my first project how about the efeciency is that quite alright or something?
Hey @SieghartSaladdin Nice work on your first project, the flow looks clean and the logic is easy to follow.
Efficiency wise, it’s already pretty solid. The main thing I’d look at is reducing extra AI calls where possible, since those are usually the slowest and most expensive steps. You might also be able to move some of the formatting into a Set/Function node to simplify the flow a bit.
Other than that, it’s a good start! If you share what the workflow is meant to achieve, happy to suggest more targeted improvements.
Thank you, @Anshul_Namdev for sharing your thoughts on my workflow. I appreciate your feedback.
If you’re interested in understanding the flow, here’s the idea behind it:
I initially built it just for fun, but my main goal is to create a system that can generate financial reports for me and also maintain good chatbot memory. The chatbot should understand simple inputs like:
“Hey chat, I just had dinner for 25k, please add it to my expenses,”
and automatically record it.
1 Like
Thanks for sharing more about your idea, @SieghartSaladdin , honestly, it’s a really cool concept, and pretty ambitious for a “first fun project”! 
A conversational expense tracker that can understand something like “I just had dinner for 25k, add it to my expenses” is a perfect mix of AI + n8n:
-> The AI part can interpret the messy, natural language input (amount, category, date, notes, etc.)
---
-> n8n can then take that clean structure and reliably save it to your sheet or database.
A few things you might find helpful as you build this out:
→ Keep “understanding” and “saving” separate
Have the AI step only extract clean JSON, something like { amount, category, date, note }.
Then let your Set/Function nodes handle formatting, and your Sheets/Airtable node handle storage.
This makes the workflow much easier to maintain and debug.
→ Use scoped memory
If you want the bot to remember context (like “that dinner yesterday…”), you can use a memory node with a session/user key. That way each conversation has its own memory without mixing things up.
→Add reports later on
Once your data is nicely structured, you can spin up a second workflow (maybe on a Cron trigger) to send yourself daily/weekly/monthly summaries through chat or email.
You’re absolutely heading in the right direction already.
If you’re comfortable sharing a (redacted) screenshot or the workflow JSON of the updated workflow which you are building, I’d be happy to give more concrete suggestions, especially around where to place the AI logic vs. the classic n8n nodes for the reporting flow.