← This weekCase studyAI platforms — data science & MLdbtGong's API was used extensively to feed AI models, leading to high token costs. By modeling the transcripts in a data warehouse using dbt, the token costs were reduced by 20 times.
MyDataWork POV — The transition to modeling for tokens instead of tables highlights a troubling oversight. While reducing token costs with dbt is impressive, it reveals a deeper issue: the reliance on APIs without considering long-term cost implications. This approach could lead to unsustainable practices where short-term savings mask the potential for runaway expenses. Organizations must scrutinize their AI data pipelines to ensure they prioritize strategic foresight over mere cost reductions.
Discussion happens on Reddit — no comments are hosted here.