Here's a typical workflow customers perform with Athena:
- Research market metrics,
- Compile a 6-year sizing estimate,
- Compile TAM-SAM-SOM for a new player.
This requires some calculations, and that's where code executor comes in.
The alternatives are:
- LLM math (GPT4o is very good at it, but the idea is horrid)
- Math tool The reason why a math tool wouldn't really cut it is, well, it can only do math.
We see our customers throw at Athena things we wouldn't think of — and didn't know it could do these! So having a higher capability ceiling is a no-brainer.
So there it is — a case for code executor as a no-brainer alternative to both LLM math and a simple math tool.
As a bonus, once you start going down the code-assisted workflows, the scope of agent capabilities grows really quickly. More on that in later posts!