How the math works
No black box. This is the entire calculation:
human cost per conversation = (handling minutes ÷ 60) × hourly agent cost current monthly cost = conversations × human cost per conversation AI-resolved conversations = conversations × resolution rate projected cost = ((conversations − AI-resolved) × human cost) + (AI-resolved × AI cost) + platform fee monthly saving = current monthly cost − projected cost break-even volume = platform fee ÷ (resolution rate × (human cost − AI cost))What the break-even number tells you
Break-even volume is the point where your savings exactly cover the subscription. Below it, the platform costs more than it saves. This is why small operations with low ticket volume should start on a free or entry plan: at 200 conversations a month, no platform fee is justified by deflection alone.
The assumptions that matter most
- Resolution rate. Every result scales with this. It is also the number vendors are most optimistic about, and the hardest to predict before you test. Use 40% until you have measured your own.
- Fully-loaded agent cost. Most people enter the wage and forget tax, benefits, tooling and management. That understates savings — and it also understates what you are currently spending, which is the number that should worry you.
- Handling time. Include follow-ups. A conversation reopened three times is not a five-minute conversation.
- Not everything is deflectable. Damaged goods, disputes and anything needing account access should be excluded from your resolution rate, not averaged into it.
What to do with the result
- If the saving is positive: trial the recommended platform, then re-measure your real resolution rate after 30 days and run these numbers again with the true figure.
- If the saving is negative: either your volume is too low for a paid platform, your AI cost per resolution is too high, or too little of your queue is deflectable. The first is fine — start free. The other two mean AI support is not your problem right now.
- In every case: measure deflection, not ticket reduction. The number of conversations does not usually fall; what falls is the number a human has to handle.
Start free on Tidio Text pricing HighLevel trial
Affiliate links — we may earn a commission at no extra cost to you.
FAQ
Is any of my data sent anywhere?
No. The calculation runs entirely in your browser in JavaScript. There is no form submission, no analytics call from this page, and no server involved. Close the tab and the numbers are gone.
What resolution rate should I use?
Start at 40%. Industry benchmarks put first-month autonomous resolution at 30–45% and mature, well-maintained deployments at 50–70%. The honest answer is that your ticket mix decides it, so trial a platform and measure before you commit a budget.
Why does the platform fee barely change my result?
Because in most support operations the subscription is small relative to the cost of human handling time. At 1,000 conversations and 8 minutes each, you are buying roughly 133 hours of agent time a month — which is why the resolution rate and the hourly cost dominate the answer.
Should I include the time spent maintaining the AI?
Yes, ideally. The model above does not, and knowledge-base maintenance is real ongoing work — answers go stale and need fixing. If you want a conservative result, raise your effective cost per AI resolution to account for it, or lower your resolution rate.
Related: every plan and limit compared · for SaaS support teams · for Shopify stores.