Callbacks and In-Queue Treatment in Genesys Cloud: Designing a Waiting Experience That Keeps Customers and Protects Agent Capacity
A contact centre queue is often treated as a holding pen: customers wait, hear little about what is happening, and decide whether to stay, hang up, or ask for a callback. Yet the queue is part of the service experience, not just a routing step.
Callbacks are also easy to switch on as an afterthought. Without a deliberate design for how they are offered, dispatched and measured, customers may still wait, agents may have capacity that the callback process does not use, and reporting can give an incomplete picture of service.
Genesys Cloud provides several documented ways to shape this experience: separate in-queue flows for different media, estimated wait time playback, and two callback models with distinct dispatch and reporting behaviour. For organizations modernizing their cloud telephony, the important decision is not simply whether to enable callbacks. It is how to align the waiting experience, agent capacity and measurement.
For organizations reviewing a wider platform change, Dunamis’ five-step cloud telephony migration framework offers a related planning perspective.
1. Treat the wait as a designed experience for each media type
In Genesys Cloud, an in-queue flow customizes the customer experience while an interaction waits for an agent. Call, email and message interactions use separate in-queue flows, so the treatment should reflect how customers experience each channel. The Genesys Cloud in-queue flows documentation describes the available flow types and their constraints.
For calls, Architect provides a built-in default in-queue flow, but it is not automatically published. Until someone publishes it, it does nothing. The default flow cannot be deleted, and its name and description cannot be edited, although its behaviour can be changed by users with the right permissions. It is built using a Task action; Menus and Reusable Tasks are not available in this flow type.
A useful voice experience might combine hold music with a message explaining that calls are handled in the order received, an update about high call volume, or a prompt pointing customers to online support. The purpose is to set expectations or offer a helpful next step rather than leave silence or repetition as the only treatment.
Email and message flows have no built-in default in Architect, so teams must create them. These flows can, for example, send a periodic queue update or transfer a customer to another in-queue message flow after a set time. Genesys documents a maximum duration of 72 hours before an email or message flow disconnects, and limits each interaction to 30 opened in-queue flows. Outbound email and message interactions do not launch an in-queue flow until an inbound response arrives.
Recurring states also behave differently by channel. Message flows start on a ten-second period and email flows on a 60-second period; recurring states then run every five minutes of flow runtime with an additional five-second wait increment. An outside update to participant attributes can cancel a wait and trigger a recurring state immediately. Architect flow permissions are required to add, edit and view flows.

Actionable takeaway: Publish the default call flow before relying on it, then review call, email and message treatment separately, including the duration and recurring-state behaviour of digital flows.
2. Tell the truth about the wait
The Play Estimated Wait Time action belongs in an in-queue flow’s task sequence. Genesys calculates the estimate using analytics for the same half-hour interval one week earlier, including tHandle, tAcd, tAbandon, tTalk and nOffered, alongside the interaction’s position in the queue. If no agents are active, the calculation looks at interactions already waiting and uses the longest current wait. The Genesys Cloud estimated wait time documentation explains the calculation and playback options.
The estimate has limits that matter when setting customer expectations. If the queried interval has no ASA statistics, there is no threshold on the maximum estimate returned. That presents a risk for new or thin queues. Also, the value does not increase during a flow, even if the actual wait rises. A caller who hears two minutes will not later hear five minutes from this action, and the real value is not exposed in Architect. The action is intended to play an estimate, not make routing decisions.
Architect offers four playback modes: Approximate gives a rounded, non-specific estimate; Less Than gives a more specific estimate below the actual estimate; Padded gives a range; and Precise reads the estimate to the second. Approximate and Less Than use a Round Up Interval and can also use Specific Round Up Intervals. Padded uses a Padding Interval. A computed duration under one minute is treated as one minute.
Review voice queue settings in the same pass. Genesys documents an alerting timeout from 7 to 59 seconds; after timeout, the agent goes Not Responding and the call returns to the queue. A whisper prompt can tell the agent which queue the caller entered. A flow selected on a Transfer to ACD action supersedes the queue-level in-queue flow. The queue also has a Continue voice recording during queue wait option. For messaging, documented patterns include an auto-response that updates the customer’s queue position every 30 seconds or transferring the interaction to another in-queue flow after a set time.

