Measuring the Business Value of AI Products

Why Traditional SaaS Business Cases Are No Longer Enough

“Our AI maintenance copilot reduces troubleshooting time by 60%.”

Such proclamations are common in executive meetings and vendor presentations, but when the CFO asks, “What is my return on investment?” vague assurances are not sufficient.

Organizations have long used financial frameworks to understand the value of traditional software products. These methods quantify the intangible benefits of software such as reduced risk exposure or improved regulatory compliance while factoring in tangible benefits like implementation costs and productivity gains.

Return on Investment (ROI) measures the value an investment generates over time, while the Payback Period (PBP) indicates how long it takes to recover the initial investment. Both are useful metrics that have been applied to traditional software products/SaaS with a degree of success.

Why? Because such products tend to operate in a largely deterministic manner - once implemented, they behave in more or less predictable ways, producing consistent results. As a result, the value these systems deliver tends to be more-or-less straightforward to estimate, distribute throughout the organization, and realize.

Realized Business Value (RBV)=Potential Business Value (PBV)×Adoption\text{Realized Business Value (RBV)} = \text{Potential Business Value (PBV)} \times \text{Adoption}

Where:

Potential Business Value (PBV) is the aggregate value that the implemented solution can deliver to the organization. It can include both tangible elements (efficiency gains, cost reductions, revenue growth opportunities) and intangible elements (reduced risk exposure, improved quality, enhanced customer experience, etc.).

Adoption refers to the percentage of the target audience/workflow that uses the solution for its intended purpose.

AI products, however, are probabilistic and non-deterministic, introducing greater uncertainty in both outcomes and realized value. Unlike traditional software, they don’t just execute predefined business logic—they work alongside people. An AI maintenance copilot helps engineers troubleshoot equipment, while an AI visual inspection system identifies potential defects for human review. Rather than replacing people, these systems augment their capabilities.

This changes how value is created. Traditional software delivers value by automating and standardizing well-defined workflows. AI is different. It creates value by helping people make better decisions, work faster, and take on more complex tasks. But organizations rarely give AI full autonomy from day one. They start with high human control and low AI agency, using AI to assist rather than act. As trust in the system grows, the AI gradually takes on more responsibility. That means the value of an AI product depends not just on whether it works, but on how much of the job it actually does. That’s something traditional ROI models were never designed to measure.

The Job Completion Factor (JCF)

The missing piece is how much of the work the AI actually does. Let’s call this the Job Completion Factor (JCF) – the percentage of a business workflow that AI can meaningfully complete.

Note that JCF should not be conflated with accuracy – the percentage of correct answers in an AI model. A given AI model may have a high accuracy rate, but only automate a small portion of a given business process, or the reverse may be true. Analysts should examine how much work the AI can complete automatically while being able to deliver acceptable results within regulatory and quality control constraints.

Consider an example use case and scenario: a maintenance engineer utilizes an AI-based copilot to troubleshoot equipment.

ActivityAI Contribution
Retrieve manuals100%
Analyse sensor trends90%
Identify likely fault70%
Recommend corrective actions50%
Final engineering decision10%

Looking at the overall business process, we see that the AI completes approximately 70% of the work, leaving 30% that requires human judgment. The appropriate JCF for this scenario would be 70%.

This leads us to an expanded definition of realized business value:

Realized Business Value (RBV)=Potential Business Value (PBV)×Job Completion Factor (JCF)×Adoption\text{Realized Business Value (RBV)} = \text{Potential Business Value (PBV)} \times \text{Job Completion Factor (JCF)} \times \text{Adoption}

Where:

  • Potential Business Value (PBV) remains the same as for traditional software—it represents the total value the solution could create if it were fully deployed and used as intended.
  • Job Completion Factor (JCF) measures the percentage of the business workflow that the AI can reliably complete while meeting the required quality, safety, and compliance standards.
  • Adoption measures the percentage of target users or workflows that actively use the AI solution.

