copilot removes the searching.

  • Reply suggestions grounded in your knowledge base, every one carrying the citation it was drawn from
  • Sentiment and intent, turn by turn
  • The matching article surfaced beside the conversation
  • A next best action proposed from the customer record and the policy that applies to this account
  • A "why this answer" view that opens the clause and the groundedness score behind the suggestion

See the groundedness gates and guardrails

Illustrative interface. Scores, intents and article references are sample data, not a performance claim.

2. Know the Client before the conversation

Most desks leave the agent to dig the history out of other systems while the customer waits. Every previous conversation on every channel arrives in one timeline instead, alongside the orders, the tickets, the CRM record, the traits, the tags and the note somebody on your team left last year, all read from the systems that own them. The agent opens the conversation already knowing the customer, because the platform does.

  • One timeline across voice, WhatsApp, web chat, SMS, social and email
  • Orders, tickets and CRM records read live from the system of record, never copied into a second database
  • Traits, tags and team notes
  • Consent and channel preference, shown before the agent replies rather than after

See ticketing and case management

Illustrative interface. Names, order numbers and dates are sample data, not a performance claim.

3. Voice and chat, one place

Answer, hold, mute, transfer, conference and record from the same workspace that handles the WhatsApp threads and the emails. Warm transfer carries the full context bundle and a conversation summary to whoever picks it up. No "let me explain the situation.".

  • Warm, cold, conference and supervised transfer
  • WhatsApp threads and emails sitting in the same queue as the calls, worked by the same person
  • Availability states, capacity limits and skill assignment
  • Per-agent event rooms, so the desk updates without anyone hitting refresh

See all eight channels in one inbox

Illustrative interface. Names, queue counts and timers are sample data, not a performance claim.

4. After-call work writes itself

When the conversation ends, OptiML drafts the disposition, the notes, the tags, the extracted entities and the follow-up action from what was actually said on the call. Nobody logs half a shift as "Other" any more. The agent reviews and confirms in seconds instead of typing for minutes. The CRM write-back is idempotent, so a retry after a dropped connection updates the same record rather than creating a second one.

  • Disposition, notes and tags drafted from the conversation itself, not from memory at the end of a shift
  • Entities pulled out and written to the record
  • Follow-up scheduled on the channel the customer prefers
  • Idempotent CRM sync, so a retry never creates a second record

Put your own conversation through it

Bring a call recording or a policy document. We ground an agent in it and run it live on the call.

Book a demo

Illustrative interface. Dispositions, notes and order numbers are sample data, not a performance claim.

5. Centralised KB and Library access

Most teams keep their best replies in a spreadsheet one person owns and nobody updates. Here they sit in the desk itself. Personal and team libraries hold the replies, with personalisation tokens filled from the customer record, and the knowledge article that matches the live turn is suggested beside the conversation instead of being searched for. Everything searchable, everything shareable, nothing anyone has to remember where they saved.

  • Personal and team libraries, shared without a copy-paste chain
  • Personalisation tokens filled from the customer record
  • Articles matched to the live turn
  • The suggested article can be opened, edited into the reply, or sent to the customer exactly as it stands

Inside knowledge retrieval and citations

Illustrative interface. Library entries and token names are sample data, not a performance claim.

Use Cases

Where Agent Desk delivers value

AI-to-human handover bundle

The handover nobody has to explain twice

Northfield Home Retail (illustrative)

Scenario

A customer on her second damaged delivery starts with the AI voice agent and asks for a person inside the first minute. Before the call rings on the desk, the receiving agent already has the transcript, the summary, the order references, the falling sentiment line and the stated reason for handover on screen.

Outcome

The human opens with the fix instead of asking her to go through it again. And the reason the conversation left the AI agent stays on the record, so the review afterwards has something to look at.

Presence, capacity and one queue

One agent, a live call and three chats

Meridian Telecom (illustrative)

Scenario

Evening volume is mostly WhatsApp with a thin line of voice underneath it. Rather than splitting the floor into a voice team and a chat team, each agent carries one voice slot and three chat slots, with skills and languages set on the same profile.

Outcome

Chats keep moving while a call is live. A supervisor changes the mix by editing capacity, not by dragging people between queues at six in the evening.

Grounded suggestions with citations

The policy answer nobody had to look up

Ashford Credit (illustrative), a mid-market lender

Scenario

A hardship query turns on a clause that changed last quarter, and half the floor is still answering from the version they learned in induction. The copilot matches the live turn to the current article and drafts the reply, with the clause it drew from and a groundedness score sitting beside it.

