Virtual Event • August 13 • 12 PM PDT
How do we know if AI is actually good at data analytics?
Benchmarks are the obvious place to look. But writing SQL against a public dataset is nothing like trusting AI with a real analysis.
The best data analysts are skeptical. They pick the right aggregation levels, catch the confounds, and know when the data can't answer the question at all.
We built a benchmark that grades models on analytical judgment and we score cost and latency too, since "right answer" and "worth what it costs" aren't the same thing.
Join us live on August 13 at 12 p.m. PDT, when Izzy Miller (AI Engineer) and Rachel Herrera (Product Evangelist) share what we learned building a new way to test whether AI can actually be trusted with data analysis.
We'll walk through:
- What "trustworthy AI analysis" actually looks like
- A few examples of how we’re putting analytical judgement to the test
- Where today’s models land
- What it means for how you evaluate and roll out AI on your own team