5b. Take a green approach to AI tools used by the team
AI tools can support productivity for teams.
For example, AI-augmented development can help software engineers design, code and test applications. This can be an a cost effective way to help teams build software faster.
Sub-actions
5b. (i) Use AI tools only when appropriate
5b. (ii) Quantify the environmental impact of AI tools used
5b. (iii) Run AI ‘pipeline’ tasks only once where possible
(i) Use AI tools only when appropriate
Choose the right AI tool for the problem to be solved. Check that non AI approaches have been considered before selecting AI tools.
Environmental benefit:
Avoiding the use of AI tools when these are not necessary or the best tool for the job can save on resource consumption.
Read more
(ii) Quantify the environmental impact of AI tools used
Tools for measuring the environmental impacts of AI with confidence are in their infancy. A good starting point is tracking token use and expenditure per transaction or user task.
Environmental benefit:
Estimating the impacts of AI tools helps inform decisions on their appropriate use.
Read more
(iii) Run AI pipeline tasks only once where possible
Pipeline tasks include data collection and preprocessing, model training and inference. Aim to run tasks once and reuse results, cache intermediate results and store processed data.
Run tasks at times of day with lower carbon intensity electricity where possible.
Environmental benefit:
Running tasksing with lower-carbon electricity reduces carbon footprint.