Okta Measures and Offsets AI Energy Use With Custom Methodology
Identity management software firm Okta adopted a strategy in April to manage the energy consumption of artificial intelligence tools used by its employees. The approach emerged from a two-year evaluation of emerging technologies, prompted partly by questions from staff members.
Okta's sustainability team partnered with engineering and IT colleagues to develop the program. The collaboration ensures sustainability decisions integrate into how the company manages its technology relationships rather than existing as a separate reporting function.
Building a measurement system from scratch
Okta created its own methodology to calculate emissions tied to its top AI tools because vendors rarely provide this data. The company evaluated how many lines of code were generated using AI, then converted that to "tokens"-blocks of data the model processed-to estimate carbon footprint.
The calculation factors in additional AI model training, employee locations, and data center locations. Okta treats this as a temporary approach until industry standards solidify, and it's sharing the methodology to accelerate that development.
One challenge emerged quickly: AI vendors resist disclosing granular energy data. Another lesson: productivity tools used daily require different measurement approaches than AI services built for specific projects. Each type offers different opportunities to reduce environmental impact.
Guiding employee decisions
Okta compiled best practices for staff deciding whether to use AI. The guidance recommends limiting AI use to projects where it delivers clear efficiency gains and selecting the lowest-energy model for the task.
The company uses OpenWebUI, which ranks AI tools by energy efficiency and compute requirements. Okta's sustainability team built an automated agent that suggests better options when employees select less efficient models, flagging alternatives at the moment of use.
Offsetting consumption
Okta plans to estimate AI-specific energy consumption annually. Based on initial measurements, the company expanded renewable energy certificate purchases to cover the additional power needed for its two most-used third-party AI tools.
The company won't include these offsets in its greenhouse gas emissions inventory yet. The methodology remains too new for that level of official accounting, though the company expects to move toward real-time tracking that enables faster efficiency improvements.
For managers implementing AI for Management strategies, Okta's approach demonstrates how to build accountability into technology adoption decisions. The company will present additional details at the AI x Sustainability program at Trellis Impact 26, running June 23-25 in San Francisco.
Your membership also unlocks:
