Key Takeaway:
Measuring AI ROI (return on investment) comes down to five approaches used by real finance leaders at companies doing this well: tie AI outcomes to existing OKRs rather than building separate AI metrics; segment spend by team and activity type to create measurable baselines before adoption; use the Double-Click method – keep asking “so what?” after “time saved” until you reach a real financial outcome; track team output against headcount over time; and combine AI and headcount costs into one budget per outcome.
Improving your AI return on investment starts with understanding the measurement frameworks before you need it. Measuring ROI after a vendor’s renewal quote arrives with a 30% increase attached puts you behind the curve.
What Makes AI ROI Hard to Measure?
The core difficulty with measuring AI value is that the benefit typically adds up at the individual level long before it compounds into something a business can report on. And most companies feel AI working before they’ve built any way to measure it. By the time they want to quantify the value, the baseline is gone and the before-and-after comparison is near impossible. Three things make this structurally difficult:
- The value goes to your employees, not your P&L: When someone uses AI to finish work faster, that extra time hardly becomes more output for the company. It usually tracks as slightly faster outputs or better work-life balance for employees (which is not a bad thing). The employee wins, but the organization absorbs the cost and the P&L shows nothing. There is a gap in how organizations capture this value.
- You didn’t set a baseline before you started: ROI requires a before and an after. Most AI rollouts happened fast – a tool was bought, people started using it, and nobody documented what “before” looked like. Now the tool has been in use for a year, everyone agrees it’s helpful, and no one can say by how much. According to the IBM Think Insights from February 2026, 79% of executives report productivity gains from AI. Only 29% can measure them confidently. The gap is almost entirely a baseline problem.
- Most AI is still experimental, not embedded: Repeatable ROI comes from AI that’s embedded in core workflows at scale, not from a handful of people using a tool they like. The IBM CEO Study (2025) found only about 25% of AI initiatives deliver expected ROI, and just 16% have scaled enterprise-wide. The rest are still in pilot territory, which is fine early on, but it means the financial case is still being built.
How to Measure AI ROI: 5 Frameworks from CFOs at Zapier and Nium
An AI ROI framework built around goals your organization already tracks – rather than separate AI-specific dashboards nobody checks – is the foundation both of these approaches share.
These five frameworks come from a Tropic webinar (available on demand here) with Ryan Roccon, CFO at Zapier (where 97% of the organization uses AI), and Andre Mancl, CFO at Nium (a global payments infrastructure company). Both leaders and their companies have made using AI an organizational mandate. They’ve scaled AI usage and ROI measurement with growing success.
1. Tie AI to your existing OKRs, not separate metrics
Andre Mancl’s framework at Nium is elegantly simple: don’t build a separate AI measurement system. Instead, ask whether AI investments are accelerating the timelines on goals you’re already tracking. If you’re spending on an AI tool that’s supposed to speed up sales cycles, that improvement should show up in your existing sales OKR – not in a dashboard labeled “AI impact.” If it doesn’t show up there, you have your answer.
2. Segment your AI spend before you can’t
Ryan Roccon at Zapier segments AI usage by team, user, problem type, and activity. That granularity creates baselines. Without baselines, you’re just guessing. The time to start tracking is before you need the data, not after a vendor sends a renewal quote with a 30% increase attached.
3. Keep asking “so what?” until you hit a real number
This is the Double-Click Framework. When someone says “time saved,” ask further: time saved doing what? And what did they do with that time? And what business outcome did that produce? Keep going until you land on something measurable – revenue, cost, cycle time, error rate. If you can’t get there, “time saved” is an anecdote, not a return.
4. Watch output vs. headcount over time
This one cuts through the self-reporting problem entirely. If your team’s output is rising while headcount is flat or shrinking, AI is creating real organizational capacity – regardless of whether anyone can articulate exactly how. This is the Team Output Framework: compare what a function produces against how many people are producing it, and watch the ratio over time.
