May 28, 2025 Rahul Sharma
<aside> đź’ˇ
tl;dr Traditional workflow automation requires the user to define the rules (e.g., do X, when Y happens given Z constraint). Agentic AI goes one step further and takes care of the logic itself by using an LLM (to think), memory (to remember), and tools (to act).
</aside>
With the rise of agent-based workflows, it’s natural to compare them against traditional workflow automation tools. These traditional tools depend on a user to define rules and logic for execution e.g., do X when Y happens. Each branch of the execution flow needs to be laid out to ensure proper resolution.
In contrast, the agentic workflow tools do NOT need a user to define the logic flow. The agent can reason and think on how to execute the task, avoid errors, and cover edge cases. All it needs is access to the tools (e.g., Calendar, Email, Weather API).
Let’s consider a sample situation and develop both agentic and traditional workflows to resolve it.
<aside> đź“„
Josh enjoys organizing group hikes on weekends with his neighbors. He wants to build a solution that automatically sends an email to the group every Friday evening with details about the upcoming hike based on the latest weather and air quality conditions.
For instance, if the forecast predicts a hot and humid day, Josh prefers to plan a shorter hike in a shaded area. If it's cool and breezy, he opts for a longer hike through open meadows. If the air quality is poor, he’d rather postpone the hike altogether and so on.
</aside>
As you can observe, the workflow can have multiple branches based on each condition. Here’s sample pseudo code to achieve the desired result.
-- Illustrative
...
IF air_quality = 'poor' THEN
SET hike_status = 'postponed';
ELSE IF temperature = 'hot' THEN
SET hike_length = 'short';
SET hike_type = 'shaded';
ELSE
SET hike_length = 'long';
SET hike_type = 'open';
END IF;
...
Let’s create this workflow using an automation tool. I am using n8n but you can pick any other option (e.g., Make, Flowise, Zapier).
The “traditional” workflow below is fetching calendar schedules, checking weather, checking air quality, reviewing list of hikes, and finally sending out an email with the right details. There’s a lot of if-else
nesting to get the desired result which makes it harder to debug and maintain long term.
Traditional workflow automation using n8n