WFM guide

WFM for blended agents

Q: What is a blended agent in a contact centre?

A blended agent handles contacts from more than one channel simultaneously or in sequence during a single shift. The most common blending model is voice plus chat: an agent handles one live voice call at a time, but can simultaneously manage 2–3 live chat conversations in the gaps between voice contact interactions (hold time, customer reading, etc.). Blending also covers voice plus email (where email is worked between calls) and inbound plus outbound (where agents handle inbound calls during peak hours and work outbound tasks during quieter intervals).

Q: How does blending affect Erlang C and headcount calculations?

Erlang C was designed for single-channel, single-contact queues — one call per agent at a time. For blended agents, two changes are needed: (1) Voice staffing: calculate voice agents using Erlang C exactly as for a dedicated voice team. (2) Chat capacity uplift: blended agents handling simultaneous chats can handle more contacts than their headcount implies. An agent managing 2 concurrent chats provides 2× the chat throughput of a dedicated single-chat agent. The WFM model must account for the chat capacity available from the voice team during their voice contact gaps. Occupancy must be modelled separately for voice and for the blended digital load — combined occupancy above 85% risks quality degradation on both channels.

Q: How many chats can an agent handle at once?

The standard concurrency benchmark for chat-only agents is 2–3 simultaneous conversations. For blended agents (voice + chat), the practical concurrent chat capacity is lower: 1–2 simultaneous chats when also handling voice, because voice demands continuous attention and interrupts chat response cadence. In practice: if an agent handles 6 voice contacts per hour (each 10 minutes average), the interstitial chat capacity is roughly 20–40% of a dedicated chat agent's throughput — more during low-voice periods, much less during peaks. Blending works best when voice and digital peaks are offset, not coincident.

Q: What are the risks of blending voice and chat agents?

Three primary risks: (1) Attention splitting — agents managing a voice call and 2 live chats during a complex query will produce lower quality on all three. Quality scores for blended agents typically run 5–12 points below dedicated-channel agents for the same type of work during high-concurrency periods. (2) Occupancy inflation — the combined load of voice plus chat contacts can push effective occupancy to 90%+ without the ACD flagging it, because the ACD only sees the agent as 'in a call' or 'available'. The digital load is invisible to the voice occupancy calculation. (3) Service level misattribution — if chat response times degrade while voice SL is green, the root cause (agent attention split by voice) is not visible in standard reporting.

When agents handle voice and chat simultaneously, the standard Erlang C model breaks down. A blended operation requires a separate staffing approach for each channel, a concurrency uplift model for digital capacity, and explicit occupancy management across the combined load — or you will consistently misstate your effective headcount.

Dedicated vs. blended: what changes in the model

Dedicated channel teams

→ Voice team: staffed with Erlang C on voice volume + AHT
→ Chat team: staffed with Little's Law / concurrency model on chat volume
→ Occupancy calculated per-channel independently
→ SL targets set and tracked per-channel independently
→ Simple to model; requires more total headcount (teams cannot flex between channels)

Blended agents (voice + chat)

→ Voice staffing: still Erlang C on voice volume (voice is always single-threaded)
→ Chat capacity: derived from blended agent interstitial availability
→ Occupancy: must sum voice occupancy + digital load — ACD only sees voice
→ SL targets: must be monitored separately or channel blending masks problems
→ Potentially lower total headcount; higher complexity; higher quality risk

The efficiency case for blending is real — but only when voice and chat peaks are offset. When both peak simultaneously, blended agents are forced to choose which channel to deprioritise.

Concurrency uplift: how much chat capacity does the voice team provide?

The interstitial capacity available to a blended voice agent for chat depends on how much of their time is spent in active voice conversation vs. hold, waiting, and ACW states.

Worked example: 10 blended agents, voice + chat

72%

Voice occupancy

Agents in voice-related activity 72% of the hour

28%

Interstitial availability

Available for chat during voice gaps

~2.8 chat FTE

Max chat throughput

10 agents × 28% = equivalent of 2.8 dedicated chat agents

Caveat: interstitial availability is not evenly distributed. During peak voice intervals, availability may drop to 10–15%; during quiet intervals, it may rise to 50%+. Chat capacity from blended agents therefore peaks and troughs inversely to voice. If your chat volume peaks when voice peaks, the model breaks — you need dedicated chat agents for peak coincidence.

Chat concurrency by voice occupancy level

Voice occupancyInterstitial %Max concurrent chats (per 10 agents)Recommendation

60%40%%~3–4 FTE equivalentGood blending conditions; peak offset maintained

72%28%%~2.5–3 FTE equivalentWorkable; monitor chat response time closely

80%20%%~1.5–2 FTE equivalentMarginal; quality risk on both channels

88%+<12%%<1 FTE equivalentDo not blend at this occupancy; add dedicated chat resource

Blended occupancy: the invisible overload problem

The ACD measures voice occupancy only. An agent handling two live chats in parallel with a voice call appears as “in a call” in the ACD — the digital load is invisible to the occupancy calculation. This creates a systematic blind spot.

