Business case calculator

What would AI actually save your contact centre?

A first-pass estimate in two minutes, with every assumption on show. Replace the numbers with your own: the defaults are illustrative, not benchmarks. Nothing you enter leaves your browser.

Your numbers

The symbol only changes the labels. Enter costs in whichever currency you budget in.

Your operation

Contacts handled per month

Calls, chats and emails that reach a person today.

Average handle time

min

Talk or chat time plus wrap-up, per contact.

Fully loaded cost per agent hour

Salary, on-costs, management, seats and systems, divided by paid hours.

Occupancy

Share of paid time agents spend handling contacts. Every handled minute costs more than a minute of pay.

What AI takes on

Contacts in scope for AI

The share of contacts the AI will attempt: simple, repeatable, low-risk reasons.

Resolved by AI, of those in scope

Genuinely resolved, not merely contained. A customer who calls back was not resolved.

Handle time saved by agent assist

Summaries and suggested answers on the contacts people still handle. Use 0 if not in scope.

Extra agent time when AI hands over

sec

Customers who tried the AI first often arrive frustrated or needing to repeat themselves.

What AI costs

AI cost per contact attempted

Platform, model usage, telephony and speech, per in-scope contact, resolved or not.

One-off implementation cost

Design, integration, testing, training and change management.

Annual running and governance cost

The people who tune, monitor, audit and improve it. This is never zero.

Paid hours per agent per year

hrs

Used only to turn hours into full-time equivalents.

Net annual benefit, central case

to pay back the implementation

net over three years, after implementation

agent hours freed a year

full-time equivalents

Three scenarios
	Cautious	Central	Optimistic

Cautious assumes 60% of your resolution rate, half the assist saving, 50% more handover time and 25% higher AI and running costs. Optimistic assumes 130% of your resolution rate (capped at 95%), 50% more assist saving, half the handover time and 15% lower AI cost.

a year · payback Results ↓

How the numbers work

Simple arithmetic, shown in full.

Today's cost is contacts, times handle time, divided by occupancy, times the hourly cost. Occupancy matters: at 80%, every minute spent on a contact costs a minute and a quarter of paid time.

Savings come from two places: contacts the AI resolves outright, and minutes that agent assist takes off the contacts people still handle.

Costs come from three: the AI's per-contact charge on everything it attempts, the extra agent time when it hands over, and the team that runs and governs it.

The net annual benefit is the savings less those costs. Payback divides the implementation cost by the monthly net benefit.

What this leaves out

The parts that decide whether it works.

Customer outcomes. Satisfaction, complaints and churn can move either way. A badly designed bot can cost more in lost customers than it saves in wages.
Repeat contact. If "resolved" really means "gave up", the calls come back. Measure resolution, not containment.
How savings are realised. Freed hours only become cash through attrition, rostering or redeployment. Hours that nobody plans for simply disappear.
Ramp-up. Results rarely arrive on day one. Expect months of tuning before the central case holds.
Risk and regulation. Data residency, disclosure duties and audit requirements add cost, particularly in regulated sectors and the GCC.

Why containment is the wrong number to chase What 24/7 customer service really costs We built an AI contact centre. Here is what it took.

Want this done properly, with your real data?

A real business case starts from your contact reasons, your volumes and your systems. We build them with clients in a few weeks, and we will tell you if the answer is not yet.

Start a conversation