5. Put AI and headcount in the same budget envelope
The Unified Budget Framework: combine AI tool costs and headcount into one budget for each outcome area. Let the team lead decide the most efficient mix. This removes the artificial wall between “software budget” and “headcount budget” and forces a real conversation about what AI is actually replacing or enabling (versus what it’s sitting alongside at full cost).
What Metrics Do CFOs Actually Use to Measure AI ROI?
Here are the specific signals and indicators Ryan Roccon and Andre Mancl watch in practice as CFOs. These aren’t generic AI ROI metrics pulled from a survey; they’re the signals two finance leaders actively use to make budget decisions and evaluate whether a given AI investment is earning its cost.
OKR-Driven Metrics (From Andre Mancl, CFO at Nium)
Output and Outcome-Driven Metrics (From Ryan Roccon, CFO at Zapier)
For a practical breakdown of how to benchmark and control AI vendor costs across your full software portfolio, see the buyer guide on how to manage AI costs.
The shared signal: both CFOs watch seat utilization and adoption closely. Andre flags low utilization as overspending. Ryan measures AI fluency in Zapier’s hiring process. If your licensed users aren’t using the AI features you’re paying for, that’s your strongest negotiating position at renewal – it’s evidence the value case hasn’t been made by the vendor or by your own rollout.
Hard AI ROI vs. Soft AI ROI: What the Difference Means in Practice
Hard AI ROI shows up in your P&L, soft AI ROI doesn’t. But what Ryan and Andre’s experience adds is a more useful frame – the real question isn’t what type of value AI creates. It’s whether the organization captures it, or the employee does.
Hard AI ROI
Hard ROI is the stuff that shows up in your P&L. It’s directly measurable, financially defensible, and what your CFO can actually point to in a budget review. Both CFOs have concrete examples where the organization unambiguously got the benefit.
For Ryan at Zapier, the accounting team’s story is the clearest: over three to four years, the team got smaller while complexity exploded through multiple entities, international expansion, Big Four audits and month-end close got shorter. Nobody walked away with more free time. The organization captured the efficiency. And the credit calculation example adds a revenue dimension. The surface metric is time saved on manual spreadsheet work. Ryan doesn’t stop there: “Now we’re giving customers an answer immediately, which means our probability of upselling them is incredibly higher.” Time saved is the input. Upsell rate is the hard ROI. Same AI investment, different question asked.
For Andre at Nium, the reconciliation example is equally concrete: a 1-2% manual tail in payments reconciliation that used to require spreadsheet work now resolves instantly. There’s a before and an after, and the difference is measurable. Andre’s nuance here is worth noting: hard ROI for Nium isn’t primarily about cost reduction. “It’s about improving operational efficiency, not reduction in spend.” Speed, accuracy, and time to close are his hard metrics, not just dollars saved. That’s a broader and more useful definition of hard ROI than the standard P&L framing.
Soft AI ROI
Soft ROI is real – it’s just harder to put a number on. Employee morale, decision quality, customer experience improvements. These definitely matter to the business, but they don’t show up cleanly in a quarterly P&L.
Ryan described how when employees get to automate the work they don’t want to do, the mundane, repetitive, low-value-add stuff, they naturally focus on the opposite – “and the company benefits from them focusing on the most valuable work.”
This example of soft ROI – an employee spending time on higher-value work instead of manual processing – can generate hard ROI over time. The catch is that the chain has to be traced. Without the Double-Click framework, it stays anecdotal.
Ryan also flagged a harder version of this: “It’s very easy to see if engineers are writing more code per engineering hour. But the quality could be poor, or you need more review.” Even apparently hard metrics can mask soft outcomes if you don’t go one level deeper.
The procurement test: ask vendors what happens after “time saved.” If the answer is “employees will be happier” or “work will feel more meaningful,” that’s the soft ROI case. Real and valuable, but not sufficient to justify a 20–37% renewal increase. The hard ROI case requires a downstream business outcome that survives the Double-Click: a faster close rate, a measurable upsell lift, a documented process elimination. If the vendor can’t name one, the premium hasn’t been earned.
