AI ROI in Practice: Three Case Studies
AI Value Assessment — Lesson 9 of 10
Theory becomes conviction only when tested against reality. The preceding lessons have built a comprehensive framework for measuring AI value — four layers of returns, four classes of intangible assets, structured attribution methodologies, and a dashboard architecture. This lesson applies those frameworks to three detailed case studies drawn from different industries, investment scales, and maturity levels.
Each case study illustrates how the 4-Layer Framework reveals value that simpler measurement approaches miss, and how the intangible asset perspective transforms the investment calculus from "did this project pay for itself?" to "what has this programme built?"
In all three case studies, the traditional first-year ROI calculation captured less than 30% of the total value created by the AI investment. The 4-Layer Framework — when applied rigorously and measured over appropriate time horizons — revealed 3-5x more value than conventional metrics. The difference is not accounting tricks; it is the systematic inclusion of revenue growth, competitive advantage, and strategic optionality that simpler frameworks ignore.
Case Study 1: SaaS — AI-Powered Customer Success Platform
Company Profile
A B2B SaaS company with $45 million ARR, 2,200 customers, and a 120-person team. The company invested $3.8 million over 18 months to build AI capabilities across three use cases: predictive churn scoring, automated onboarding personalisation, and intelligent feature recommendations.
Investment Breakdown
| Component | Cost | Phase |
|---|---|---|
| Data engineering and integration | $1.1M | Months 1-6 |
| Model development (3 use cases) | $1.4M | Months 4-14 |
| Infrastructure (MLOps, monitoring) | $0.6M | Months 6-12 |
| Change management and training | $0.4M | Months 10-18 |
| Ongoing operations (annualised) | $0.3M | Year 2+ |
Layer 1: Cost Reduction — $1.2M per year
The predictive churn model identified at-risk accounts 45 days earlier than manual monitoring, enabling proactive intervention. The automated onboarding reduced customer success manager time per new account from 12 hours to 3.5 hours. Net savings after AI system operating costs: $1.2 million per year.
Layer 2: Revenue Growth — $3.6M per year
The churn model's early intervention reduced net revenue churn from 8.2% to 5.1%, protecting $1.4 million in annual recurring revenue. The personalised onboarding increased 90-day activation rates from 64% to 81%, accelerating time-to-value and increasing first-year expansion revenue by $1.2 million. The feature recommendation engine increased upsell conversion by 28%, contributing $1.0 million in incremental expansion revenue.
The company's traditional ROI calculation — ($1.2M Layer 1 savings) / ($3.8M investment) = 32% first-year ROI — led the board to classify the AI programme as "marginally successful." When Layer 2 revenue was included, the first-year return was ($1.2M + $3.6M) / $3.8M = 126%. The characterisation shifted from "marginal" to "one of the highest-returning investments in the company's history."
Layer 3: Competitive Advantage — Estimated $8-12M in enterprise value
The AI capabilities created measurable competitive differentiation. Customer retention for AI-engaged accounts was 94% versus 87% for the industry benchmark. Competitive win rates increased from 34% to 41% in deals where the AI features were demonstrated. The proprietary data asset — 18 months of customer behaviour data across 2,200 accounts — was estimated at $3-5 million replacement cost. The competitive moat value, using a With-and-Without analysis, was estimated at $8-12 million in incremental enterprise value.
Layer 4: Strategic Optionality — Identified but not yet quantified
The AI infrastructure and data assets created options for: a self-service analytics product (estimated $5M TAM), an industry benchmarking service, and an AI-powered marketplace matching customers with complementary solutions. These options were catalogued but not yet valued — a task for the annual strategic review.
Case Study 2: Manufacturing — Predictive Quality and Maintenance
Company Profile
A precision manufacturing company with $120 million revenue, 800 employees, and three production facilities. The company invested $2.1 million over 12 months to deploy AI-driven predictive quality inspection and predictive maintenance across its primary production line.
Layer 1: Cost Reduction — $2.8M per year
The predictive quality system reduced scrap rates from 3.8% to 1.2% of production volume, saving $1.9 million annually in material waste and rework costs. The predictive maintenance system reduced unplanned downtime from 142 hours per year to 38 hours, saving $0.9 million in lost production and emergency maintenance costs.
Layer 2: Revenue Growth — $1.5M per year
Improved quality consistency enabled the company to win two contracts that required defect rates below 1.5% — contracts worth $1.5 million in annual revenue that were previously inaccessible. The faster turnaround from reduced downtime also improved on-time delivery from 91% to 97%, which the sales team attributed to winning three additional competitive bids.
Layer 3: Competitive Advantage — Estimated $4-7M in enterprise value
The company's quality data — over 12 months of sensor readings, defect patterns, and process correlations — created a proprietary data asset that competitors in the sector did not possess. The predictive quality capability became a sales differentiator: the company could offer statistical quality guarantees that competitors could not match. The moat assessment estimated that a competitor would require 18-24 months and $3-4 million to replicate the capability, even with access to equivalent sensor hardware.
