If you’ve worked in IT over the past year, you’ve probably asked ChatGPT, Claude or Cursor to write some code for you. I certainly have.<\/p>\n
AI has become an incredibly useful programming assistant. Need a Python script to rename files? Ask AI. Need to scrape a website? AI can usually get you most of the way there in seconds. Even experienced developers are finding themselves writing less code from scratch and spending more time reviewing, refining and testing code generated by AI.<\/p>\n
So does that mean traditional automation tools like Macro Scheduler are becoming obsolete?<\/p>\n
I don’t think so.<\/p>\n
What AI has changed is how we create automation<\/strong>, not necessarily how automation gets done<\/strong>.<\/p>\n When people think about automation, they often assume writing the code is the biggest challenge.<\/p>\n In reality, it often isn’t.<\/p>\n The hard part is understanding the application you’re trying to automate.<\/p>\n If you’re automating a modern web application with well-structured HTML and APIs, AI has an enormous advantage. It can reason about the page, generate Playwright or Selenium code, and often produce something that works remarkably well.<\/p>\n But many businesses don’t live entirely in the browser.<\/p>\n They still rely on Windows desktop software that may have been written ten, fifteen or even twenty years ago. Internal tools. Accounting systems. Manufacturing software. Medical applications. Legacy databases. Remote desktop sessions. Software where there is no API, no documentation, and no obvious way to automate anything.<\/p>\n That’s where the real challenge begins.<\/p>\n Anyone who has automated Windows applications knows that the process is rarely as simple as writing a few lines of code.<\/p>\n Can the button be identified directly?<\/p>\n If not, can it be found using OCR?<\/p>\n Would image recognition be more reliable?<\/p>\n Has Windows scaling changed?<\/p>\n What happens if the dialog appears a second later than expected?<\/p>\n What if the application has been updated and the button has moved?<\/p>\n These aren’t really coding problems.<\/p>\n They’re automation problems.<\/p>\n And solving them is usually an iterative process. You try an approach, test it, refine it and gradually build a reliable solution.<\/p>\n Ironically, AI makes this process easier rather than replacing it.<\/p>\n Today I regularly use AI to:<\/p>\n It’s like having another experienced developer sitting beside you.<\/p>\n The important word there is assist<\/strong>.<\/p>\n AI still needs context. It often needs screenshots, recorded steps, snippets of code or descriptions of what you’re trying to achieve. Building reliable desktop automation remains a collaborative process.<\/p>\n This is exactly why we’ve been experimenting with AI ourselves.<\/p>\nThe hard part isn’t writing the code<\/h2>\n
Desktop automation is often an investigation<\/h2>\n
AI is becoming a brilliant assistant<\/h2>\n
\n
Where we’re experimenting<\/h2>\n