I am building an Amazon Quick Agent on top of space with Topic which can answer data calculation questions. It answers high level questions well. but for slightly more complex questions- I have only been able to get some decent results when I explicitly model the possible question themes/intents in the backend. It feels pretty non-scalable as the scope grows. Any suggestion how to make my agent smarter to handle complex data manipulation questions.
Hi @komal_sharma,
Good to see you back in the community!
There are a few things you can do to make it handle more complexity without manually modeling every intent. Adding synonyms helps the agent understand different ways users phrase things. Pre-computing complex metrics as calculated fields at the dataset level (ratios, YoY changes, running totals, etc) means the agent can reference them directly without needing to figure out the math. Adding descriptions to your fields gives the agent context about what each one means and how it should be used.
You can also use your Agent’s reference documents to include instructions on how to approach complex calculation, for example you can have something like, “When asked about margin, calculate as (revenue - cost) / revenue.” This gives the agent explicit guidance without needing to model every possible question pattern. Combining a topic for structured queries with a space containing business logic documentation lets the agent pull from both to answer more nuanced questions.
Hope this helps and feel free to reply with any updates/additional questions!
Hi @komal_sharma,
Just checking back in since we haven’t heard from you in a bit. I wanted to see if the guidance shared earlier helped resolve your question, or if you found a solution in the meantime.
If you still have any additional questions related to your initial post, feel free to share them. Otherwise, any update you’re able to provide within the next 3 business days would be helpful for the community.
Thank you!