In essence, the three factors outline three basic questions:

  • PBV answers “How valuable is solving this problem?”
  • JCF answers “How much of the work can the AI actually do?”
  • Adoption answers “How many people are actually using it?”

Traditional software tends to focus on the first two factors, but AI-based solutions require that analysts also consider adoption and utilization.

Case Study: AI Maintenance Copilot

A manufacturing plant running 250 assets, already using a CMMS, evaluates an AI Maintenance Copilot. Estimated annual potential benefits — reduced troubleshooting time, reduced downtime, faster onboarding, better knowledge retention — total $1.2 million per year, split between tangible and intangible value:

BenefitTypeAnnual Value
Reduced troubleshooting timeTangible$480,000
Reduced downtimeTangible$540,000
Faster onboardingIntangible$100,000
Improved knowledge retentionIntangible$80,000
Total Potential Business Value$1,200,000

Operational testing shows the AI meaningfully completes 65% of the troubleshooting workflow. Six months in, 80% of engineers actively use it.

Realized Business Value=$1.2M×65%×80%=$624,000 annually\text{Realized Business Value} = \$1.2M \times 65\% \times 80\% = \textbf{\$624{,}000 annually}

That’s a far more realistic basis than assuming the full $1.2 million lands the moment the tool goes live — but a CFO calculating ROI and Payback Period wants the harder number: tangible value only, since intangibles are proxy-based and don’t belong on a P&L line.

Gain from Cost Saving (Monthly)

ItemAmount ($)
Troubleshooting time saved40,000
Downtime avoided45,000
Total85,000

Investment Needed for One Year

ItemAmount ($)
DataOps development cost80,000
Training cost60,000
Integration cost160,000
Subscription cost ($10,000/month × 12)120,000
Total420,000

Realized Monthly Gains = $85,000 × 65% × 80% = $44,200.

From these tables, a back-of-the-envelope yearly ROI:

One Year ROI=One Year GainsOne Year InvestmentOne Year Investment×100%\text{One Year ROI} = \dfrac{\text{One Year Gains} - \text{One Year Investment}}{\text{One Year Investment}} \times 100\%

One Year ROI=($44,200×12)$420,000$420,000×100%26%\text{One Year ROI} = \dfrac{(\$44{,}200 \times 12) - \$420{,}000}{\$420{,}000} \times 100\% \approx \textbf{26\%}

A 26% one-year ROI is solid — not the “fantastic” number a vendor pitching the undiscounted $1.02M tangible potential would show, because these Gains already carry the JCF discount.

Now the payback period:

PBP=Total InvestmentOne Month GainsPBP = \dfrac{\text{Total Investment}}{\text{One Month Gains}}

PBP=DataOps Cost+Training Cost+Integration CostMonthly Cost SavingMonthly SubscriptionPBP = \dfrac{\text{DataOps Cost} + \text{Training Cost} + \text{Integration Cost}}{\text{Monthly Cost Saving} - \text{Monthly Subscription}}

PBP=$80,000+$60,000+$160,000$44,200$10,0008.8 monthsPBP = \dfrac{\$80{,}000 + \$60{,}000 + \$160{,}000}{\$44{,}200 - \$10{,}000} \approx \textbf{8.8 months}

A 26% ROI and a nine-month payback are numbers a CFO can actually underwrite. This is how an AI company should justify its tangible value. Once the tangible case is made, the intangible benefits — faster onboarding, better knowledge retention — get pitched on top.

Conclusion

Traditional SaaS and AI products shouldn’t be evaluated on identical assumptions. Traditional software delivers value by digitizing a process — once adopted, it performs its function consistently and predictably. AI products are fundamentally different: they collaborate with people, automate only part of a workflow, and rely on human judgment for the rest.

That changes the question executives should ask. Not “how accurate is the model?” but: “What percentage of the business job can this AI product meaningfully complete?”

Combine the tangible and intangible benefits organizations already know how to estimate with the Job Completion Factor, and you get more realistic business cases and better decisions about where AI actually creates value. AI rarely replaces the job — it completes part of it. The value it creates should be measured accordingly.