I have been working in the generative AI space “since the beginning”—speaking to my non-computer-science friends here—and I have always raised an eyebrow at the hype.
LLM chatbots have been useful, but they were definitely not the magic pill being promised by influencers. It still took a lot of effort to get to a quality output: crafting the prompt, checking the work, iterating, and completing the parts of the workflow that still depended on a human. Even automated workflows took significant time and effort to plan, design, and build.
That changed this summer when I started pushing myself to do as much as possible through Codex and ChatGPT. One afternoon, I looked at my watch and thought, “Wow, I’m done with my to-do list already.”
What one day actually included
- Running analytics on a SQL database and creating a new business dashboard
- Working through marketing and local-search strategy for a website
- Ordering my weekly groceries online—finally
- Running a quarterly sales-tax report and preparing the return for filing
- Fixing several small but frustrating IT problems that had lingered for months
- Assessing accreditation readiness and developing an action plan
- Finalizing travel planning for an upcoming trip
- Working on advertising and creative strategy for a deadline
- Completing a tedious data-entry project in seconds
I easily saved five hours. More important, I ended the day excited and energetic. I usually finish a full workday feeling like I have not completed everything on my list and I am already drained before tomorrow begins. This time, the mental load felt different.
That does not mean the earlier concerns disappeared. I still had to direct the work, protect sensitive information, review outputs, and make the consequential decisions. The difference was that the system could now carry much more of the workflow instead of handing every intermediate step back to me.
By 3:00 p.m., I looked at my watch and thought, “Wow, I’m done with my to-do list already.”
For the first time, the promise felt concrete: not a flashy output, but hours returned and enough energy left to notice the difference.