Hi there,
The hardest part of working with AI is knowing which parts of its answers to trust. That's why we've been working on a solution to empower you to use AI for the work that matters, with confidence.
Today we're excited to announce Reasoning Chains. They show you the logical steps the Count agent takes for every stat or claim it makes, so you can be confident it's right.
This goes well beyond checking SQL or Python. Every sentence the agent writes can be checked for logical consistency, and anything that looks inconsistent or unsupported is flagged for you and your team to review.
No AI gets everything right, including the Count agent. Here are a few things Reasoning Chains have caught for us recently:
- A revenue analysis that drew a confident conclusion from a date the data didn't contain
- A calculation where the agent made up its own methodology
- Research from call transcripts that applied findings about one audience to a different one
To try it, ask the Count agent a question. Claims in its answer are underlined: click any of them to see the chain behind it, or open the lineage panel on the left to review every flag in one place.
It's available now on all plans. You can read the story behind Reasoning Chains on our blog, or find more detail in the docs. I'd love to hear what you think, so just reply to this email.
Ollie
CEO, Count
|