The biggest lesson I learned from client number 5
This Friday I got a reply from a person on Upwork who wanted me to create a fully automated social media content machine. So I offered to make a free MVP within 24 hours, just to check that what they were asking for was possible and to show proof of work, you know. When I delivered the MVP on Saturday, they were satisfied with the results. So I started hacking away the entire weekend. But here’s where things started to take a turn. As I’m building out the entire system and sending updates, I’m realizing that there are some red flags going on. First of all, the scope keeps changing. And while the client thinks it’s plug and play, the APIs and apps he wants to use are not connected in any way whatsoever. So today, Monday, I’ve completed what I think is 70% of what he wants. And I’m damn happy for keeping up and completing the work so fast. Until I send over a Loom, and then the client says that we can’t use any other tools other than the ones they’ve picked. Okay, no worries. We can manage. I have a few workarounds. But there’s one problem, and that is that the client wants to use Canva. But why Canva? Because Canva lets you create 5 AI clips a month. So because of that, the client wants Canva to receive the prompts from GPT and then create AI clips. So that we can merge the content with clips from Pexels or Pixabay that we get via the API. If you’ve ever touched Canva, you know that Canva has no such endpoint that lets you automatically create AI content. It’s at this point I have the full picture of what they actually want: Send AI prompts from GPT → To Canva to create AI content → Export it to DaVinci Resolve and use a Python script to edit the video → Then upload it to all social media platforms → And finally keep track of all analytics. So what’s the lesson I learned from this? 1. Push for the call. The client didn’t want to meet on a call, but calls help avoid miscommunication. And when it comes to automation, understanding what people want and how they want it is 90% of the work. 2. Map out all the steps, every single one of them. When we know what the steps are, we can automate things. Nothing should be unknown, as there can be hidden issues there.