← work

ServiceNow · Omnichannel voice

Callback

From waiting on hold to a call that comes back. The technology existed, but the experience didn't. I designed how customers ask for a callback and how agents act on it, then extended the same model to new scenarios.

role

Product designer, design owner

team

PM, UX research, engineering

timeline

Late 2024 – 2026

status

Shipped in Zurich and Australia

customer asks from

Phone menu (IVR)PortalVirtual AgentMessenger

callback type

Immediatekeeps their place in the queue Scheduledat a time the customer picks

agent

Callback context cardCall · schedule · cancel

outcome

ConnectedNo answer → retryTransferredRescheduled

The callback system I designed the experience for, from request to resolution.

01

the problem

Customers were stuck on hold when queues were long, and the callback experience for agents was largely undefined.

02

what I did

Took over the flows from a previous designer and turned a technical framework into end-to-end workflows for customers and agents.

03

the result

Shipped in Zurich and Australia, then reworked for a Q1 2026 scope change through stories, testing and release enablement.

context

why customers ask to be called back

When demand spikes or no one is free, customers are asked to wait on hold. Our research linked long waits to frustration and to customers leaving, and showed callback matters most during spikes like outages, recalls and travel disruption.

Needs also differed by business. Consumer brands leaned toward immediate callback, while B2B and complex products leaned toward scheduling a time.

4

customer entry points designed for

2

callback types: immediate and scheduled

2

releases shipped, then a Q1 2026 rework

the question

How might we let customers leave the queue without losing their place, and give agents everything they need to call back with confidence?

what I inherited

a framework, not an experience

When I took over from a previous designer in late 2024, the product team had defined the technology: immediate and scheduled callback, routing through the contact-center platform, and whether the agent or the customer is dialed first. Much of what agents and customers would actually experience was still an open question.

already defined
  • Immediate vs scheduled callback
  • Routing through the contact-center platform
  • Agent-first and customer-first connection models
still open
  • What context does an agent need before calling back?
  • What happens when the customer doesn't answer?
  • How should retry and transfer work?
  • How do agents get back to the callback, or remember a later one?
  • What lives on the context card vs the record?

my role

I was the design owner for callback, partnering with a PM, UX researchers and engineering across several releases. I designed the customer entry points, the agent's callback actions, the context card, and the integration of callback into the native and global voice controls. I also built and presented a callback prototype at ServiceNow's annual company-wide planning event, which received positive feedback for its clarity and helped align stakeholders on next steps.

how it grew

one model, extended release by release

  • Zurich

    Leave the queue, keep your place

    Immediate callback from the phone menu, with callback context, actions, transfer and retry for agents.

  • Australia

    Call at a better time

    Scheduled and agent-initiated callback, callback-type selection, rescheduling and cancelling, and flows for guest customers.

  • Q1 2026

    Adapt to a scope change

    When agent-scheduled callback became native-only, I reworked the actions, customer callback lists, and callbacks that start from a case.

  • Australia · voicemail

    Hear the message, then call back

    I designed the voicemail playback experience: play, pause, seek and the player's other states, so agents can listen before returning the call. It shipped in Australia, and the product later began modeling voicemail as a type of callback task, reusing the same context and follow-up patterns.

key decisions

and what each one traded off

decision 01

Show callback context only when it matters

considered

The card under the voice controls

Kept callback details next to the call, but took a lot of space from the customer section agents rely on.

chosen

A context card at the top, on callback calls only

The first thing the agent sees on a callback, including the scheduled callback, and gone on calls where it doesn't apply.

why

Agents need enough to call back with confidence: who, why and when. At the top, that answer comes first, and the customer section keeps its space. Showing it only on callback calls keeps every other call clean. I also worked through what belonged on the card versus the underlying record, including the number, time, reason, source and related case.

trade-off

Some details stay one click away on the record. Technical limits, like reliably deriving a customer's time zone, also shaped what the card could show.

Callback card placed under the voice controls, compared with the chosen card at the top of the page
decision 02

Know who to call, even when the customer is a guest

the problem

A signed-in customer comes with a profile. A guest doesn't, so the agent had no reliable way to know who they'd be calling, or whether the number was really theirs.

the design

The chat asks the guest for their number and sends a verification code before the callback is set up. The agent gets a verified number, and the customer doesn't need an account to be called back.

Guest callback in chat: share a number, verify a one-time code, callback confirmed

solution

the shipped experience

Customer portal chat offering Call me now or Call me later

01Request a callback

When the wait is long, customers can ask to be called back now, keeping their place in the queue, or pick a time that suits them.

Agent view of a callback with callback actions shown close up

02Callback context and actions

The agent sees who they're calling and why, then calls, schedules or cancels from the same place.

Global phone controls showing callback actions while the agent is on another page

03Callback in the voice controls

Callback lives in the native and global voice controls agents already use on every call.

edge cases

where a callback can go wrong

  • no answer

    An alert tells the agent the customer didn't pick up. They wrap up with a "didn't pick up" outcome and try again later.

  • needs another agent

    The agent can transfer the callback, so the customer isn't sent back to the start of the queue.

  • guest customer

    The number is verified with a code before the callback is created, so agents never dial an unconfirmed number.

  • starts from a case

    Reworked in Q1 2026, so callbacks created from a case carry that context to the agent.

impact

Callback moved from a technical framework to a shipped experience in Zurich and Australia, then through a Q1 2026 scope change to accepted stories, end-to-end testing and release enablement.

The model held up as the product grew. New entry points, scheduling, guest customers and later voicemail all fit the same context card and follow-up patterns instead of needing new ones. Usage tracking was planned from the start, with daily callback requests measured by source and type.

reflection

what worked

Designing a model, not a feature

Treating callback as one system meant each new scenario extended what agents already knew instead of adding something new to learn.

what I'd change

One model from the start

Voicemail and callback converged later. I'd design them as one model from day one, so shared patterns didn't have to be retrofitted.