What changed
Channing Ferrer, Brevo’s Americas CEO and head of revenue, argues that SaaS companies need to treat AI responsibility as an operating-cost and governance problem. Every prompt, recommendation and automated workflow carries compute costs, while the infrastructure behind them consumes energy and water. The practical levers are model choice, retrieval architecture and caching.
Why it matters
AI can quietly change the economics of a SaaS product. Subscription pricing assumes relatively stable costs, but a few heavy users can create a disproportionate computing load. That may push companies toward usage-based pricing or restricted access to compute-intensive features.
The pressure is not only financial. Data centers accounted for about 1.5% of global electricity consumption in 2024, and the International Energy Agency projects their electricity use could approach 945 terawatt-hours by 2030, just under 3% of global electricity use. Water demand also varies sharply by location, infrastructure and cooling system; reported water usage effectiveness can reach 2 liters per kilowatt-hour.
The useful shift is to measure AI like any other core infrastructure: energy per request, carbon per request, system efficiency and cost by customer segment. Smaller models may handle many jobs adequately. Retrieval can prevent systems from recomputing information they already have. Caching can stop repeated actions from triggering full processing.
Responsible AI in SaaS is therefore less a slogan than a bookkeeping discipline. It exists when operators can see what each feature costs and govern it accordingly. Without that visibility, responsibility is mostly an intention.
Comments
No comments yet.