Deployment blueprint

A Contact Centre platform for a BPO that answers for a dozen brands.

Agents at home and in one hub, several client programs with their own numbers, greetings and service targets, supervisors who need to hear and coach in real time. This blueprint shows how 3CTel designs queues, callbacks, quality sampling and AI deflection for an outsourced contact centre.

  • One queue set per client program
  • Whisper, barge and wallboards for supervisors
  • AI agents for first-line FAQ and after-hours

For an outsourced contact centre, 3CTel configures a separate number, greeting and skills-based queue set for each client program, routes calls to agents by skill and language, offers callbacks when wait times climb, and gives supervisors wallboards, whisper and barge. Calls are recorded with an announcement and sampled for quality review. AI agents answer first-line FAQ and after-hours calls, and the CRM or ticketing tool is connected via API, scoped during planning.

Deployment blueprint: a representative, anonymized scenario compiled from how 3CTel implements this kind of organization. No client is named and no result figures are stated; every parameter is a configuration choice we design with you.

Sector
Business process outsourcing / contact centre
Users
140 agents, 12 supervisors, 6 QA and admin
Sites
1 hub office plus agents working from home
Client programs
12, each with its own numbers and queues
Languages
English and French, with per-queue skills
Modules
Contact Centre, business phone, SMS, call analytics
AI agents
First-line FAQ, after-hours intake, callback scheduling
Integrations
Client CRM and ticketing via API and automation platform

Situation

An outsourced contact centre answers on behalf of other companies. Each client program has its own phone numbers, its own greeting, its own hours and its own definition of a good call. Volume arrives in waves that follow the clients' marketing calendars rather than the BPO's own, and a single agent may be trained on three programs.

Most agents work from home on a softphone. A hub office holds supervisors, QA staff and a smaller floor of agents. Clients expect reporting per program, recording for disputes and quality review, and evidence that their callers were told the call may be recorded. Client data must stay separated program by program.

Requirements

  • A distinct number, greeting, hours and queue set for each of twelve client programs, with agents assigned by skill and language.
  • Callbacks offered automatically when estimated wait exceeds a threshold set per program.
  • Wallboards per program and per team, with supervisor whisper, barge and silent monitoring.
  • Recording with a caller announcement, and a sampling workflow so QA reviews a fixed share of calls per agent.
  • Agent softphones on home computers with headsets, plus the mobile app as a fallback.
  • Screen-pop and call logging into each client's CRM or ticketing tool, without mixing data between programs.
  • AI agents that handle first-line FAQ and after-hours intake for programs that opt in.

Call-flow design

Every program number lands on its own greeting and menu. The menu maps to skills rather than to named agents, so staffing changes do not require routing changes. Language is captured once and carried as a skill requirement into the queue.

Conceptual illustration of the 3CTel interface — not a screenshot

WhenThen
A call arrives on a program number during that program's hoursProgram greeting plays, recording notice is announced, and the caller picks a reason or language
The caller chooses FrenchThe call queues for agents with the French skill on that program; other agents are not offered the call
Estimated wait exceeds the program's thresholdThe caller is offered a callback that holds their place; the queue dials them when an agent is next
No agent on the program is logged inThe call overflows to a cross-trained backup team, then to the AI agent if the program has enabled it
The program is closedThe after-hours AI agent takes the call, or voicemail with transcript is delivered to the program's inbox
A supervisor flags an agent for coachingWhisper is used on live calls; barge is reserved for escalations and logged

AI agents

AI agents are enabled program by program, only where the client has approved the script and the escalation rules. Each AI agent is trained on that program's FAQ and nothing else, so an answer from one brand never leaks into another.

  • The AI never quotes prices, makes commitments on a client's behalf or discusses account details it cannot verify.
  • Any caller who asks for a person, expresses distress or raises a complaint is transferred immediately, with the transcript attached.
  • Every AI conversation is logged to the program's record so QA can review it like any other call.

