Moving the Contact Centre to Cloud UC for Remote Clinicians

The position we inherited

The network ran its contact centre on an on-premises platform with agents tied to fixed desks. Handsets were wired back to a switch in the main data center, and every queue belonged to one physical site. Clinical work never fit that shape. Specialists consulted from satellite clinics, senior nurses triaged after hours from home, and administrative teams covered each other across buildings. Each of those arrangements required a named person at a named desk, or it did not happen at all.

Patient scheduling and triage felt the constraint first. A call arriving at a busy site could not be answered by an idle agent at another one, because the queue had no way to cross sites. Recruiting followed geography rather than skill, and the same demand spikes repeated on a weekly rhythm that everyone could see and nobody could act on.

The IT team carried the other half of the problem. The platform was approaching end of support, spares for the handset estate were getting harder to source, and each site had drifted into its own configuration. Reporting arrived as flat files exported from the switch after the fact. There was no safe way to test a routing change without touching a live queue.

What we designed

We moved the contact centre onto a cloud unified communications platform, and treated the network underneath it as part of the deliverable rather than an assumption.

Agent voice runs over WebRTC in the browser and over softphone clients on managed laptops, with mobile clients for clinicians who need to take calls between appointments. Inbound PSTN traffic terminates at a session border controller, which handles SIP normalization, number translation, and the security boundary between the carrier and the platform. Sites that still needed analog lines for lift phones and clinical record fax were given their own gateway instead of being forced through the new stack.

Routing moved from hardware queues to a policy layer. Queues are built around skills rather than sites, so a triage call can land with any qualified nurse on shift regardless of building. Overflow rules, priority treatment for clinical lines, after-hours behavior, and voicemail-to-email are configured once and applied everywhere. Call recording is scoped per queue with defined retention, and recording consent is announced to the caller before the agent connects.

Media quality was specified rather than hoped for. Opus is preferred end to end, with G.711 as the fallback on PSTN legs. Voice is marked DSCP EF and video AF41, so the underlay can protect real-time traffic when a link saturates. We tuned jitter buffers per site instead of per platform, because clinic Wi-Fi and hospital LAN behave differently and a single global setting always penalizes one of them. Jitter, packet loss, and round-trip delay are collected from RTCP reports and reviewed weekly against a defined threshold, which turns call-quality complaints into something the network team can measure.

Access follows the directory. Single sign-on with multi-factor authentication, role-based permissions for supervisors and agents, and administrative actions written to an audit log. Media is encrypted in transit with SRTP and signaling with TLS, and patient-identifying data stays inside the platform’s contracted region.

How we rolled it out

Cutover was phased by site, and the old platform stayed licensed and live until the last one moved. One team ran on the new stack first for two weeks while the rest of the network continued as before, which gave us real call volumes to tune routing against instead of a test script.

Training was built for how people actually work: short sessions for agents, a separate one for supervisors covering queue changes and reporting, and floor walkers available during the first week at each site. Mobile clients were enrolled before cutover so nobody discovered the app on the day itself.

Emergency call handling at each site was verified independently before the legacy switch was retired.

Results

Detailed metrics are shared under NDA once the customer signs off, so this section covers what changed operationally.

Remote clinicians were brought onto the platform without desks, handsets, or site-specific configuration, which means extending coverage to a new location no longer carries a procurement lead time. Supervisors gained a live view of queue depth and abandonment across the whole network rather than one building at a time. Routing changes moved into a change window with testable configuration instead of a physical switch and a hope.

The IT team’s work shifted from maintaining a diminishing estate to improving service: tuning skills groups, tightening reporting, and connecting the contact centre to the clinical workflows it was always meant to serve.