Outcome

The agent reads the clause before she speaks. Later, the quality reviewer can see which document the answer came from instead of guessing at it from the audio.

At a glance

Specification

The numbers and limits, without the sales copy

Specification for Agent Desk
Specification Detail
Channels handled Voice WhatsApp web chat SMS social email
Call controls Answer, hold, mute, transfer (warm/cold/conference/supervised), conference, record
Copilot Grounded reply suggestions with citations, next-best-action, live sentiment and intent
Customer 360 Cross-channel timeline, CRM records, orders, tickets, traits, tags
Handover from an AI agent Transcript conversation summary extracted entities sentiment trajectory reason for handover
Assignment Automatic, on skill, priority, geography and availability
Wrap-up AI disposition notes tags entity extraction follow-up scheduling
CRM write-back Idempotent; outcome, notes, tags and extracted entities written to the record
Macros Personal and team reply libraries with personalisation tokens
Realtime Per-agent event rooms with push updates, no page refresh
Presence Availability state, per-channel capacity limit, skill set
FAQ

Questions,
answered

What teams ask us before they roll out Agent Desk — how it works, what it needs from your side, and what happens when it gets something wrong

Still not sure?

Talk to a specialist and get a straight answer.

Ask our team

No, the opposite. Agent Desk reads from the CRM you already run: the customer 360 panel pulls the account, the orders, the open tickets, the traits and the notes from whatever systems hold them, then writes the outcome back when the conversation closes. That write-back is idempotent, so a retry after a dropped connection updates the same record rather than creating a second one.

Nothing reaches the customer unless your agent sends it, because every Agent Desk suggestion is a draft rather than an action. Each one carries the article or clause it was drawn from and a groundedness score, and the "why this answer" view opens that source in place, so a weak answer is visible before anyone uses it. A suggestion the groundedness gate cannot support is held back instead of being shown as a confident answer.

Yes. Agent Desk presence carries an availability state, a capacity limit per channel type and a skill set, so one agent can hold a voice conversation alongside three chats, or be restricted to one thing at a time. Assignment respects those limits. An agent already at their chat limit is passed over until a slot frees, and a supervisor changes the mix by editing capacity rather than moving people between queues.

Four things: the telephony you already run for voice, the messaging channels you already own, the record systems the customer 360 panel should read, and the knowledge the copilot should ground on. Channels and record systems connect to Agent Desk through OptiML adapters and APIs. The knowledge side usually takes the most setup time, and it is the part worth spending time on, because the quality of every suggestion depends on what you point it at.

We run the platform and you choose the region the data sits in, so the transcripts, the recordings and the customer records Agent Desk touches stay in the jurisdiction you nominate. Copilot suggestions are grounded on your own knowledge sources rather than on anything pooled with other tenants. Every handover, every suggestion and every wrap-up edit is written to the conversation record, so a reviewer can reconstruct what happened months later.

Keep exploring

Additional supporting modules

Native modules that share the same record, knowledge and governance

Platform module

Supervisor Console

The live floor above the desk. Listen, whisper or take over any conversation, with SLA-at-risk alerts arriving before the breach.

Explore
Platform module

Chat & Omnichannel

Eight channels behind one conversation model, so the desk shows one customer instead of six fragments of one.

Explore
Platform module

Quality Management

Every interaction scored on 15 dimensions, so coaching the desk starts from evidence rather than from a 2% sample.

Explore
Connected solutions

Where Agent Desk is used

The Solutions pages that lean on this module, and what it looks like once it is configured for a particular floor, job title or job to be done.

12 solutions built on this module

By industry

Sector floors that run on it

4

By role

Job titles that answer for it

3

By use case

Jobs it is put to work on

5
Browse every solution
Talk to a specialist

See it handle one of your real conversations

Send us one recording and the policy your agents answer from. We'll put it through the desk live on a call with you, so you can watch the context load and the copilot suggest with its citation attached. The wrap-up comes back drafted by the time the conversation ends. Not a canned demo.

  • 30 minutes
  • One of your conversations, end to end
  • Bring the agent who would have taken it

Book your slot

Leave your email and our team will come back to you within one business day.

or reach us directly
Call our team Email us

Your details stay private. We never share them.

This website uses cookies.

Cookies are small text files that allow us to create the best browsing experience for you on our site. By continuing to use this website or clicking "Accept & Close", you are agreeing to our use of cookies. To understand how we use cookies or how to manage them, please see our cookies policy.

Ask OptiML

Powered by RMT Engineering