AI-Enabled Commercial HVAC Asset-Management Model

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.

1. Base assumptions

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

Cost per request

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

Daily and annual AI processing cost

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

2. Connecting AI to HVAC asset management

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.

Additional assumptions

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

Potential annual benefits

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.

3. Operate versus replace

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

Operating-cost formula

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.

4. Asset-management architecture

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

Per-request AI cost

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.