WFM guide

Contact centre technology stack

Q: What technology does a contact centre use?

A contact centre typically uses seven technology layers: (1) ACD (Automatic Call Distribution) — the routing engine that queues and routes contacts to agents based on skills, availability, and priority rules. (2) WFM platform — workforce management software for forecasting, scheduling, and real-time adherence monitoring. (3) QA and recording — interaction recording (screen + voice), quality evaluation forms, and optionally speech analytics for automated scoring. (4) CRM — the agent desktop application showing customer history, case management, and integration with back-end systems. (5) Knowledge base — agent-facing internal guides and customer-facing self-service articles. (6) IVR and AI — interactive voice response for call routing and self-service; AI chatbots for digital channel deflection. (7) Analytics/BI — operational reporting on volumes, SLAs, agent performance, and queue state.

Q: What is an ACD in a contact centre?

ACD stands for Automatic Call Distribution. It is the routing engine at the heart of a contact centre: it receives incoming contacts (calls, chats, emails, messages), queues them, and assigns them to agents based on rules. These rules typically include agent skills (language, product knowledge, complaint handling authorisation), agent availability, queue priority, and time-based overflow logic. Modern ACDs also handle omnichannel routing — applying the same queuing logic to voice, digital, and back-office work items. The ACD produces the interval-level data (calls offered, calls handled, ASA, abandoned rate) that feeds WFM forecasting.

Q: What does WFM software do in a contact centre?

WFM (Workforce Management) software handles three core functions: (1) Forecasting — using historical ACD data to predict future contact volumes at interval level (typically 15 or 30 minutes). (2) Scheduling — applying Erlang C or equivalent algorithms to the volume forecast to produce optimised agent schedules that meet service level targets at minimum cost. (3) Real-time adherence — comparing what agents should be doing (schedule) with what they are actually doing (ACD state), generating exception alerts when agents are off-schedule. Enterprise WFM platforms also include long-range capacity planning, absence management, voluntary overtime management, and performance reporting.

Q: How does AI and automation affect contact centre staffing?

AI affects contact centre staffing through deflection (fewer contacts reaching agents) and augmentation (reducing handling time for contacts that do reach agents). Well-implemented IVR deflects 20–45% of inbound calls to self-service. AI chatbots for digital channels achieve 10–50% containment depending on design quality and query complexity. AI agent-assist tools (real-time knowledge base suggestions, automated case summarisation, guided next-best-action) reduce ACW by 15–35% and AHT by 10–20%. Each of these shifts the WFM calculation: lower volume or lower AHT both reduce the headcount required to meet a service level target, but the benefits do not scale proportionally with Erlang C — occupancy may need adjustment to maintain service level with fewer agents.

A contact centre depends on at least seven distinct technology layers — and the quality of each one affects staffing requirements. AHT, FCR, deflection rate, and schedule adherence are all shaped by the tools agents use. This guide explains what each layer does and its direct WFM implications.

📞ACD

📅WFM Platform

🎙️QA and Interaction Recording

🗂️CRM

📖Knowledge Base

🤖IVR and AI Chatbot

📊Analytics and Reporting

📞

ACD — Automatic Call Distribution

Routes and queues contacts to the right agent at the right time

Key capabilities

→Inbound call queuing with hold music and position announcements
→Skill-based routing (language, product, authorisation level)
→Priority queuing (VIP, complaint, vulnerable customer flags)
→Overflow rules (route to alternative skill group or voicemail after N seconds)
→Omnichannel blending: voice, chat, email, SMS routed through the same engine

WFM impact

Produces the interval-level volume data that feeds forecasting. Skill groups, overflow rules, and priority settings directly shape the queue patterns WFM must plan for.

Staffing note

Skill group design is a WFM decision. Narrow skills (too many distinct queues) produce small, volatile queues that are difficult to staff efficiently. Broad pools (high multi-skill) reduce queue fragmentation but increase training cost.

📅

WFM Platform

Forecast demand, schedule agents, and track real-time adherence

Key capabilities

→Historical volume import from ACD for interval-level forecasting
→Erlang C or proprietary algorithm for schedule generation
→Shift pattern management, break scheduling, and leave management
→Real-time adherence: compares planned schedule with live ACD agent state
→Exception management: automated alerts when agents go off-schedule
→Long-range capacity planning and headcount modelling

WFM impact

The operational core of workforce management. The forecast accuracy of the WFM platform directly determines schedule quality and therefore staffing efficiency.

Staffing note

The WFM platform is only as good as the data it receives. ACD integration quality, correctness of shrinkage inputs, and discipline in maintaining agent skill records are the three most common failure points.

🎙️

QA and Interaction Recording

Record, evaluate, and improve the quality of agent interactions

Key capabilities

→100% call and screen recording for compliance and coaching
→Structured evaluation forms with weighted scoring dimensions
→Speech analytics: automated call categorisation, keyword detection, silence and talk-over analysis
→Sentiment analysis: customer tone and escalation pattern detection
→Calibration workflow: multi-assessor scoring for inter-rater reliability tracking

WFM impact

QA scores affect adherence interpretation (legitimate ACW vs. avoidance), ramp trajectory measurement, attrition prediction, and training investment decisions.

Staffing note

Speech analytics reduces the sampling requirement for QA (automated scoring supplements manual evaluation) and surfaces systemic issues — high ACW patterns, repeat caller detection, compliance gap identification — faster than manual sampling alone.

🗂️

CRM — Customer Relationship Management

Agent desktop showing customer history, case management, and back-end integration

Key capabilities

→Screen pop: customer record automatically presented when call connects
→Case creation and routing: automatic ticket for back-office follow-up
→Order, account, and policy data surfaced on a single agent screen
→Integration with back-end systems (billing, fulfilment, policy admin) via API
→After-call work guidance: structured disposition codes and follow-up actions

WFM impact

CRM performance directly affects AHT. Every second of screen load, duplicate data entry, or manual lookup in a fragmented CRM adds to AHT — and therefore to headcount requirements under Erlang C.