What the ACD reports

→ Voice occupancy: 75% ✓ (acceptable)
→ All agents: "in a call" or "available"
→ No alert triggered
→ No escalation

Actual agent load

→ Voice: 75% occupancy (ACD visible)
→ Chat: 30% additional load (ACD invisible)
→ Combined occupancy: ~105% of comfortable capacity
→ Agent is overloaded; quality and response time degrade

How to detect blended occupancy overload

Chat response time: track median response time per chat interaction within a session — if it degrades during voice peaks, agents are attention-split
ACW spikes: agents using post-call work time to catch up on chat sessions they deferred during voice contact
QA scores by hour: if blended quality drops during voice peak intervals, the concurrent load is too high
Agent-reported capacity: pulse check with blended agents — do they feel they can handle both channels adequately during peak?

Inbound–outbound blending

Inbound–outbound blending is a different problem from voice–chat blending. Here agents handle inbound calls during peaks and work outbound tasks during quiet periods. The WFM challenge is switching accurately between modes.

Outbound tasks are interruptible; inbound calls are not

An agent can pause an outbound dialling task when a queue builds. A live inbound call cannot be interrupted. The ACD must be set to automatically suppress outbound dialling when the inbound queue reaches a threshold — typically when agents available drops below the Erlang minimum.

Outbound productivity is non-linear with dialling time

An agent who works outbound for 15 minutes per hour produces disproportionately less than one who works outbound for 45 minutes per hour, because the number of contacts reached scales with connected-call opportunity. Short windows are inefficient for predictive dialling models.

Inbound–outbound switching has a productivity cost

Each mode switch (inbound to outbound, outbound to inbound) has a context cost of approximately 30–60 seconds. In high-frequency switching environments, this adds materially to ACW and reduces effective talk time across both modes.

Service level obligations differ by mode

Inbound calls have a hard SL commitment (answer within X seconds). Outbound contacts do not — they are target-driven (reach N contacts per hour). The WFM model must protect inbound SL first and treat outbound as the residual absorber.

When blending helps — and when it doesn’t

Blending works well when...

→ Voice and digital peaks are significantly offset (voice peaks morning, chat peaks afternoon)
→ Chat conversations are asynchronous enough to tolerate 2–3 minute response gaps
→ Contact complexity is moderate — both channels are transactional, not advisory
→ Agent capacity is genuinely underutilised on voice during quiet periods
→ Training allows agents to maintain quality across both channels simultaneously

Blending fails when...

→ Voice and chat peaks coincide — agents must choose which channel to fail
→ Contacts on either channel are complex or emotionally demanding (complaints, vulnerable customers)
→ Chat SLA requires near-real-time response (under 30 seconds) — incompatible with being on a call
→ Voice occupancy is already above 80% — no interstitial capacity available for chat
→ QA scores for blended agents fall significantly below dedicated channel scores

Blended agent WFM questions

What is a blended agent in a contact centre?

A blended agent handles contacts from more than one channel simultaneously or in sequence. The most common model is voice plus chat: agents handle one voice call at a time while managing 1–2 live chats in the gaps. Blending also covers voice plus email (email worked between calls) and inbound plus outbound (outbound dialling during quiet inbound periods).

How does blending affect Erlang C and headcount calculations?

Erlang C works correctly for the voice component of a blended team — voice is always single-threaded. The chat component requires a separate concurrency model. The key calculation is interstitial availability: if voice occupancy is 72%, agents have 28% availability for chat. 10 agents at 28% interstitial availability ≈ 2.8 FTE equivalent of chat capacity. Occupancy must be monitored across both channels — the ACD only sees voice.

How many chats can an agent handle at once?

Chat-only agents: 2–3 simultaneous conversations. Blended agents (voice + chat): 1–2 simultaneous chats, declining to near-zero during high-voice-occupancy intervals. The practical concurrent capacity depends on voice occupancy — at 80%+ voice occupancy, there is insufficient interstitial time for quality chat handling.

What are the risks of blending voice and chat agents?

Three risks: (1) Attention splitting — quality degrades on both channels when agents are managing voice and multiple chats simultaneously; QA scores typically run 5–12 points lower. (2) Occupancy inflation — the ACD only sees voice occupancy; combined load may reach 100%+ without the system flagging it. (3) Service level misattribution — chat response time degradation during voice peaks is not visible in standard ACD reporting.