Codex is in a bit of a funny place. Just as with Claude Code, the fact that the word "code" appears in the product's name is often all that's needed to scare away people who haven't ever seen a line of code in their life before. Ironically, I think that group would get more out of Codex than most developers do. If you've spent the last two years prompting ChatGPT, Gemini, Claude, or any LLM for that matter, you already know how to talk to Codex.

The only difference really is that Codex doesn't just talk back to you — it has the capability to plug right into your workflow and actually go and do the thing you were going to ask it for help with anyway (instead of just walking you through how you'd do it yourself). This is exactly why Codex has quietly turned into one of the tools I rely on most, and almost none of what I do with it involves writing code (or at least code that I'll ever actually look at). Here are a couple of ways I use Codex that have nothing to do with coding…

Document creation

Slide decks, sheets, docs, take your pick

While you can now create outputs directly with the ChatGPT web interface and go from idea to a polished prototype, the process isn't always seamless. You need to manually add the context every time. Upload the template, paste in the data, explain the audience, then do it all over again the next time you want a new version. What's special about tools like Codex is that they can plug right into the folders where your work already lives. You can point Codex at the directory with your draft notes, the template you'd like to reuse, last quarter's version, the spreadsheet your numbers are in, and so on.

You don't need to upload anything to the interface itself — you simply point it to the folder and begin prompting away. Codex can create a range of documents for you like .docx files, .txt, .md, .xlsx, .csv, .tsv, .pptx, PDFs, and so on. For instance, I was recently working on a project for a course I'm taking this semester, and I wanted to create a slide deck for it. I already had a folder created that included the report I had written for it, the sources I had used for my research, the requirement list my professor had provided, and so on. The last step here was generating a deck I could use to walk my class through my findings, and instead of spending time explaining all of that to ChatGPT from scratch, uploading each file, and re-pasting the requirement list, I just opened Codex in that folder and asked it to put together a slide deck.

The best part about this is that Codex can connect to the tools you use daily too. So if you're creating a slide deck for, say, a work project and a lot of the context lives on a Slack group thread, you can connect Codex to your Slack and let it pull the relevant messages itself, instead of you copy-pasting them in chunk by chunk. Same thing for Gmail, Google Drive, Notion — if it's somewhere you actually work, Codex can probably reach it.

Batch rename files

Goodbye, Screenshot 2026-05-14 at 11.47.32 AM

I hate to admit this, but I'm not a very organized person. A quick glance at my MacBook's Desktop will tell you all you need to know. Unfortunately, this also means that I can never find what I'm looking for, when I'm actually looking for it. A few weeks ago I went hunting for some visa documents I'd scanned and saved months earlier, and it took me close to twenty minutes of scrolling through "Screenshot 2026-05-14 at 11.47.32 AM.png" type filenames to track them down.

So now, every couple of weeks, I just open Codex in my Desktop folder and let it sort itself out. The text-based files are the easier half — Codex can scan the first page of a PDF or a Word doc and rename it based on what the file actually is. The prompt I use is something like:

Scan the first page of every PDF, .docx, and .md file in this folder. Rename each one to a short, descriptive name based on what the document actually is. Keep it under five words and use hyphens. If you can't tell, leave it alone.

The image side is what genuinely changed things for me though. Codex (and other agentic tools like Claude Code) can actually see images and rename it based on the contents. I take a lot of screenshots for every article I write and then spend a lot of time renaming each screenshot manually. With Codex, I can simply drop all the images in one folder and say something like:

Look at each .png and .jpg in this folder, and rename it to a short, natural-language description of what's actually in the image. Keep it under 80 characters, no hyphens. For screenshots of text, prioritize the most useful identifying info — sender, subject, or topic.

Keep in mind that when Codex looks at your images or files, they are being sent to OpenAI's servers for analysis. If anything on your desktop is sensitive, I'd recommend using a local model and a tool like OpenCode for this step instead!

Screenshot OCR + local search

Did I screenshot that or imagine it

Renaming gets you to the point where you can browse a folder and recognize what's in each file. But there's a different kind of finding problem renaming doesn't solve — when you remember the contents of a screenshot but not enough about it to know what its filename would even be. "I know I screenshotted a tweet from XYZ person about XYZ thing, but I can't remember when." Scrolling through thumbnails to find one specific screenshot is its own kind of torture. What you actually need is to search the text inside the images, not the filenames around them. Codex handles this surprisingly well. I point it at my screenshots folder and ask:

Go through every .png and .jpg in this folder and run OCR on the contents. Create an index.md at the root that lists each filename alongside the text extracted from that image. Skip the ones that don't have any readable text in them.

What I get back is a single markdown file I can open in any editor, Cmd+F through, and find the exact screenshot I was thinking of. And sometimes, instead of going through the hassle above, I simply describe what I'm looking for, and 99.9% of the time, Codex manages to find the screenshot for me on the first try.

Create daily work briefs

The morning briefing I always wanted

The single Codex workflow that's saved me the most time is having it generate a daily work brief for me, and this is a tip I stole from OpenAI's own Codex for Everyday Work guide on its Academy site. It's the first prompt they walk you through, and once I actually tried it, I understood why.

My mornings used to start with tab cycling including my calendar, Gmail, Asana, Slack, Notion, the draft folder open on my desktop. By the time I'd assembled a rough mental picture of the day, I'd lost twenty minutes and most of my willpower. The fix turned out to be the connector setup I mentioned in the document creation section, but pointed at the right question. This is the prompt OpenAI suggests, which I tweaked according to my requirements:

Set up a weekday work brief that starts in the morning and keeps checking throughout the day. At the start of the day, review today's calendar, unread direct messages and mentions from the last 24 hours, unread email from the last 24 hours, my running list of open follow-ups, and any recent context that affects today's priorities. Create a short brief with priorities, meeting prep, messages that need replies, decisions I owe, and useful FYIs. Then check back every hour until the end of the workday for new replies, meeting changes, or follow-ups that need attention. Only update me when something changes or needs action. Draft replies only when the next step is clear, and flag anything you cannot access or cannot confirm.

What I get back is a short brief meetings that I open with my coffee instead of cycling through tabs. The reason why this workflow is more effective than the standard AI assistant pitch is that Codex isn't waiting for me to prompt it again. The check-back-every-hour part of that prompt means it keeps an eye on things through the day on its own, and only pings me when something actually changes or when there's a reply that needs drafting. That's the version of "AI assistant" people have been promising for years and the first time I've actually experienced it being more useful than annoying.

Don't let the name fool you

If you've dismissed Codex because you assumed all it can do is build your apps and websites, it's worth giving it another look. Most of what I use Codex for has nothing to do with building like you ust saw above, and the name is doing it a disservice. So, trust me, give it a go again.