First-line FAQ

Answers order-status, hours, return-policy and 'where do I send' questions from the program's approved knowledge. Anything outside the list is transferred to the queue with a summary.

After-hours intake

Takes the caller's name, reason and callback window, sends a transcript to the program inbox and, where the client allows, creates a ticket via API.

Callback scheduling

When a caller declines to wait, the AI confirms a callback slot within the program's hours and books it into the queue.

Bilingual

Answers in English or French based on the caller's choice or first words, and carries the language into the transfer.

Supervision and quality

Supervisors watch wallboards per program: calls waiting, longest wait, agents available and service level for the interval. Whisper lets a supervisor coach an agent without the caller hearing; barge is used only when a call needs to be taken over, and each use is recorded in the call log.

QA sampling is configured as a percentage of calls per agent per week, drawn at random across programs. Reviewers score from a form agreed with each client, and scores are reported by program. Recordings are searchable by program, agent, number and date.

Integrations

Each program's CRM or ticketing tool is connected separately via API, webhooks or an automation platform, scoped during planning and confirmed before activation. When a call is answered, the agent's screen shows the matching contact in that program's tool; when the call ends, the recording link, disposition and duration are written back.

Programs that share a tool are separated by pipeline or workspace so agents only see the records of the brand they are answering for.

Security, privacy and continuity

  • Canadian law generally requires that callers be informed that a call is recorded; 3CTel configures the announcement per program with the BPO and its client.
  • Recording retention is set per program with the client, and access to recordings is limited by role: agents cannot open recordings, QA and supervisors can.
  • Agent logins use the desktop app with per-user credentials; agents are removed from queues the day they leave.
  • Emergency calling is configured with a registered service address for each home-based agent, confirmed at setup and updated when they move.
  • If the hub office loses connectivity, home agents continue unaffected; if a program's queue is unreachable, calls forward to a number the client designates.
  • Outbound calling limits and international restrictions are set per user to reduce fraud exposure.

Rollout

  1. Program inventory

    List every client program, its numbers, hours, languages, service targets and the CRM it uses. This becomes the routing plan.

  2. Build and test on temporary numbers

    Each program's greeting, menu, queues and callbacks are configured and called through before anything is ported.

  3. Pilot with one program

    The first program goes live on the new platform with its agents, supervisors and QA sampling running for a week.

  4. Port program numbers in waves

    Remaining programs port in groups agreed with each client, with dates confirmed before any switch.

  5. Enable AI agents per program

    AI FAQ and after-hours intake are switched on only after the client approves the script and escalation rules.

What we measure

Per program and per interval: calls offered, answered and abandoned; service level; average and longest wait; callbacks requested and completed; AI conversations handled, transferred and escalated; QA sample coverage and score distribution; agent availability and occupancy. Reports are exported per client.

How does an outsourced contact centre keep client programs separated on one phone platform?

Each program gets its own numbers, greeting, queues, recording retention and CRM connection. Agents are assigned skills per program, so a call for one brand is never offered to someone untrained on it. Reports and recordings are filtered by program, and 3CTel scopes the separation during planning so a client only ever sees its own data.

Can agents working from home use the same queues as the hub office?

Yes. Agents log in from a desktop softphone with a headset, or from the mobile app as a fallback, and join the same skills-based queues as staff in the hub. Supervisors see them on the same wallboards and can whisper or barge in the same way. Emergency calling uses a registered address for each home location.

Where do AI agents fit in a BPO that sells human service?

They take the calls no client wants a human paid to answer: order status, hours, return policy and after-hours intake. Each AI agent is limited to its program's approved knowledge, transfers on request or distress, and hands the transcript to the queue. Clients opt in program by program, and the AI never commits on their behalf.

Frequently asked questions.

The routing plan is built per program, and there is no practical limit that a BPO of this size would reach. Each program adds numbers, a greeting, queues and skills; agents are assigned to whichever programs they are trained on.

Bring us your program list.

Numbers, hours, languages, service targets and the tools each client uses are enough to design the queues, price the seats and plan the port.

Call 3CTelBuild My System