A few weeks ago I was running a Lindy community session.
We were deep in it — someone had just shared their setup, and a few people were riffing on how they’d handle a specific problem: searching through years of email, calendar data, and text messages to build a prospect list from scratch.
And I said something that caught people off guard.
“I would actually not use Lindy for most of this.”
The room got quiet. Which made sense — we were in a Lindy session.
But here’s why I said it, and why I think this matters for how you think about AI tools in general.
The Wrong Way to Think About Automation Tools
Most people pick a tool they like and then try to make everything fit that tool.
I’ve watched this play out over the years with Zapier. With n8n. With Python scripts. With no-code platforms. Someone discovers a tool that works really well for one thing, then they spend the next six months forcing every problem through that same mold.
It’s understandable. Learning a new platform takes time. It’s easier to get better at one thing than to maintain expertise across several.
But the cost is real. When you use the wrong tool for a job, you either can’t build it at all, or you build something fragile that breaks when something slightly unexpected happens.
The fix isn’t finding the perfect universal tool. It’s building a mental model for routing work to the right tool.
The Framework: Deterministic vs. Exploratory
Here’s the question I ask before building any automation:
Is this task deterministic or exploratory?
Deterministic tasks have predictable inputs and predictable steps. The same trigger happens, the same process should follow. An email comes in, you summarize it and log it. A form gets submitted, you notify three people. A meeting ends, you generate a follow-up. These tasks repeat on a cadence or on a trigger. You can define the logic once and walk away.
This is where Lindy shines. It’s built for rinse-wash-repeat workflows. Set it up, let it run, don’t think about it again.
Exploratory tasks are different. You don’t know exactly what you’ll find when you start. You’re searching through messy, unstructured data. You’re making judgment calls based on what turns up. You need a tool that can reason in real time, adjust its approach, and handle things it wasn’t explicitly programmed for.
Searching through years of emails and texts to build a prospect database? Every step depends on what the previous step found. That’s exploratory. And Claude Code — which can write and run code, iterate based on results, and handle ambiguity — is a much better fit.
A Real Example of Getting This Right
When I was working on connecting different data sources a while back, I spent about two weeks trying to make a Python script do it reliably. It worked in testing. It fell apart in production because inputs weren’t as clean as I expected.
Then I moved the recurring part of that workflow to Lindy. Cost comparison wasn’t close — and the maintenance overhead basically disappeared. Lindy just handled it every time without me touching it.
But I kept Python and Claude Code around for the exploratory jobs. The one-off research tasks, the data cleanup projects, the things where I needed to investigate first and build second.
Both tools are in regular rotation. They handle different jobs.
The RATs Diagnostic
Before building any automation, I run what I call a RATs check. Is this task:
- Redundant — something you do repeatedly on a schedule or trigger?
- Annoying — low-value work that pulls your attention?
- Time-sucking — taking longer than the output justifies?
If yes to all three: it’s an automation candidate.
Then the question becomes which tool. And the deterministic/exploratory framework tells you where to start.
Redundant + predictable = Lindy. Exploratory + one-off = Claude Code. Something in between — a task that repeats but also needs real-time reasoning — might be a hybrid where Lindy handles the trigger and scheduling, and Claude Code does the thinking.
The Broader Point
I keep saying this in workshops: the more tools you know, the more leverage you create.
Not because you use all of them at once. But because you can route work to whichever one fits best.
A carpenter doesn’t use a hammer for everything just because they’re really good with a hammer. They know what a chisel does, what a router does, what a hand saw does. The expertise isn’t in one tool — it’s in knowing which tool the job calls for.
Same logic applies here. Lindy is excellent. Claude Code is excellent. They’re excellent at different things.
The highest-leverage thing you can do right now is build that mental model. Figure out the shape of each tool’s strengths. Then, when a new problem shows up, you’ll know within five seconds which direction to go.
That routing instinct is worth more than being really, really good at any single platform.
Want help figuring out which tasks in your workflow are worth automating first? The 4-Day AI Sprint gives you a systematic way to identify your highest-leverage opportunities.
