Model it yourself
Your numbers, not ours. Including the ones that make us look bad.
Every assumption in this calculator is yours to change, and one of them can turn the answer negative. That is deliberate. A model you cannot argue with is a sales tool; a model you can break is worth taking to your CFO. Nothing you type here reaches us.
Your book
Everything your floor tries to dial in a month, not just the calls that connected.
The unworked tail: small balances, low-value renewals, the queries nobody gets to. Leave this at zero and the calculator gives you no credit for it at all. On most books it is the largest number on this page.
The seats actually working this book today.
Salary plus supervision, infrastructure, attrition and training. Use the number your finance team uses, not the salary.
35%
Share of attempts that reach a live person.
12%
Of the people you actually speak to, how many end in the result you wanted: a payment, a renewal, a save, a qualified meeting.
The two assumptions that decide the answer
65%
Default set at 65 as a deliberately conservative placeholder. Push it past 85 and the model stops being credible; we say so in the panel on the right.
Default set at 3 rupees per attempt as a neutral placeholder, not a quoted rate. Put whatever figure you would defend internally. If you have a quote from anyone, use theirs.
Recovered amount, renewal premium, retained margin or qualified-lead value. Use the contribution, not the top line.
Your model · runs entirely in this browser₹
Annual difference in run cost
₹0
Cost of the floor today, less the cost of the smaller floor plus the platform. Change the completion rate on the left and watch how fast this moves; that sensitivity is the real finding.
Seats still needed for handed-over calls0
Cost of those seats₹0
Cost of AI attempts₹0
Run cost today₹0
Run cost after₹0
Monthly difference₹0
Outcomes per month today0
Outcomes per month after0
Cost per outcome today₹0
Cost per outcome after₹0
Value of additional outcomes₹0
What this model does, and the four places it is wrong
- It assumes handed-over calls convert at the rate your floor manages today. In practice an agent who arrives with the full context usually does better, because the call starts at the decision instead of the introduction. We left that upside out rather than argue for it.
- It holds your connect rate constant. Real programmes usually improve it, because attempts move to the hours a segment is actually reachable and retries stop being spent on dead numbers. Again, left out.
- The unworked tail is an input, and it defaults to zero. Leave it there and the outcome side of this page will barely move, which is exactly what you should expect. On most real books that tail is the largest number available: the volume nobody can afford to call today becomes serviceable when cost per attempt collapses. If you know roughly how big yours is, put it in.
- It says nothing about implementation effort. Integration, list hygiene and handoff design take real time from your team. If your data is stale, the first fix is the data, and no model on a web page will tell you that.
Taken together, those four mean the honest reading of a positive result here is directional. The cost side is reasonably solid. The outcome side is conservative. Neither is a promise, and we will not quote a price off this page.
The next step is the only one that settles it
A model is an argument. A pilot is evidence.
Bring the numbers you just typed to a call and we will tell you which of them we think are wrong, and where. Then, if it is worth doing, we agree a baseline on one real process before anything runs, so the result cannot be argued about afterwards.