The ROI Framework

Core Concepts

The Four-Dimension ROI Framework

AI investments create value in four distinct ways. Each dimension is measurable, but they require different evidence and different methods. A complete ROI case addresses all four.

Dimension 1: Cost Direct reduction in spend. This includes labor hours replaced or reduced, vendor tool consolidation, reduced rework, and infrastructure efficiencies. Cost is the most familiar ROI dimension and the easiest to quantify, but it is rarely the most important one for growth-stage organizations.

Example at X-company: Three product managers each spend six hours writing a PRD. AI drafting reduces that to roughly two hours per PRD. At twelve PRDs per month across the team, and a fully-loaded cost of $80 per hour, the math is: 3 PMs x 4 hours saved x 12 PRDs x $80 = $11,520 per month. At 80% adoption (accounting for the PRDs where PMs still work manually), that lands at approximately $17,280 per month in recovered capacity. This is not headcount reduction. It is hours shifted from drafting to higher-value work.

Dimension 2: Time Cycle time compression. This captures how AI shortens the time between a trigger and a useful output, whether that's a feature reaching users, a new hire becoming productive, or a support issue getting resolved. Time value is often larger than cost value, but it requires a credible baseline.

Example at X-company: New client onboarding documentation previously required a five-day cycle: requirements gathered from the account team, written by a PM, reviewed by the client success lead, revised, finalized. With AI drafting from a structured intake form, that cycle compresses to one day. With two onboardings per month, X-company recovers eight person-days per month. The downstream effect: clients begin using the product sooner, reducing the time-to-value gap that their churn analysis consistently flags.

Dimension 3: Quality Reduction in error rates, defect rates, mis-routing, or any outcome variance that creates downstream cost or risk. Quality gains are often invisible until you measure the baseline: they don't show up on a P&L until you connect them to a cost or time consequence.

Example at X-company: Before AI-assisted support triage, 18% of incoming tickets were mis-routed: sent to the wrong team, requiring a handoff, a re-read, and a delayed response. Average resolution overhead per mis-routed ticket is $45 (support staff time). AI triage brings mis-routing down to 6%. At 400 tickets per month, that's 48 fewer mis-routed tickets. 48 x $45 = $2,160 per month in recovered support capacity, plus a measurable improvement in time-to-first-response for the affected tickets.

Dimension 4: Capability New things you can now do that were not economically or operationally feasible before. This is the hardest dimension to quantify and the most strategically significant. It is where AI creates competitive differentiation rather than operational efficiency.

Example at X-company: X-company's product team identified that professional services clients wanted a per-client project health summary: a weekly digest combining milestone progress, resource utilization, and risk flags for each of their active projects. Building this manually was not feasible: it would have required a dedicated analyst role per client segment. AI makes it feasible. The feature ships. It becomes a differentiator in competitive deals. One enterprise contract renewal, where the health summary was cited explicitly in the renewal conversation, represents more revenue impact than all three other dimensions combined.

How the Dimensions Interact

The four dimensions are not alternatives. They stack. A strong ROI case uses all four because:

• Cost and time are the easiest to defend with data: use them to establish credibility
• Quality improvements are often the most defensible at the unit level: use them to show rigor
• Capability gains are often the most strategically significant: use them to make the case memorable

An ROI narrative that only cites cost savings looks like an efficiency play. One that only cites capability looks unquantified. The combination is harder to dismiss.

What Makes an ROI Claim Credible

Before building a number, you need three things:

1. A baseline. You cannot measure improvement without a starting point. If you don't have a measured baseline, you need an estimated one with a stated assumption.
2. An adoption rate. Benefits only materialize when people use the tool consistently. Build in a realistic adoption rate (60-85% for well-supported rollouts, lower for unsupported ones).
3. A causal link. The AI change must be the plausible cause of the improvement, not a coincidence. If ticket volume dropped and you also hired three support staff, AI triage deserves partial credit at most.

What Makes an ROI Claim Inflated

Watch for these patterns in vendor pitches, internal proposals, and your own analysis:

Key Points

• The four ROI dimensions are: Cost, Time, Quality, and Capability
• Cost and Time are easiest to quantify; Capability is most strategically significant
• Every credible claim requires a baseline, an adoption rate, and a causal link
• A complete ROI narrative uses all four dimensions; single-dimension cases are incomplete
• Inflated projections share common patterns: full-replacement math, 100% adoption assumptions, uncaused attribution

---

Actionable Takeaways

1. Pick one AI initiative you are currently running or evaluating. Write down which of the four dimensions apply. If you can only name one, your ROI case is incomplete.
2. Find your baseline. For the initiative you chose, identify the current state metric for each dimension you named. If you don't have measured data, write down your best estimate and label it as an estimate.
3. Apply an adoption discount. Take your projected benefit and multiply it by 0.70. If the case no longer holds, the underlying initiative may not be as strong as it appears, or you need a better change management plan.
4. Identify your capability dimension. Ask: what does this AI initiative make possible that was not feasible before? If the answer is "nothing new, just faster," the strategic case is weaker than it could be.
5. Write one paragraph that combines all four dimensions into a single ROI narrative. Use the format in the Implementation Workflow below.

---

Practical Examples

Example 1: A PM Defending Budget

X-company's Head of Product is presenting the AI tooling budget to the CFO. She has two options.

Version A (single dimension): "We're saving about $17K per month on PM time from the AI PRD drafting tool."

