Make an Existing Workflow More Agentic

Over the last few months, we’ve built a lot of small AI tools and workflows in Tiny AI Lab. This week, instead of starting from scratch, I want you to go back to something you’ve already built and see if you can make it a little more agentic.

You don’t need to build a fully autonomous agent if you have never done that before. The goal this week is to understand what actually changes when you give an AI system a little more responsibility for figuring out what to do next and where that can be useful for you.

The Challenge:

Pick an AI workflow you already use. It could be something you built for one of these challenges, something you use at work, or even just a prompt you find yourself coming back to. If you are stuck on where to start, try the executive brief we built in one of our first Tiny AI labs. 

Then map out how it works today and look for a step where you are still making the decision about what happens next. Maybe you use AI to summarize meeting notes, but you decide which things actually need follow-up. Maybe it helps you research companies, but you tell it exactly what to search for each time. 

Choose one of those decisions and see if you can hand it over.

The change can be small. Your research workflow might decide when it doesn’t have enough information and go looking for another source. Your meeting notes tool might identify which action items are urgent and which ones are unclear. Your inbox workflow might decide whether something needs a reply, can be ignored, or should be turned into a task.

The prompt to get you started

Drop this into your AI of choice to get started:

I have an AI workflow I already use and I want to experiment with making it more agentic.

Here’s how it works today:

[describe your existing workflow]

Help me break the workflow into its individual steps and identify where I am currently making decisions, giving the AI another instruction, or manually moving information from one step to the next.

Then suggest a few ways I could give the AI more responsibility within this workflow. I want to start small, so focus on one additional decision or action the AI could reasonably handle on its own while keeping appropriate human oversight. Once we choose one, help me rebuild the workflow step by step using tools I already have access to where possible.

Too easy?

Give the system a loop to check its work. 

For example, instead of researching a topic once and returning an answer, it could review what it found, identify missing information and decide whether it needs another round of research. A writing workflow could draft something, check it against a set of criteria and revise anything that doesn’t meet them. 

You’re essentially adding a step where the AI can evaluate its own progress and decide what needs to happen next rather than waiting for you to give it another instruction. In some cases this type of workflow will improve your output.

Why this matters?

There is a lot of conversation right now about agents, agentic workflows and increasingly autonomous AI systems, and without experimenting with these things it can be hard to understand what people mean.

One of the easiest ways to understand the difference is to actually build “agency” into agents.

There’s a difference between asking an AI to complete a task, building a workflow where AI completes several predefined steps, and building a system that can make some decisions about which steps it should take in order to reach a goal.

Most people probably aren’t going to jump straight from ChatGPT prompts to fully autonomous agents but there are a lot of small ways to reduce work by understanding agentic workflows and taking a small step can help open up some possibilities in your other work.

That’s it for this week.

Hit reply and tell me what you thought of the challenge this week.

I read every response, and your experience helps me shape these challenges. If you get stuck or have questions, reach me at hello@tinyailab.ai