Actionable takeaway: Choose a playback mode that matches the level of certainty the queue can support, and test the estimate against new or thin queues before using it to set customer expectations.
3. Choose Agent First or Customer First on evidence
A callback is a request for a call to be returned when an agent is unavailable to take it immediately. Genesys Cloud treats callbacks as a separate media type from voice, and a single callback can generate reporting activity on both. In the default Agent First Callback (AFC) model, an agent answers the callback offer and then calls the customer. The agent’s engagement with the callback record appears on callback media; talk and after-call work for the customer call appear on voice media.
With Customer First Callback (CFC), the system calls the customer before connecting the interaction to an agent. There is no Answered event on callback media. If the customer answers, the system routes the interaction to an agent as an inbound-routed interaction, and the agent’s handle time is on voice media. The Genesys Cloud Customer First callback documentation describes the settings and limitations.
CFC settings determine what happens at key decision points. Reserve Agent before Callback Dialing holds the best available agent before the customer is called. For a live voice answer, the system can transfer to the queue and use its in-queue flow, or transfer to a selected inbound Architect flow. For an answering machine, it can transfer to a flow to play a message or hang up. CFC requires persistent connection and auto-answer for agents in the queue to support a timely connection. Genesys lists CFC as unsupported with voicemail, preview campaigns and callback-specific agent scripts; it uses the standard voice script.
The operational choice depends on the constraint. AFC places an agent on the callback before the customer call. CFC dials the customer first, so it avoids involving an agent in the pre-dial step, but the customer must answer before an agent is connected. Organizations should compare these trade-offs with their customer reachability needs and agent-capacity priorities. For related design considerations, see Dunamis’ article on conversational IVR in Genesys Cloud.
Actionable takeaway: Choose AFC or CFC based on whether agent involvement before dialing or customer availability at connection time is the more important constraint, then confirm the model’s documented dependencies and exclusions.
4. Set dispatch and retries against real capacity
For CFC, a queue-level pacing modifier determines the probable number of callbacks dispatched at once. Genesys applies a multiplier to the number of agents on queue. Working down the documented scale, a modifier of 10 gives a multiplier of 1.00 and needs only one agent for a guaranteed first dispatch. Modifiers of 9, 8, 7 and 6 give 0.73, 0.52, 0.36 and 0.23, with minimum agent counts of 2, 2, 3 and 5. Modifiers of 5, 4, 3, 2 and 1 give 0.14, 0.08, 0.08, 0.03 and 0.02, requiring 8, 13, 22, 38 and 50 agents respectively. The Genesys Cloud queue configuration documentation details the pacing model.
For instance, Genesys documents that 10 agents on queue with modifier 10 dispatches 10 callbacks; 100 agents on queue with modifier 5 dispatches about 14. The calculation counts all agents on queue regardless of routing status, but a callback is only dispatched when an agent is available or idle and capable of taking the call. If no agents are on queue, nothing dispatches. Eligibility is checked again when agent or callback activity occurs, such as an agent joining the queue, completing a call without going off queue, or a callback arriving.
Retry settings allow 0 to 20 retries, with up to 24 hours between attempts. They apply to unsuccessful calls, such as no answer or a busy line, but not to answering-machine scenarios. Agent ownership is available for scheduled callbacks for periods from one hour to 30 days, with an option to return an unanswered callback to the queue at expiry. Ownership only takes effect when the queue has at least one Preferred Agent Routing rule, and agents can own callbacks only in an Agent First scenario. Non-owned callbacks can be scheduled up to 30 days in advance; that limit cannot be changed.
Queue settings also include Automatically start and end callbacks. That option requires Enforce Communication Level After Call Work and provides timers for when an assigned callback auto-dials and when a dispositioned callback auto-ends.

Actionable takeaway: Set pacing and retries using the queue’s actual available capacity, and validate ownership and automatic callback settings against their prerequisites before deployment.