The CFO's first question: "What happens to those hours?" If she can't answer, the case collapses into a cost number that the CFO will reclassify as a nice-to-have.

Version B (four dimensions): "The PRD drafting tool recovers roughly $17K per month in PM capacity at 80% adoption. That time is being redirected to customer discovery, which reduced our average feature cycle from 11 weeks to 8. Support triage is reducing our mis-routing rate from 18% to 6%, worth about $2,100 per month in resolution overhead. And the per-client health summary, which wasn't feasible to build without AI, is now a named reason in two of our last four enterprise renewals."

The CFO still asks hard questions. But this version has four entry points, and only one of them needs to land.

Example 2: An Engineer Evaluating an AI Testing Tool

A senior engineer at X-company is assessing an AI-assisted test generation tool. He structures his evaluation across the four dimensions:

• Cost: Test authoring currently takes 2 hours per story. AI generation would reduce that to 30 minutes. At 60 stories per sprint, two-week sprints, that's 90 hours per sprint recovered at 80% adoption.
• Time: Faster test authoring shortens the PR review cycle, which currently bottlenecks at test coverage gaps. Potential to reduce review cycle by one day per sprint.
• Quality: Baseline: 14% of production bugs in the last quarter were in areas with low test coverage. AI coverage recommendations may reduce that. No projection yet because there is no intervention data.
• Capability: The tool generates contract tests automatically from API schemas. This was previously a manual, rarely-done task. It enables a testing practice X-company currently does not have.

He presents this as an evaluation matrix, not a promise. The capability dimension is what wins the approval: it is not a marginal efficiency but a new engineering practice.

Example 3: A Founder Pitching an AI Feature to the Board

X-company's CEO is presenting the per-client health summary feature to the board. She resists the temptation to frame it purely as a capability story and structures it across all four dimensions:

• Capability: This feature was not economically feasible before AI. It required synthesizing data across five internal systems in real time per client, per week.
• Cost: No additional headcount required to operate it at current scale.
• Time: Feature shipped in six weeks. Pre-AI, a comparable feature would have required a six-month data engineering engagement.
• Quality: Two beta clients reported a 40% reduction in the number of status meetings they needed with their own stakeholders, based on the summary replacing manual status reporting.

The board approves continued investment. The quality dimension, which came from a customer survey rather than an internal metric, is the most persuasive element in the room.

---

Implementation Workflow

Use this workflow to build a complete ROI case for one AI initiative you are currently running or evaluating.

1. Name the initiative. Write one sentence describing the AI initiative: what it is, who uses it, and what it is intended to change. Example: "AI-assisted PRD drafting tool used by three product managers to generate first-draft requirements documents."

2. Map the four dimensions. For each dimension (Cost, Time, Quality, Capability), write one sentence describing how the initiative creates value in that dimension. If a dimension does not apply, write "N/A" and explain why.

3. Establish baselines. For each dimension that applies, write down the current state metric. Use measured data if available. If not, write an estimated baseline and label it "est." Example: "PRD drafting time: 6 hours per document (est., based on PM time tracking in last sprint)."

4. Project the improvement. For each dimension, write the expected new metric after the initiative is fully adopted. Cite the basis for your projection: vendor benchmark, pilot data, internal estimate.

5. Apply an adoption discount. Multiply each projected benefit by your expected adoption rate. If you do not have adoption data, use 70% as a conservative default for a supported rollout.

6. Calculate the cost dimension. Identify your unit of labor cost (hourly rate, fully loaded), multiply by hours saved per period, multiply by volume, multiply by adoption rate. Express as a monthly figure.

7. Calculate the time dimension. Identify the cycle that is being compressed. Calculate the days or hours saved per cycle, multiply by cycle frequency per month. Express in person-days or hours recovered per month.

8. Calculate the quality dimension. Identify the error or defect rate being reduced. Calculate the unit cost of each error (resolution time x labor rate, or penalty cost). Multiply: (baseline rate - new rate) x volume x unit cost.

9. Describe the capability dimension. Write two to three sentences describing what is now possible that was not feasible before. If there is a revenue or retention connection, name it explicitly, even if you cannot quantify it precisely.

10. Write the ROI narrative. Combine all four dimensions into a single paragraph of four to six sentences. Use this structure:

    • Sentence 1: Cost dimension with a number
    • Sentence 2: Time dimension with a number
    • Sentence 3: Quality dimension with a number
    • Sentence 4: Capability dimension (qualitative or with a revenue reference)
    • Sentence 5 (optional): Total monthly quantified value and strategic framing

    Example for X-company: "AI-assisted PRD drafting recovers approximately $17,280 per month in product team capacity at 80% adoption. Onboarding documentation cycle time has compressed from five days to one, recovering eight person-days per month and accelerating client time-to-value. AI-assisted support triage reduces mis-routing from 18% to 6%, saving roughly $2,160 per month in resolution overhead. The per-client project health summary, which was not operationally feasible before AI, now ships as a product feature and has been cited in enterprise renewal conversations as a differentiating capability. Combined quantified monthly value is approximately $21,440, with a capability dimension that directly supports net revenue retention."

11. Stress-test the case. Apply the credibility checklist: Does each claim have a baseline? Is there an adoption rate applied? Is the causal link defensible? Are any inflated projection patterns present? Revise any claims that fail the check.

12. Identify your strongest dimension. Review the four dimensions and mark the one that is most defensible with current data. This is the one you lead with in a presentation. The others provide depth, but your anchor is the number you can defend under questioning.