I’d model the asset as a commercial HVAC system in Saudi Arabia, with an AI/LLM layer processing 10,000 maintenance-related requests per day.
For the token price, I’ll use an illustrative $0.002 per 1K input tokens plus $0.008 per 1K output tokens. This provides a transparent model without tying the example to a particular vendor.
| Variable | Assumption |
|---|---|
| Asset | Commercial HVAC system |
| Location | Saudi Arabia |
| Requests | 10,000 per day |
| Days per year | 365 |
| Requests per year | 3,650,000 |
| Input tokens per request | 1,500 |
| Output tokens per request | 500 |
| Input price | $0.002 per 1K tokens |
| Output price | $0.008 per 1K tokens |
| HVAC operating life | 15 years |
| Analysis horizon | 10 years |
The cost per request is calculated as follows:
Costrequest =
(1,500 / 1,000 × $0.002) +
(500 / 1,000 × $0.008)
Costrequest = $0.003 + $0.004 = $0.007 per request
At 10,000 requests per day:
10,000 × $0.007 = $70 per day
Annual AI processing cost:
3,650,000 × $0.007 = $25,550 per year
Tokens × price × volume represent only the AI operating cost. The asset-management model should connect that cost to the HVAC system’s physical operating decisions.
| Variable | Assumption |
|---|---|
| Existing HVAC replacement cost | SAR 500,000 |
| Existing HVAC annual maintenance | SAR 80,000 |
| Reduction in unexpected failures | 15% |
| Reduction in HVAC energy consumption | 5% |
| HVAC annual electricity cost | SAR 600,000 |
| AI processing cost | Approximately SAR 95,800 per year using an illustrative SAR/USD rate of 3.75 |
The annual AI cost is therefore:
$25,550 × 3.75 = SAR 95,813 per year
Maintenance savings:
SAR 80,000 × 15% = SAR 12,000
Energy savings:
SAR 600,000 × 5% = SAR 30,000
Total quantified annual benefit:
TotalBenefit = SAR 12,000 + SAR 30,000 = SAR 42,000
In this scenario, the AI operating cost is greater than the directly quantified savings, so the project would not yet justify itself.
This is useful for the model because it forces the analysis to identify additional value from avoided HVAC downtime, emergency repairs, and asset life extension.
The decision should be modeled over time. Suppose the existing HVAC system is becoming increasingly unreliable:
| Year | HVAC age | Maintenance | Failure cost | Energy cost | Decision |
|---|---|---|---|---|---|
| 1 | 10 years | SAR 80,000 | SAR 30,000 | SAR 600,000 | Operate |
| 2 | 11 years | SAR 90,000 | SAR 40,000 | SAR 615,000 | Operate |
| 3 | 12 years | SAR 105,000 | SAR 60,000 | SAR 635,000 | Evaluate |
| 4 | 13 years | SAR 125,000 | SAR 90,000 | SAR 660,000 | Evaluate |
| 5 | 14 years | SAR 150,000 | SAR 130,000 | SAR 690,000 | Replace |
The annual cost of continuing to operate the existing HVAC system is:
Costoperate,t =
Maintenancet +
Failuret +
Energyt +
AIt
The annual cost after replacement is:
Costreplace,t =
ReplacementCost +
Maintenancet,new +
Failuret,new +
Energyt,new +
AIt
The model then calculates the cumulative 10-year cost of operating versus replacing the HVAC system.
The 10,000 daily requests are not the asset itself. They are the observations and events that feed the asset-management model.
A suitable architecture is:
10K daily transactions →
time-series features →
HVAC condition →
predicted failure, energy, and maintenance cost →
operate-versus-replace decision →
NPV decision
For each period, the AI processing cost can be represented as:
AI Costt =
Tokenst ×
Pricet ×
Volumet
The actual asset decision is then based on:
NPV(Operate) versus NPV(Replace)
Illustrative model only; the figures are not measured Saudi HVAC cost data.