Can I See How My AI Agent Reached Its Answer?
Where I looked on October 7, 2026 to see what my AI agent was doing, and a three-question review built from those screens.
Yes, in the builder I used on October 7, 2026. I said: "As it's reasoning, you'll see exactly what it's doing. You can double-click into those later into the reasoning and review that."
On October 7, 2026, in the Women Applying AI session, I showed three places to look when you want to know what your AI agent did. I like building in this platform for that reason (38:55).
The builder in that session is Gumloop, and On my relationship to it, I said this:
From the recording, 33:22. I don't work for Gumloop. I found this tool over the summer through Lenny's podcast, and I started using it and have since made myself an ambassador for them. I've created a relationship with Gumloop. I am not an employee.
More on that is in Do you work for Gumloop? (the answer from the session)
What does the AI agent show while it works?
After I pressed the up arrow on the first dictation, the screen changed and my AI agent began to think (17:12).
From the recording, 17:12. What you'll see on your screen is that your agent starts to think. And you'll see that it actually dictates and transcribes what it's doing based on the instructions that you provided on my screen.
That narration is one of the screens I pointed to. The AI agent writes out what it is doing as it works.
Can you look back at the reasoning later?
I pointed at it a second time when my AI agent was working on the two skills.
From the recording, 38:41. So it thinks for a few moments, while it's thinking, I want to call attention to two things. As it's reasoning, you'll see exactly what it's doing. You can double-click into those later into the reasoning and review that.
On October 7, 2026 I also compared this with other tools, and I hedged it myself, so I give it as I put it:
From the recording, 38:59. You can build agents inside of ChatGPT and Claude. They don't always allow you to click into the reasoning. It may show you the reasoning, but I can actually read the source that it pulled from.
That comparison is what I said on October 7, 2026, in my own hedged words.
What are three questions to ask of the screen?
Replace these with yours. Each question points to a screen I showed.
- 01
What did it do? The narration while it worked.
- 02
Where did it pull from? I said I can read the source it pulled from (39:07).
- 03
What did it write for itself? The instructions and skills. I opened the instructions and showed where to open the skills.
Can you read what the AI agent wrote for itself?
The third question has a second screen. After the skills, I told my AI agent to update its instructions (40:49), and then I opened what came back.
From the recording, 42:34. So if we're on the screen now, you'll see I've shown you where that we can double click into skills to read that. I told my agent a few moments ago to update the agent instructions. You'll now see, if you click into that box, my agent has operating rules. It has core jobs and it has a rhythm. So all of this is editable.
I said the chat feature is the easiest place to edit both skills and instructions (43:03).
Checked on 2026-11-05 against Gumloop's agent documentation, which says you can tell the AI agent to update its instructions or turn the work into a skill. Screens change, so the steps on this page are dated.
What goes wrong
I named one situation: self-updating instructions that nobody reads. I said to put a reminder on your calendar at a frequent cadence to go in and read the AI agent's instructions if you give it access to auto update (43:10).
What the recording does not show: I did not describe how the reasoning steps look on screen, open a source from a Gmail step, or show a case where my AI agent was wrong.
Where this comes from
This article grows from the session I taught live with Women Applying AI on October 7, 2026. The quoted passages are my own words, taken verbatim from the recording. The questions and steps around them were drafted by my team's AI agents from those words and my published pages, and checked against the recording before publishing. Platform details that have a public source carry a dated Checked on line.
Build Your First AI Agent in One Hour: the full session with the clips, the prompts and the transcript.
Related reading: How to Test an AI Agent and Give Feedback, Why Build an AI Agent on Gumloop, Not Only ChatGPT or Claude, and the session page, Build Your First AI Agent in One Hour.
Frequently asked questions
Where do I look to see what the AI agent is doing?
At the narration on screen while it works. I said you see exactly what it is doing as it reasons (38:47).
Can the reasoning be reviewed afterward?
I said you can double-click into the reasoning later and review it (38:51).
Can I read what the AI agent wrote for itself?
Yes. I showed where to open the skills and opened the instructions it wrote, which had operating rules, core jobs and a rhythm, and I said all of it is editable (42:34).
Build something that actually runs your workflow.
A focused, free 60-minute live session with Jacqueline. You build alongside her, on your own real task, and leave with an AI agent that is already running.