Staffing note

AHT savings from a well-integrated CRM are significant. Operations moving from fragmented multi-screen environments to a unified agent desktop commonly report 30–60 second AHT reductions — equivalent to 8–15% fewer agents needed for the same volume.

📖

Knowledge Base

Give agents accurate, current answers quickly; give customers self-service answers

Key capabilities

→Agent-facing internal knowledge base with search and guided pathways
→Customer-facing help centre and FAQ (the self-service channel)
→Version control and review workflow: ensure articles are current
→Smart search and agent-assist: article suggestions during live calls (AI-enhanced in some platforms)
→Analytics: search terms that return no results signal knowledge gaps

WFM impact

Internal knowledge base quality reduces AHT (agent finds the answer faster), reduces ACW (less post-call research), and reduces repeat contacts (complete, accurate answers at first contact).

Staffing note

Knowledge base maintenance is an often-neglected WFM cost. Outdated articles increase AHT (agents verify information via alternative routes), increase error rates, and drive repeat contacts. Content ownership and review cadence should be part of the WFM planning cycle.

🤖

IVR and AI Chatbot

Deflect routine contacts before they reach an agent

Key capabilities

→IVR (Interactive Voice Response): DTMF menu navigation and basic self-service (balance enquiry, order status, appointment booking)
→Conversational IVR: natural language voice recognition for intent detection and self-service routing
→Digital chatbot: scripted or LLM-powered for web/app/WhatsApp channels
→Call authentication: pre-agent identity verification reduces agent AHT
→Proactive outbound: appointment reminders, delivery notifications to reduce inbound WISMO volume

WFM impact

Deflection reduces contact volume, which reduces agent headcount. But deflection is non-linear under Erlang C — a 20% deflection does not produce a 20% headcount reduction because of occupancy effects.

Staffing note

IVR containment rates: 20–45% for well-designed transactional IVRs; 25–55% for conversational IVR. LLM chatbots for digital channels: 25–50% containment in pilot deployments. Each percentage point of sustained deflection reduces agent headcount requirement by roughly the same percentage — but re-run Erlang C to get the precise number, as occupancy changes too.

📊

Analytics and Reporting

Operational and strategic reporting on volumes, SLAs, agent performance, and trends

Key capabilities

→Real-time wallboard: live queue state, SLA status, agents available
→Interval-level reporting: 15/30-min SL, ASA, abandonment, occupancy
→Agent-level: AHT, FCR, adherence, quality score by interval and day
→Forecast accuracy reporting: WAPE or MAE against actuals for continuous calibration
→Trend and seasonality analysis: week-on-week, year-on-year volume comparison

WFM impact

Analytics is how WFM validates forecast accuracy, identifies performance patterns, and builds the business case for headcount decisions.

Staffing note

The most valuable analytics for WFM are interval-level (not daily) and skill-group-level (not operation-level). Aggregated daily reporting masks intraday peaks and multi-skill routing patterns that determine actual staffing requirements.

Why integration quality matters more than individual tools

The most common technology failure in contact centres is not bad individual tools — it is poor integration between good tools. When ACD, WFM, CRM, and QA operate in silos, the costs compound at the agent level.

ACD–WFM data gap

Forecasting runs on manual volume exports rather than live feed. Forecast is always 24–48 hours stale. Intraday changes are missed until the queue is already in crisis.

WFM–QA disconnect

Adherence data does not include QA context. A 10-minute ACW spike looks the same whether it is thorough post-call work or avoidance — the data cannot distinguish them without linking to the QA record.

CRM fragmentation

Agents use 3–5 systems during a contact. Each system load adds 10–30 seconds to AHT. The AHT the Erlang model is told does not match actual AHT because nobody has accounted for the inter-system navigation time.

Contact centre technology questions

What technology does a contact centre use?

Seven core technology layers: ACD (routing engine), WFM platform (forecasting, scheduling, real-time adherence), QA and interaction recording (evaluation, speech analytics), CRM (agent desktop, customer data), knowledge base (internal guides and customer self-service), IVR and AI chatbot (deflection), and analytics/BI (operational reporting). Modern contact centres also integrate identity verification, workforce optimisation, and AI agent-assist layers on top of these.

What is an ACD in a contact centre?

ACD (Automatic Call Distribution) is the routing engine that receives incoming contacts, queues them, and assigns them to agents based on skill, availability, and priority rules. It is the source of interval-level volume data (calls offered, calls handled, ASA, abandoned rate) that WFM uses for forecasting. Skill group design and overflow logic in the ACD are WFM decisions — they directly shape the queue patterns that must be staffed.

What does WFM software do in a contact centre?

WFM software handles three functions: (1) Forecasting — using ACD historical data to predict future contact volumes at 15/30-minute interval level. (2) Scheduling — generating optimised agent schedules using Erlang C or equivalent algorithms to meet service level targets. (3) Real-time adherence — comparing the planned schedule with the live ACD agent state, generating exception alerts when agents deviate. Enterprise platforms also handle absence management, long-range capacity planning, and overtime management.

How does AI and automation affect contact centre staffing?

AI reduces contact volume (deflection via IVR/chatbot: 10–50% of contacts, depending on channel and query complexity) and reduces handling time (agent-assist tools: 10–20% AHT reduction, 15–35% ACW reduction). Each reduction shifts the Erlang C calculation — but occupancy changes non-linearly, so re-run the model rather than applying a simple percentage reduction to headcount.