Today’s Practical AI Tip: Let AI Catch Your Forgotten Drafts

Today’s Practical AI tip comes from a mistake I made recently.

I wrote what I thought was a great email, got pulled away, and never sent it. Then I sat there wondering why the other person had not responded. Eventually, I realized the problem was not their inbox. It was my drafts folder.

When I shared the story with my latest Copilot cohort, I found out I was not alone. Several people had experienced that awkward, “Didn’t you get my email?” moment, only to discover the message had never been sent.

So, we turned the mistake into a simple AI workflow.

I created a scheduled prompt in Copilot that checks my drafts every two days. It looks for messages that appear ready to send, tells me who they are addressed to, and recommends whether I should send or delete them based on my sent items and the surrounding context.

This is a great example of practical AI. It is not about replacing your judgment. It is about creating a safety net for the small mistakes that cause confusion, delays, and embarrassment.

Sometimes the best AI use case is not flashy. It is simply making sure a forgotten draft does not become a missed opportunity.

ChatGPT Agent vs. ChatGPT Connector: Who Wins the Outlook Email Search Showdown?

I ran a real-world test to see which ChatGPT tool could deliver better insights from my Outlook account a deep research connector or the new agent feature. The task? Find every conversation related to AI or training from the past six months. Capture each contact’s name, title, last message date, and a summary. **First up: ChatGPT Connector using Deep Research** This approach tried to use an API-like method to extract content from your inbox. It failed. Why? Deep Research, while powerful for web research, isn’t optimized for structured email search. It likely ran into permission issues, API throttling, or simply lacked the depth to interact with your mailbox accurately. **Next: ChatGPT Agent using an Embedded Browser** This version mimicked how a human would search. You logged into Outlook inside a secure browser session within ChatGPT. The agent scrolled, searched, and analyzed content in real time—just like a human assistant. And it worked beautifully. **The Results** The agent gave me a clean list of contacts with last interaction dates and summaries. Essentially, it built a mini-CRM out of my mailbox in minutes. **The Verdict** Deep Research fell short due to system limitations and design intent. The agent triumphed thanks to direct, browser-based access. The human-like approach wins when APIs can’t keep up.