When automation saves a company more than ₹3 crore in resources, the board does not ask which library you used. They ask: what counted as a resource, who agreed the baseline, and is the saving durable? Crore-scale claims only stick when you can show the math — hours, leakage, and avoided cost — not bot volume.
Here is a measurement model that holds up in leadership rooms, written for people who aren’t automation specialists but need to trust the number.
Start with the current process cost
Before any script ships, price the status quo. This is the baseline everything else depends on. Get it wrong and your entire ROI story collapses at the first finance review.
- Labour capacity — hours per cycle × volume × fully loaded cost (or opportunity cost of skilled time). If a finance analyst spends four hours per vendor setup and you process 200 vendors a month, that’s 800 hours — price it.
- Rework — percent of cases that bounce, plus the cost of each bounce. A 15% bounce rate on 500 monthly cases means 75 cases need a second pass. Each pass has a cost.
- Delay cost — payment holds, late decisions, stockouts, or customer wait — only where you can tie delay to money or capacity. Don’t invent numbers; use what finance already tracks.
- Leakage — duplicate spend, missed discounts, avoidable penalties, or uncontrolled access risk if finance owns a number for it.
Write assumptions in plain language. Invite challenge. A contested baseline is healthier than a silent one. If ops and finance disagree on hours per case, resolve it before you automate — not after you claim savings.
Work through a simple example
Suppose vendor onboarding takes five hours per case at a fully loaded cost of ₹800 per hour. You process 300 vendors per month. That’s 1,500 hours, or ₹12 lakh per month in labour capacity — ₹1.44 crore annually, before rework.
If automation cuts handling time to 1.5 hours and reduces bounce rate from 20% to 5%, you reclaim roughly 1,050 hours per month plus 45 fewer rework cases. At the same rate, that’s roughly ₹10 lakh per month — but only if adoption reaches 90%+. At 60% adoption, the real saving is closer to ₹6 lakh. Adoption isn’t a footnote. It’s the multiplier on everything else.
Separate one-time from recurring
Build and change-management costs are real. Recurring savings are what compound. Report both, clearly labelled.
A one-time data cleanup might save ₹30 lakh in the first quarter — that’s a windfall, not a run rate. Don’t mix it into the annual figure without saying so. Leadership respects honesty about what repeats and what doesn’t. Inflating run-rate with one-time wins is how automation programs lose credibility at year two.
Track adoption as a gate
If 40% of volume still runs on chat and spreadsheets, your theoretical saving is fiction. The automation might work perfectly on the 60% that uses it — but the 40% workaround still costs what it always cost.
Pair rupee estimates with adoption percentage and open exception count. Show all three on the same slide: “Theoretical saving at 100% adoption: ₹1.2 Cr. Current adoption: 72%. Realized saving: ₹86 lakh. Open exceptions: 23, average clear time 1.4 days.” That is how you protect a ₹3 Cr+ story from becoming a slide that dies in QBR.
Account for exceptions honestly
Exceptions aren’t failures — they’re the design working. But they have a cost. If 10% of cases hit the exception queue and each takes 45 minutes of human time to clear, that cost belongs in the model.
As adoption grows and data quality improves, exception rates should fall. Track that trend. A shrinking exception queue is evidence that savings are durable. A growing one is a warning that your run-rate number is optimistic.
Phrase savings as resources conserved
“Headcount cut” is a political landmine. “Capacity returned to higher-value work,” “rework avoided,” and “leakage closed” are clearer and more honest. Leadership can still convert capacity to rupees; you don’t need to oversell job loss.
Finance understands capacity. HR understands capacity. Framing savings as resources conserved — hours, cases, rupees — keeps the conversation productive. Framing it as people eliminated turns a operations win into a political fight.
Review the number on a rhythm
Quarterly, refresh volume, rate, adoption, and exception trends. If the process changed — new policy, new volume, new team — update the model. Durable crore outcomes come from governance, not from a one-time calculator in a kickoff deck.
The number should get more accurate over time, not less. Early estimates are directional. By quarter three, you should know within 10% what you actually saved. That discipline is what turns a project into a capability.
The takeaway
Measure before you build. Baseline with finance in the room. Multiply by adoption, not theory. Subtract exception cost honestly. Separate one-time from recurring. Review quarterly. A ₹3 Cr+ claim that survives scrutiny is worth more than a ₹5 Cr claim that doesn’t.