Chatbot ROI Calculator

A chatbot ROI model works by multiplying your current per-conversation cost by the number of conversations the bot will fully resolve without a human agent (containment rate), then subtracting the platform and maintenance costs. The result is your net monthly saving, which amortises the build cost to give a payback period.

What is containment rate?

Containment rate is the percentage of conversations the chatbot handles from start to resolution without escalating to a human. For well-scoped, FAQ-style support (returns, order status, password reset, opening hours), 60–75% is achievable. For lead qualification flows with structured questions, 70–85% is realistic. General enquiries with high variability are harder to contain — start at 40–50% as a conservative baseline.

What drives per-conversation cost?

Agent handling cost per conversation depends on your average handle time (AHT) and the loaded hourly cost of your support staff. A 6-minute conversation at £24/hour effective cost = £2.40. A 20-minute conversation at £35/hour = £11.67. Include overhead, management, and QA time in the hourly rate — typically 1.2–1.5× base salary. Use your AHT from your helpdesk system to set this figure accurately.

Beyond cost savings, chatbots typically provide 24/7 availability (reducing overnight human cover costs), faster first response (seconds vs minutes), and consistent messaging (no agent variation). These benefits are not modelled here but are often worth as much as the direct savings for businesses with significant out-of-hours volume.

Chatbot business cases live or die on one number: the deflection rate — the share of contacts fully resolved without a human. It is also the number most often overstated.

Vendor claims above 80% usually count any conversation the bot touched, including ones the customer abandoned in frustration. Measure resolution, not containment: did the customer's problem go away, or did they simply give up?

Where the savings actually come from

Rarely from removing headcount. In practice the gains are absorbing volume growth without hiring, cutting first-response time, covering nights and weekends without a rota, and freeing experienced agents from repetitive questions so they handle the complex work that retains customers.

The cost side

Build cost is one line. Also include model API fees, which scale with conversation volume, hosting, and — the one people forget — content maintenance. A chatbot is only as good as the knowledge behind it, and that needs owning as products, prices and policies change. Budget continuing effort, not a one-off build.

The number that decides it

Multiply your monthly contact volume by your fully loaded cost per contact, then by a realistic deflection rate. If that annual figure is not comfortably above build plus running costs, the case rests on service quality rather than savings — which can still be a good reason, but should be argued honestly.