The Right Question Is Not “How Much Do We Get Back from Investing in AI?”
Most executives begin with the question, “If we invest in AI, how much will we get back?” It sounds reasonable, but it starts from the wrong premise.
A more powerful question is “If we do not invest in AI today, what cost will the organization continue to pay over the next 3 years?”
The cost of doing nothing (Cost of Inaction) is often overlooked—whether it is employees still doing repetitive manual work, fragmented data slowing down decision-making, or competitors that have already adopted AI and are widening the competitive gap every day.
This article presents an AI ROI calculation framework that Thai enterprise executives can apply in practice, covering costs, returns, and the traps to watch out for.
The 4 Layers of AI Cost You Must Account For
AI investment is not just a monthly license fee. The real cost consists of 4 layers.
Layer 1: License and Subscription
The most visible expenses include platform licenses, API usage, cloud infrastructure that supports AI workloads, and model costs if you use services from external providers. This layer is typically presented in the proposal and is the first number executives see.
Layer 2: Implementation and Integration
These are the costs of deploying and connecting AI systems to existing systems, whether ERP, CRM, or various legacy systems. This phase often takes 30–50% more time and budget than expected, especially in organizations with multiple older systems that were never designed to integrate with one another.
Layer 3: Data Preparation and Training
AI performs only as well as the data fed into it. Preparing clean, complete, AI-ready data is a hidden cost that is often underestimated. This also includes training the model to understand the organization’s specific context, which may take several months.
Layer 4: Change Management and Ongoing Maintenance
The most frequently forgotten costs include employee training, workflow redesign, resistance-to-change management, and long-term system maintenance. This also includes monitoring AI model performance and retraining when data or business context changes.
The 3 Types of Returns You Need to Distinguish
Type 1: Hard Savings — Returns That Can Be Measured Immediately
- Reduced labor hours — A manual task that used to take 100 hours/month may drop to 20 hours
- Fewer errors — A lower error rate translates into reduced correction costs
- Lower resource consumption — Paper, storage space, or legacy systems that can be retired
Type 2: Revenue Uplift — Returns from Increased Revenue
- Higher sales — AI helps personalize offers and improve conversion rates
- Lower customer churn — Predictive analytics can identify at-risk customers before it is too late
- New revenue streams — Products or services made possible by AI
Type 3: Strategic Value — Strategic Benefits That Are Harder to Quantify
- Competitive advantage — Capabilities competitors do not yet have
- Speed to market — Faster launch of new products or services
- Innovation capacity — Organizations that use AI effectively can innovate faster
- Talent attraction — AI-enabled organizations attract top talent more easily
Although Type 3 returns are difficult to quantify, they often determine whether an organization can remain competitive in the long term.
The Basic ROI Formula and a Real Calculation Example
The basic ROI formula is:
ROI (%) = (Total Gains - Total Costs) / Total Costs x 100
Calculation example
Assume an organization invests in AI to automate the invoice processing workflow:
- Total Costs (Year 1): THB 2,400,000 (license 600,000 + implementation 1,200,000 + training 300,000 + change management 300,000)
- Total Gains (Year 1): THB 3,600,000 (reduced staff time 2,400,000 + reduced error cost 600,000 + reduced late payment penalties 600,000)
Year 1 ROI = (3,600,000 - 2,400,000) / 2,400,000 x 100 = 50%
From Year 2 onward, costs decline to mainly license and maintenance (approximately THB 900,000/year), while gains remain stable or increase, resulting in significantly higher cumulative ROI.
Case Study: A 500-Person Organization Investing in AI Workflow Automation
Consider a mid-sized organization with 500 employees that decides to invest in AI Workflow Automation covering document processing, approval workflows, and reporting.
