Know the carbon cost of every AI call
SAVI estimates the carbon footprint of each AI call from published model energy figures and Australian grid intensity, so you can see it by team and by model.
AI compute belongs in your emissions picture
As more work moves to AI, so does a share of your electricity use. Provider invoices show spend, not carbon. SAVI adds an estimate by call, team and model, clearly labelled as an estimate, so AI can be part of the numbers you already report on.
Climate reporting is growing
Many organisations now report on climate. AI is a growing part of electricity use, and it helps to have a documented way to estimate it.
Models differ in energy use
In SAVI's current factors, a call to a large model uses around four times the energy per token of a small one. The gap varies by model, so a per-call view beats a rule of thumb.
Wasteful prompts cost money and energy
A prompt that carries far more text than the task needs costs more and uses more energy. SAVI points these out so engineering can tighten them.
A carbon estimate on every AI call
Every AI call that passes through SAVI gets a carbon estimate as it arrives. SAVI combines the tokens used with published energy figures for the model and with Australian grid intensity, refreshed regularly. The energy figures are reviewed each year as models and hardware change. These are estimates, not meter readings, and SAVI labels them that way.
- Worked out as each call arrives, with no separate data job
- Model energy figures reviewed each year
- Based on published research, and labelled as an estimate
How each estimate is built
Swipe to see all columns →
Carbon views available. Figures are estimates.
See carbon by team and by model
SAVI shows carbon on the same footing as cost, by team and by model. Finance and sustainability teams can see which teams and models drive the most. Procurement can weigh carbon next to price when choosing a model.
- Carbon by team, month by month
- A model comparison to support choosing between options
- A 13-month trend you can use in your own reporting
Spot prompts that waste tokens
When a workload uses far more tokens than its task needs, SAVI points it out with an estimate of the savings. Fewer tokens means a smaller bill and a smaller footprint.
- Wasteful prompts surfaced for engineering to review
- An estimated saving shown with each recommendation
- Fewer tokens per task lowers cost and energy together
From insight to saving
Could you include AI in your climate numbers today?
5 questions. Answer honestly.
Do you know how much carbon your AI usage produced last quarter?
Can you attribute AI carbon to specific teams and workloads, not just a total?
If your board asked for AI climate data today, could you produce it without a manual exercise?
Can you compare the carbon footprint of different models running the same task?
Do you know which workloads are your biggest emitters, and can you act on that?
This checks one slice of your AI estate. Take the full AI Cost Ownership Assessment →
Your AI gap score
Answer to begin
Complete all questions to see your result
See the carbon cost of your AI.
SAVI estimates carbon on every AI call and shows it by team and by model. Book a demo to see yours.