Sustainability

Measured returns. Measured footprint.

AI runs on electricity and water. We measure what your AI uses, cut what it wastes, and report it next to the financial return, so you can make the case for AI with the full picture.

Why it matters

Your share is small. Spend it well.

Data centers used about 415 terawatt-hours of electricity in 2024, and the International Energy Agency expects that to more than double by 2030.1 Any one company's part of that is small, which is a reason to spend it carefully, not to ignore it.

Customers, lenders and larger buyers are also starting to ask what your AI uses. That question is easy to answer when measurement is built in from day one, and hard to answer after the fact.

What we measure

Three numbers, each with a range and a source.

Electricity

Estimated kilowatt-hours for each kind of AI work: quick lookups, chat questions, document processing and multi-step AI agents.

Emissions

That electricity converted to carbon emissions (CO2e) using published US grid figures from the EPA.5

Water

Cooling water at the data center, plus the water used to generate its power, which is often more than ten times the cooling water.6

Published estimates differ by up to 38 times depending on what each AI provider counts.2,3 That is why we report a low, typical and high figure, show where each one comes from, and update them when your provider publishes its own.

What drives the number

How you use AI matters more than how many people use it.

A single task run by an AI agent can use hundreds of times the energy of a chat question.2,4 In our benchmarks, agent work accounts for most of an organization's AI electricity at every size.

84–92%

Share of AI electricity from agent work in our small, mid-size and larger organization benchmarks.8

How we reduce it

  • Use the smallest AI model that does the job well.
  • Filter out spam and duplicates before they reach a model.
  • Let the AI act only when it is confident, and hand the rest to a person.
  • Reuse work already done instead of repeating it.
  • Choose AI providers that publish their energy and water data.
  • Recommend against AI where the return doesn't justify the cost.

Worked example

A typical engagement, both sides of the ledger.

The same client as our ROI example: 12 people spending 10 hours a week on repetitive work, 40% of it automated with one Sprint and one Automation Build.

Each year8

2,304

hours returned to the team

~220 kWh

of AI electricity, with a range of 65 to 890 kWh

~76 kg

of CO2e, about the same as driving a car 190 miles7

~95 kWh

of AI electricity for every 1,000 hours returned

We report returned hours as time, not as an energy saving, because people keep working. A resource saving counts only when something is switched off, retired or no longer made, such as a server, a software tool, a trip or a reprint. We measure those from your Sprint baseline.

Benchmarks by size

What AI typically uses in a year.

OrganizationElectricityEmissionsSame CO2 as driving
Small (under 50 staff)~210 kWh~75 kg CO2e~190 miles
Mid-size (50 to 500)~5,500 kWh~1.9 t CO2e~4,800 miles
Larger (500+)~70,000 kWh~24.6 t CO2e~61,000 miles

Typical estimates from illustrative usage profiles that mix chat, document processing and AI agents. Your own figures depend on what you run and where.8

Included in every engagement

Built into the work, at no extra cost.

AI Opportunity Sprint

A footprint baseline, and every opportunity ranked by its return and its resource cost.

Data & KPI Foundation

Energy, emissions and water on your dashboard, per unit of finished work.

Automation Build

Built to waste less from the start, with a footprint report at handoff.

Fractional AI & Product Lead

A quarterly footprint report, and the right questions asked in every AI purchase.

Already using AI?

AI Footprint Check

1–2 weeks · Fixed fee

We measure what your current AI tools use, find where they waste it, and give you a short report you can share with customers, lenders or your board.

The full fee is credited toward an AI Opportunity Sprint signed within 60 days.

$3,500

  • Inventory of your AI tools and usage
  • Electricity, emissions and water estimates, with ranges
  • Waste reductions ranked by effort and impact

Our position

We use AI to run this business too. We don't call AI green. We measure it, use it where it earns its place, and tell you plainly what it costs.

Want your AI footprint measured?

Book a 30-minute call. We'll tell you whether a Footprint Check or a Sprint is the better place to start.

Book a 30-minute call

Sources

  1. International Energy Agency, Energy and AI: Executive summary (2025). Data center electricity use in 2024 and the outlook to 2030.
  2. Google Cloud, Measuring the environmental impact of AI inference (2025). Energy, emissions and water for a typical Gemini text prompt.
  3. Mistral AI, Our contribution to a global environmental standard for AI (2025). Lifecycle emissions and water per response, including hardware manufacturing.
  4. Zeke Hausfather, The real energy use of agentic AI (2026), citing Watershed's corporate AI emissions framework (Bistline et al., 2026). Energy per AI agent task compared with a chat prompt.
  5. US Environmental Protection Agency, eGRID 2023 summary data. US average grid emissions rate.
  6. Lawrence Berkeley National Laboratory, 2024 United States Data Center Energy Usage Report, as summarized by Latitude Media (2026). Water used to generate data center power compared with on-site cooling water.
  7. US Environmental Protection Agency, Greenhouse gas emissions from a typical passenger vehicle. About 400 g of CO2 per mile.
  8. BloomRail AI Footprint Benchmarks (October 2026). BloomRail estimates built from sources 2 to 7, with low, typical and high cases. Methodology available on request.

Figures are estimates, not guarantees.