Why Measuring AI ROI is a Financial Problem To Fix ASAP
Right now, software vendors are imposing 20–37% AI-driven price increases at renewal. Their justification is always some version of “AI adds significant value.” The effective counter to that isn’t “we disagree” – it’s “Prove it. What measurable outcomes have customers like us actually seen, and what will you commit to?”
That counter only lands if you’ve done your own homework. If you don’t have adoption data, usage metrics, or any baseline showing what AI is doing for your team, you’re negotiating on feel.
The teams that negotiate well on AI pricing are the ones who walked into the renewal conversation with their own numbers. Adoption rate. Output metrics. A clear answer to “what has this tool actually changed for us?” That’s the leverage you need.
Tracking AI investment ROI before renewal conversations is the difference between negotiating from a position of evidence and absorbing increases based on vendor claims alone.
The contract terms that lock in that leverage – renewal caps, SKU protection, and credit consumption definitions – are covered in SaaS & AI Contract Negotiation: 8 Terms to Negotiate.
FAQs: How To Measure AI ROI
What is AI ROI?
AI ROI (return on investment) is the measurable business value an organization captures from its AI software investments – revenue growth, cost reduction, time saved, and capacity created. For finance and procurement teams, it’s also the number needed to evaluate whether AI-driven SaaS price increases at renewal are actually justified. Fewer than one-third of companies can demonstrate AI ROI financially today, despite most reporting that AI creates value.
How do you measure AI ROI?
Measuring AI ROI comes down to five frameworks used by CFOs at Zapier and Nium: tie AI outcomes to existing OKRs rather than building separate AI metrics; segment spend by team and activity type to create baselines before adoption; use the Double-Click method – keep asking “so what?” after “time saved” until you reach a real financial outcome; track team output against headcount over time; and combine AI and headcount costs into one budget per outcome.
What is the Double-Click Framework for measuring AI ROI?
The Double-Click Framework is a method developed from CFO practice at Zapier: when someone says “time saved,” keep asking “so what?” until you reach a concrete business outcome – revenue, cost, cycle time, or error rate. If the chain can’t reach a measurable financial outcome, “time saved” is an anecdote, not a return. Ryan Roccon at Zapier applied this to credit calculation time savings and traced it to a measurably higher upsell conversion rate.
What is the difference between hard AI ROI and soft AI ROI?
Hard AI ROI shows up in your P&L – faster close cycles, lower headcount for the same output, measurable upsell rate improvements, documented process elimination. Soft AI ROI is real but indirect – employees doing more meaningful work, improved morale, better decisions – and doesn’t translate to short-term financial results. The distinction that matters most at renewal: soft ROI cannot justify a 20–37% vendor price increase. If a vendor’s value claims don’t survive the Double-Click test and land on a hard business outcome, the premium hasn’t been earned.
What metrics do CFOs actually use to measure AI ROI?
CFOs at Zapier and Nium track distinct but complementary metrics. Andre Mancl at Nium watches time to delivery, time to close, accuracy, seat utilization rate, and manual process elimination rate. Ryan Roccon at Zapier tracks team output vs. team size over time, month-end close cycle time, deal and upsell conversion rate, and cost per AI workflow. Both CFOs flag low seat utilization as the clearest early signal that ROI won’t materialize – and the strongest negotiating position at renewal.
How does AI ROI affect software vendor negotiations?
Vendors justify 20–37% AI-driven renewal increases by claiming AI adds significant value – the effective counter requires your own numbers, not disagreement. Adoption rate, output metrics, and a clear answer to what the tool has actually changed for your team is the evidence that shifts the conversation. Buyers who walk into renewal conversations with that data negotiate significantly better outcomes than those relying on gut feel.
What should you ask a vendor before accepting an AI-driven price increase?
Ask vendors what downstream business outcomes – not just “time saved” – similar customers have achieved, and at what adoption rate. Ask for a reference customer who can share before-and-after metrics. If the vendor can only point to soft ROI claims like improved employee satisfaction or better decisions, the hard ROI case hasn’t been made and the premium hasn’t been earned.
