How do I build a business case for AP automation that survives CFO scrutiny?

Credible: labor time on automatable tasks (verifiable against your actual process), duplicate and error reduction (measurable from your history), and discount capture (computable from your terms). Fantasy or inflated: best-customer automation rates applied to your different mix, "hours saved" multiplied by fully-loaded rates as if every saved hour converts to cash, and soft benefits assigned hard dollar values without basis. Stress-test by substituting your own numbers for the vendor's defaults, halving the optimistic assumptions to see if the case still holds, and demanding the vendor justify any rate they quote against a customer with your mix. A business case that only works at the vendor's best-case assumptions isn't a business case.

Gather before you talk to vendors, so your baseline is yours, not theirs. Collect: AP labor (salaries times the share of time actually on AP, including partial allocations), current AP software costs, measured error costs (duplicates paid, late fees, rework time), and a defensible overhead share. Compute current cost per invoice fully loaded. This baseline is the foundation of every ROI claim - a case built on a real current-state number is defensible; one built on a vendor's assumed baseline is theirs to revise. Owning the baseline means owning the argument.

Anchor each soft benefit to something measurable even if imperfectly: errors to recovered duplicates and avoided late fees, faster close to days saved times the cost of close-period effort, vendor relationships to reduced inquiry volume and captured discounts, audit readiness to hours saved in audit prep. Present them as a secondary tier with stated methodology and conservative values - "here's the basis, here's the conservative estimate." Soft benefits sound hand-wavy when asserted; they sound credible when tied to a number with a method, even a rough one.

Payback varies with volume and current-state cost, but the case strengthens with invoice volume because savings scale while software cost grows slower. The math gets hard at low volume - too few invoices to generate enough savings against the platform cost - which is why complexity and volume, not just headcount, define the fit. Below a certain volume and complexity, the honest answer is that automation may not pay back, and a credible business case says so rather than forcing it.

Rebuild on your own measured baseline rather than vendor projections, propose a paid pilot on your invoices to generate real before/after evidence, and present conservative savings with sensitivity analysis showing the case holds even at half the optimistic assumptions. "Unproven savings" is a rejection of vendor assumptions presented as fact - counter it with your data and a pilot that proves the savings on your operation before the full commitment. Evidence the CFO can trace beats projections the CFO can't verify.

Be precise about which you mean: avoidance (absorbing growth without the hires you'd otherwise need - the most defensible and least disruptive), redeployment (moving people to higher-value work - credible if you specify the work), and reduction (actual cuts - real but morale-costly and to be stated plainly if it's the plan). Mixing them or implying avoidance while planning reduction destroys credibility. Honest framing names the actual mechanism; most strong cases lead with avoidance and redeployment.

When cuts are coming regardless, the automation case reframes from "savings" to "capacity preservation" - automation lets the smaller team maintain output that would otherwise collapse, protecting close quality, controls, and payment reliability through the reduction. The argument shifts from ROI to risk: without automation, the cut team can't sustain the work, and the failures (late payments, control gaps, errors) cost more than the software. Automation becomes the thing that makes the RIF survivable rather than catastrophic.

Quantify each from your data: discount capture as the annual dollars currently missed (computable from terms and payment timing), duplicate prevention as historical duplicate payments times a prevention rate, fraud avoidance as exposure reduced (harder, but estimable from risk). Presented as line items with methodology, these often rival the labor savings - and they're more durable, because they're cash recovered rather than soft efficiency. Move them out of the footnotes; they frequently carry the case.

Define the metrics before go-live (cost per invoice, cycle time, touchless rate, discount capture, error/duplicate rates), baseline them, and track the same metrics after - so realized ROI is measured against the promise, not asserted. Reporting actual results closes the loop and builds the credibility that makes the next automation investment an easy approval. Teams that prove the first case get the benefit of the doubt on the second; teams that never measure get scrutinized forever.

(1) What's your assumed automation rate, and will you commit to it on our invoice mix in a paid pilot? (2) How much of your "hours saved" actually converts to cash versus just freed time? (3) What's the baseline you're comparing against - ours or your assumption? (4) Show me a customer with our mix and their realized ROI, not projected. (5) What's the implementation cost and timeline that the ROI conveniently omits? Each forces the vendor off the marketing number and onto evidence.

Stampli's ROI story maps to three honest value categories rather than a single inflated number: faster processing (shorter cycles, fewer exceptions, faster closes without cleanup), less headcount (higher throughput per finance employee, growth absorbed without proportional hiring - framed as avoidance, not layoffs), and smarter spending (discount capture, fewer surprises, better control). The headcount framing matters for CFO credibility: Stampli's approved language is absorbing growth without proportional hiring, not "cut staff" - which is both more accurate and more defensible in a business case than headcount-reduction claims a CFO will probe.