Agent-assist tools, better knowledge bases, CRM clean-ups and automated after-call summaries are usually sold on one promise: lower average handle time. The business case then converts that into “FTEs freed”. A typical slide says that a 12% AHT reduction on a 500-agent process frees 60 agents. The arithmetic is right, and as a rough estimate it is not unreasonable. But the path from a lower AHT to a smaller payroll has several turns where the saving shrinks, grows or simply goes somewhere else.

The linear estimate and why it is a fair starting point

Workload is calls × AHT. If calls stay the same and AHT falls 12%, workload falls 12%. On a large process, required staff tracks workload fairly closely, so a 12% reduction in agents is a sensible first approximation.

It is worth checking with the queueing model that voice processes actually use. Using Erlang C for a single half-hour interval with 400 calls, a 360-second AHT and an 80/20 service level target:

AHTAgents needed for 80/20Reduction
360 s88–
342 s (5% lower)844.5%
317 s (12% lower)7811.4%

For a team this size, the result is close to linear. The model needs 10 fewer agents; the linear estimate says 10.6.

Small teams move in steps

Repeat the exercise for a small queue of 40 calls per half-hour. At 360 seconds, the model needs 12 agents. At 342 seconds it needs 11; at 317 seconds, 10. In percentage terms that is 8% and 17%, very different from the 5% and 12% AHT changes. People come in whole numbers, and in small teams each person is a large share of capacity, so savings arrive in lumps. Sometimes a 5% AHT reduction frees nobody in a particular interval; sometimes it frees one person out of nine.

Small teams also run at lower occupancy for the same service level. The 12-agent team above is only about 67% occupied, while the 88-agent team is above 90%. That is the usual pooling effect, and it means the freed time in small teams often turns into idle time rather than removable headcount.

If the process is understaffed, service level comes first

Many AHT projects start because a process is missing its service level. Then the freed capacity is not “freed” at all; it is absorbed by the gap.

Suppose the 400-call interval is actually staffed with 84 agents instead of the 88 it needs. At 360 seconds AHT, Erlang C predicts only about 55% of calls answered within 20 seconds. Cut AHT by 5% to 342 seconds, keep the same 84 agents, and predicted service level jumps to about 83%. Near full occupancy, service level is extremely sensitive to small changes in workload, so a modest AHT gain can rescue a failing SLA. In this situation the honest business case says “SLA restored”, not “FTEs released”.

Where the saved seconds go

Before converting AHT into headcount, check which part of handle time fell.

  • After-call work is the most reliable target. Automated call summaries and auto-filled disposition fields remove typing time without affecting the customer conversation.
  • Hold time falls when agents find answers faster through suggested knowledge articles. This also improves customer experience.
  • Talk time reductions deserve scrutiny. If agents are cutting conversations short to meet a new AHT expectation, repeat calls rise and the workload returns through the front door.

The real test is total workload over several weeks: calls × AHT, including repeat contacts. If AHT falls 12% but repeat calls rise 5%, the net workload saving is much smaller. A pilot that measures only AHT for the pilot group can overstate the benefit.

Pilots also tend to be run with motivated agents, extra supervision and a fresh tool. Gains often shrink when the tool is rolled out to the whole floor. It is prudent to plan on a fraction of the pilot result until the full rollout data is in.

From capacity to cost

Even when the capacity is genuinely freed, it does not appear as a saving until the roster changes. A 10-agent reduction in a peak interval does not mean 10 agents can go home all day; the saving is spread across intervals, and turning it into fewer shifts requires rebuilding schedules. In practice, headcount usually falls through attrition: the process simply hires fewer replacements over the next few months.

Who benefits depends on the contract. Under FTE-based pricing, the client needs fewer billable seats, so the client captures the saving and the BPO’s revenue falls. Under per-transaction pricing, the BPO keeps the saving as margin, at least until the next rate negotiation. This is one reason why automation projects can be politically awkward in outsourcing relationships, and why many contracts now include gain-sharing clauses that split productivity benefits.

The freed capacity can also be redeployed rather than removed: to absorb volume growth without hiring, to add time for coaching that had been squeezed out, or to take on outbound retention calls. Each is a legitimate use. The point is to decide deliberately, because unclaimed capacity has a habit of turning into lower occupancy and longer breaks without anyone choosing that outcome.

A better way to state the benefit

Instead of a single “FTEs freed” number, a credible case usually states three things: the workload reduction measured over a meaningful period including repeat contacts; what that does to service level at current staffing, using the actual queueing model; and the plan for the capacity, with dates and the mechanism (attrition, redeployment or a contract change). That is less dramatic than a headline figure, and far more likely to match what finance sees six months later.

Written by the RP Strativa Team. Figures in the examples are illustrative; service level results are from a standard Erlang C calculation for a single interval.

Leave a Reply

Trending