The manufacturing case study illustrates a pattern common in traditional industries: AI creates disproportionate value because the competitive baseline is low. In sectors where competitors have not yet deployed AI, even modest AI capabilities create significant differentiation. The competitive advantage is a first-mover benefit that compounds as data accumulates.
Layer 4: Strategic Optionality — $2-4M estimated option value
The AI infrastructure created options for: deploying to the other two production facilities (estimated $1.2M incremental savings each, at 60% of original development cost), offering predictive quality as a service to supply chain partners, and licensing the model to non-competing manufacturers in adjacent sectors.
Case Study 3: Financial Services — AI-Driven Credit Assessment
Company Profile
A specialist lending company with $80 million in annual originations, 65 employees, and a portfolio of 4,200 active loans. The company invested $4.5 million over 24 months to build an AI-driven credit assessment platform replacing its manual underwriting process.
Layer 1: Cost Reduction — $1.8M per year
The AI system processed 82% of applications automatically, reducing underwriting staff from 18 to 7 (with the 11 redeployed to portfolio management and business development). The average time from application to decision dropped from 5.2 days to 4 hours for standard applications. Net cost savings, after AI system operations: $1.8 million per year.
Layer 2: Revenue Growth — $6.2M per year
The faster decision time increased application-to-funded conversion from 38% to 54% — applicants who would have gone to competitors during the 5-day wait now received rapid decisions. The AI model also identified profitable lending opportunities in customer segments that the manual process had systematically rejected as too complex to assess. Together, these effects increased annual originations by $18 million, generating approximately $6.2 million in incremental net interest income.
Layer 3: Competitive Advantage — Estimated $15-22M in enterprise value
The 24 months of credit performance data — covering originations, defaults, prepayments, and recovery patterns — created a proprietary dataset that dramatically improved model accuracy. The company's default rate was 1.8%, compared to a peer average of 3.4%. This risk-adjusted return advantage was the primary value driver in a strategic investor's valuation, which placed the company's AI capability (model plus data) at $15-22 million.
The Acquisition Perspective
When a PE firm evaluated acquiring the lending company, the purchase price allocation analysis separately identified the AI credit model ($8M), the proprietary lending dataset ($6M), the MLOps infrastructure ($2M), and customer relationships enhanced by AI-driven service ($4M) as identifiable intangible assets. The residual goodwill — representing the assembled workforce and organisational capability — was $7M. The total AI-related value represented 62% of the enterprise value.
Layer 4: Strategic Optionality — $5-8M estimated option value
The AI platform created options for: credit-as-a-service (embedding lending decisions in partner platforms), expansion into adjacent lending products using the same model architecture, and a data analytics product for institutional investors seeking alternative credit exposure.
The traditional ROI calculation for the lending company — ($1.8M Layer 1) / ($4.5M investment) = 40% — suggested a respectable but unremarkable return. The 4-Layer analysis revealed a total first-year value creation of approximately $8M across Layers 1-2, with an additional $15-22M in Layer 3 enterprise value and $5-8M in Layer 4 options. The AI programme was not merely a process improvement — it was the foundation of the company's competitive position and the majority of its enterprise value.
Cross-Case Patterns
Three patterns emerge across the case studies.
| Pattern | Observation | Implication |
|---|---|---|
| Layer 1 is the minority | Layer 1 represented 15-25% of total value in all three cases | Organisations measuring only cost savings dramatically undervalue AI |
| Data assets compound | In all cases, the proprietary data created by AI became a strategic asset | AI investment decisions should account for data asset creation |
| Speed matters | Faster processes (underwriting, quality checks, onboarding) drove revenue through conversion and customer experience | Cycle time metrics are revenue metrics, not just efficiency metrics |
What Comes Next
In Lesson 10: From Measurement to Board Strategy, we bring the programme full circle — translating the measurements, frameworks, and case study insights into board-ready narratives and investment cases that secure continued AI funding, align organisational strategy, and position the company for growth.
Ivan Gowan is CEO of Opagio, the growth platform that helps businesses and investors measure, manage, and grow intangible assets. Before founding Opagio, Ivan held senior technology and leadership roles across financial services and digital platforms for 25 years. Meet the team.
Lesson 9 Quiz
4 questions to test what you've learned. Your score contributes to your overall AI Value IQ.
Ivan Gowan is CEO of Opagio, where he builds tools to help companies measure, manage, and grow the intangible assets that drive modern business value.
About the team →David Stroll is a productivity economist and AI researcher advising Opagio on AI value measurement, macro-economic productivity, and intangible asset frameworks.
About the team →Put this knowledge to work
Use Opagio to measure and value the intangible assets your AI investments create.