5. Report callbacks as a channel, not an exception
Callback reporting is not a single voice metric with a different label. Genesys defines callback outcomes as Callback Reached, Callback Not Reached and Callback Abandon. Not Reached can represent no answer, a busy or invalid number, voicemail, or maximum attempts reached. Callback Abandon applies to CFC when a live customer disconnects before an agent is selected.
Callback-specific metrics include Callback Attempts, Callback Scheduled, Callback Virtual Wait and Callback Over SLA. Active Callback, Time to First Dial and Time to First Connect apply to AFC only. Callback Virtual Wait measures the customer’s actual wait for an outcome; it excludes time waiting for a pre-scheduled callback time and does not count dwell time. These definitions are documented in Genesys Cloud callback reporting.
By default, a callback contributes to service level through the standard Answered event, which aligns with AFC because an agent answers the callback offer. The optional Include Callback Virtual Wait in SLA setting instead measures Callback Virtual Wait against Callback Reached. Separate switches determine whether Callback Abandon and Callback Not Reached count in the denominator. The option is off by default and affects callback-media conversations only; voice and other media types are unchanged.
For CFC reconnects, Genesys uses a separate set of voice metrics: CFC Offered, CFC ACD or CFC Wait, CFC Answered, CFC Abandon and CFC Flow Out. This keeps a priority-routed reconnect from being counted as a fresh inbound offer and from inflating standard service level or flattering Average Speed of Answer. In mixed AFC and CFC queues, standard voice metrics and shared callback metrics can roll up together, while callback-only measures remain specific to the model. Handle time is kept mutually exclusive between callback and voice media types.

Actionable takeaway: Document the service-level definition before comparing callback performance, and keep AFC- and CFC-specific metrics distinct when interpreting blended reports.
6. Check the traps that quietly undo a good design
Genesys advises that agents should not leave internal ACD voicemails. Doing so creates a callback segment that can tie up the agent’s utilization until the callback is answered and ended in the queue. A callback without a scheduled date and without updates for 14 days has its conversation ended automatically by analytics, although the callback may still be actionable depending on the scenario.
There is also a point to verify in configuration rather than assume. The Architect callback documentation says a callback created from an in-queue flow takes the original call’s queue position and the same skill requirements and priority settings. Separately, the Genesys Cloud callbacks overview says the platform does not retain interaction skills and priorities when placing a callback. Organizations should prove the expected skill and priority behaviour in their own setup.
The original call from which the callback was requested is not counted as an abandoned call. When an agent places a call from a callback, however, the interaction requires a wrap-up code regardless of the queue’s After Call Work setting, including Optional. Wrap-up code design therefore matters even in queues that otherwise allow agents to skip it.
The Scheduled Callbacks view gives administrators and supervisors a place to see and cancel agent-initiated callbacks. It is a useful control point for managing scheduling discipline. Callbacks also move work in time rather than remove it, so callback volume belongs in planning conversations with the people responsible for forecasting and scheduling. Dunamis explores related choices in its article on Genesys Cloud workforce management.
Actionable takeaway: Test skill and priority behaviour, review callback ownership and scheduling controls, and include callback work in forecasting and wrap-up design.
7. Decide the sequence before changing the queue
A sound decision sequence starts with the waiting experience. Publish the default in-queue call flow and design channel-specific treatment before adding callbacks. Choose the estimated wait playback mode deliberately, because it determines what customers are told. Then choose AFC or CFC according to whether agent capacity or customer reachability is the larger constraint.
Next, set pacing and retries against real on-queue strength. Only after the operating model is clear should the team settle the service-level definition. Turning on Include Callback Virtual Wait in SLA changes how callback service level is measured, so the resulting figures are not directly comparable with the previous quarter’s figures without accounting for that change.
Dunamis Consulting Inc provides consultation and gap analysis, cost analysis, project scheduling, flexible cloud staffing solutions and bulk hour purchases for cloud telephony projects, as well as managed and support services. With 15 years in the field, Dunamis helps organizations plan and improve cloud telephony and cloud communication solutions. Contact the team to start a conversation about queue design and callback strategy.
Comments