I've been testing Lindy for a few weeks now, mostly for automating some basic data gathering and email sorting. The core automation works well enough, but I keep circling back to the "human in the loop" feature they emphasize. It sounds like the perfect safety net for more complex tasks.
My main question is about its real-world usability. When an automation pauses for human input, what's that interface actually like? Is it a clunky context switch, or something relatively seamless?
For example, if I set up a Lindy to screen inbound client requests and draft replies, I'd want to approve the draft before it's sent. In practice, does the notification come quickly? Is the relevant information (the original email, the drafted response) presented clearly so I can make a fast decision without hunting through tabs? And after I give input, does the automation resume reliably?
I'm coming from building similar things with Python and APIs, where building a clean "approval step" requires a lot of custom front-end work. Lindy seems to promise this out of the box, but I'm wondering if it's smooth or feels like an afterthought. Any experiences, especially with time-sensitive tasks, would be really helpful.
Good question. I've been using the human in the loop for approval workflows, similar to your email example.
The interface is a simple Slack message or email notification. It shows you the draft and the context clearly, and you can approve or edit with one click. The context switch is minimal. It does resume reliably once you give input.
Where it gets a bit sticky is if the task is time-sensitive and you're away. The pause can hold things up. It's great for quality control, less so for urgent tasks. You might want to keep those fully automated or flagged differently.
Automate the boring stuff.
Totally agree about the time-sensitive snag. I've found you need to be really strategic about when to trigger the pause.
A workaround I use: I set up a second, fully automated flow for anything flagged as "urgent" by keywords in the initial request. That way the standard approval flow handles the bulk, but truly time-critical stuff bypasses the human check. It requires a bit more setup but solves the "I'm away" problem.
Has anyone tried using their scheduling or calendar integration to manage the pause time? Like, only allowing approval prompts during work hours?
It's usable for the core loop you described. The Slack/email notification shows the draft and original side by side. It's a fast decision.
Where it breaks is state management for complex inputs. If your edit is more than a tweak, you often have to restart the whole automation. The audit log for who approved what is also shallow, which is a compliance headache if you need it.
For your Python comparison, it saves you from building the UI but you trade off control. That's the real calculus.
Trust but verify, then don't trust.