Total cost over 3 years
| Item |
Year 1 |
Year 2 |
Year 3 |
3-Year Total |
| License/Subscription |
1,200,000 |
1,200,000 |
1,200,000 |
3,600,000 |
| Implementation/Integration |
2,400,000 |
0 |
0 |
2,400,000 |
| Data Preparation |
800,000 |
200,000 |
200,000 |
1,200,000 |
| Change Management/Maintenance |
600,000 |
400,000 |
400,000 |
1,400,000 |
| Total Cost |
5,000,000 |
1,800,000 |
1,800,000 |
8,600,000 |
Total returns over 3 years
| Item |
Year 1 |
Year 2 |
Year 3 |
3-Year Total |
| Reduced labor hours (Hard Savings) |
2,000,000 |
3,500,000 |
4,000,000 |
9,500,000 |
| Reduced Error/Rework Cost |
500,000 |
800,000 |
1,000,000 |
2,300,000 |
| Revenue Uplift (faster decision-making) |
500,000 |
1,200,000 |
2,000,000 |
3,700,000 |
| Total Returns |
3,000,000 |
5,500,000 |
7,000,000 |
15,500,000 |
ROI summary
- Year 1 ROI: -40% (not yet at break-even because implementation costs are high)
- Cumulative ROI by Year 2: +25% (returns begin to become clear)
- Cumulative ROI by Year 3: +80% (cumulative returns of THB 15,500,000 versus cumulative costs of THB 8,600,000)
- Payback period: approximately 18 months
This shows that Year 1 is often not profitable. Executives who assess ROI based only on the first year will make the wrong decision almost every time.
ROI Traps to Watch Out For
Trap 1: Counting Only Direct Savings and Forgetting Indirect Costs
Many organizations calculate ROI by counting only license fees and comparing them to labor savings, while ignoring Layers 2–4 of cost. This makes the ROI presented to the board look unrealistically attractive. When the real costs appear later, confidence in the project declines rapidly.
Trap 2: Ignoring the Opportunity Cost of Teams That Must Support the Project
When AI is introduced, IT and business teams must divert time from their normal work to support the project. That lost time is an opportunity cost that does not appear in the budget, but it affects the organization’s overall productivity. It must be included in the equation.
Trap 3: Using Too Short a Time Horizon for ROI
AI is not software that you install today and benefit from tomorrow. Systems need time to learn, employees need time to adapt, and processes need to change. AI ROI should be evaluated over at least 2–3 years. Organizations that cancel AI projects before the end of the first year because ROI is not yet positive are abandoning investments that are often just about to become worthwhile.
A 5-Step Framework for Calculating AI ROI
Step 1: Define Scope — Set Clear Boundaries
Clearly specify which processes AI will be applied to. Do not start too broadly. Select 1–2 processes with available data and high impact, such as document processing, customer service, or quality inspection.
Step 2: Baseline Current Costs — Measure the Current Baseline
Before calculating how much AI can save, you must know how much the current process costs. Measure all relevant factors: labor cost, time cost, error cost, and opportunity cost. Collect data for at least 3 months to establish a reliable baseline.
Step 3: Model AI Impact — Estimate the Impact
Create a model with 3 scenarios:
- Conservative: 20–30% improvement over baseline
- Moderate: 40–60% improvement over baseline
- Optimistic: 70%+ improvement over baseline
Use the conservative scenario as the basis for decision-making. If the project is still worthwhile under the conservative scenario, the risk is relatively low.
Step 4: Calculate TCO — Calculate Total Cost of Ownership
Include all 4 cost layers mentioned above over a 3-year period. Do not forget to include:
- Inflation in license pricing that may increase each year
- Scaling costs as usage expands
- Downtime costs during the transition period
Step 5: Decision Matrix — Evaluate Holistically
Do not make the decision based on a single ROI number. Use a decision matrix that covers multiple dimensions:
| Criteria |
Weight |
Score (1-5) |
Weighted Score |
| 3-Year ROI (Conservative) |
30% |
? |
? |
| Strategic Alignment |
25% |
? |
? |
| Implementation Risk |
20% |
? |
? |
| Organizational Readiness |
15% |
? |
? |
| Vendor Reliability |
10% |
? |
? |
Projects with a total weighted score above 3.5 out of 5 are generally strong candidates for investment.
Summary: Start by Understanding Your Organization First
Calculating AI ROI is not just about comparing cost and return figures. It is a process that helps executives understand the organization’s current state, identify realistic opportunities, and make informed decisions backed by data.
What you should do now:
- Assess organizational readiness — Start with the AI Readiness Assessment to understand where your organization stands and where to begin
- Pilot AI on a small process first — Try using the Genesis AI Platform on one process, measure the actual results, and then scale
- Consult experts — If you need a framework tailored to your organization’s specific context, contact Enersys to discuss AI strategy and ROI calculation suited to your business
Successful AI investment does not begin with technology. It begins with a correct understanding of the true costs and returns. The framework presented in this article gives executives a practical tool for making confident decisions. Whether you choose to invest now or wait, let it be a decision grounded in evidence—not guesswork.
