Member-only story

6 Must Know Agentic Workflow Design Patterns (and Their Key Trade-Offs)

A guide to agentic workflows that balance performance, complexity, and reliability

Tanmay Deshpande
Data Science Collective
7 min readFeb 17, 2025
Source — Image by Author

AI-driven apps rely on large language models (LLMs) to tackle tasks once limited to specialized rule-based systems. As these LLM-based apps become “agents” — independently choosing their steps or calling external tools — they can quickly become unreliable or inefficient.

Picture current computer systems as well-planned hiking trails: each turn and signpost is fixed, so you always know the path. Now imagine handing a map to a savvy hiker who decides where to go at each fork. That’s what LLM-based “agents” do: they can plan their routes and use different tools.

We need strong patterns that go beyond Retrieval Augmented Generation(RAG), and that also tame the agents' “autonomous” behavior to get consistent results.

In this article, you will learn about various such patterns and their advantages and disadvantages, which will help you decide which pattern to use and when.

1. The basic: Retrieval Augmented Generation(RAG)

Create an account to read the full story.

The author made this story available to Medium members only.
If you’re new to Medium, create a new account to read this story on us.

Or, continue in mobile web

Already have an account? Sign in

Data Science Collective
Data Science Collective

Published in Data Science Collective

Advice, insights, and ideas from the Medium data science community

Tanmay Deshpande
Tanmay Deshpande

Written by Tanmay Deshpande

I write about technology in simple words!

No responses yet

What are your thoughts?