How to calculate call center cost: human agents or AI?

What goes into the cost of a call center?

The most common mistake is to look only at agent salaries. The true cost of keeping a phone line staffed includes management, software, telephony, workspace and the constant cost of hiring and training. Fill in the items below from your own accounts on a monthly basis.

There is also a hidden line: calls that ring out or are abandoned in the queue. They never appear on an invoice, but they come back as lost sales, cancelled appointments and unhappy customers. Add them to the calculation as a separate line.

• Staff: gross pay, employer contributions, benefits, overtime and shift premiums
• Management and quality: team-lead time, quality checks and reporting
• Telephony: numbers, minutes and phone-system charges
• Software: phone system, CRM, recording and reporting tools
• Workspace and equipment: desks, computers, headsets, power and rent share
• Hiring and training: time to replace and onboard agents who leave
• Lost revenue: the value of unanswered and abandoned calls

How do you calculate cost per agent and cost per call?

Start with the fully loaded monthly cost of one agent: pay and employer contributions plus that agent’s share of management, software and workspace. Then work out how many calls that agent can realistically handle in a month.

No agent spends the whole shift on the phone. After breaks, leave, training, meetings and sickness, what remains is productive time. Because calls arrive unevenly through the day, not all productive time is spent talking either; take that occupancy rate from your own phone-system reports.

• Monthly agent cost = gross pay + employer contributions + benefits + (management + software + workspace ÷ number of agents)
• Productive minutes = monthly paid minutes × (1 − shrinkage)
• Handling time per call = average talk time + after-call work (notes, data entry)
• Calls handled per month = productive minutes × occupancy ÷ handling time per call
• Cost per call (human) = monthly agent cost ÷ calls handled per month

How many agents do you need?

To size the team, divide the total workload by the productive time one agent can give. That gives you an average, but averages do not set headcount; peak hours do. If calls pile up at opening time or over lunch and you are short-staffed then, callers wait or hang up.

Opening hours drive headcount directly. Once you want evening, weekend or overnight cover, one agent’s weekly hours cannot stretch to cover them and you need shifts. Running the numbers separately for business hours and out-of-hours shows exactly where the gaps are.

• Monthly workload (minutes) = monthly calls × handling time per call
• Agents needed = monthly workload ÷ (productive minutes per agent × occupancy)
• Minimum people for coverage = weekly open hours ÷ one agent’s weekly working hours
• Peak check: calls in the busiest hour × handling time per call ÷ 60 = agents needed at the same time in that hour

How do you calculate the cost of an AI voice agent?

The line items differ, but the logic is the same. Add the provider’s plan or usage fee to the cost of the phone line and call minutes. Depending on the provider, the line may be your own or one from a licensed carrier, and it is often billed separately.

Two internal costs are easy to miss. The first is the time your team spends keeping the knowledge base current and reviewing transcripts. The second is the cost of the people who take the calls the agent transfers; the agent does not make those calls disappear, it hands them over prepared and summarised.

• Monthly AI cost = plan or usage fee + line and minute costs + monthly share of setup and integration
• Internal upkeep = monthly hours on knowledge base and transcript review × internal hourly cost
• Calls resolved by AI = total calls − transferred calls
• Cost per call (AI) = (monthly AI cost + internal upkeep) ÷ calls resolved by AI

How do you put a price on missed calls?

Missed calls are often the biggest cost of all, yet they never show up on a bill. Pull a month of unanswered and abandoned calls from your phone-system logs. Then use your own records to find how many were later reached, and how many of those turned into a sale or booking.

Compare the current setup with the new one, not humans with AI in the abstract. Calculating out-of-hours misses separately, for example, shows the value of night and weekend cover on its own. If you are unsure of your figures, use a pessimistic conversion rate; if the decision still makes sense at that rate, it is a robust one.

• Lost calls = unanswered + abandoned − calls later reached
• Monthly lost revenue ≈ lost calls × conversion rate to sale or booking × average customer value
• Out-of-hours share = unanswered calls out of hours ÷ all unanswered calls

How do you compare total cost in a blended model?

A realistic comparison puts a few scenarios side by side: your current setup, a setup with extra agents, and a blended setup where the AI takes repetitive calls and transfers the rest. Use the same formulas for each and include lost revenue in the total.

In the blended model, the key variable is the share of calls the AI can resolve. Estimate it from the call-topic breakdown you built earlier: the share of topics with fixed, written answers is a reasonable starting point. Replace that estimate with real transcript data during a trial.

• Total monthly cost = (calls resolved by AI × AI cost per call) + (calls handled by people × human cost per call) + monthly lost revenue
• Scenario A: current team, current missed calls
• Scenario B: extra agents, fewer missed calls
• Scenario C: AI plus current team, including out-of-hours cover
• Sensitivity: rerun every scenario with pessimistic assumptions

What should you look at when costing Xaly?

When you add Xaly to the model, keep the plan and the phone line on separate lines: you can use your own line, or the line is provided through our licensed carrier partner. Current plans and usage options are on the pricing page, and larger organisations receive a quote based on their needs.

Two features bear directly on the lost-revenue line. Xaly retries unanswered outbound calls at the time the person is most likely to answer, calls web-form leads straight away and calls back people at the time they asked for. In outbound campaigns the assistant calls the consented contacts you upload, moves interested people to an appointment or quote, and measures which of two conversation scripts performs better; commercial calls require prior consent and a check against the national consent register.