What Do I Check First When My AI Agent Build Looks Wrong?
Eight things a first build can show you, what I said each one is, and what I did about it.
Find your symptom in the table below. My answers are short: don't be alarmed by the screen change after the up arrow, a section that grays out after save means the save worked, press save to leave the connectors screen, reauthorize after a password change, and refresh after a delete error.
On October 7, 2026 I told the Women Applying AI group, "Do not be alarmed. This is a regular part of the build process." (16:56) That was one of seven moments in the Gumloop build when I stopped to say what the screen was doing. The table below has all seven, plus one toggle that goes by three names across my materials. Find your row, and the last column links to the longer page.
I told the group this about my relationship with the platform in the same session:
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.
What does each symptom mean?
The screen details below are as shown October 7, 2026. Screens change, so your account may differ from mine. I said so at the start of the build:
From the recording, 8:12. My screen may look a little different from yours.
| What you see | What I said it is | What I said to do | Read more | | --- | --- | --- | --- | | Your screen does not match mine (8:12) | My screen may look a little different from yours | I stated no fix | The full session | | The screen changes after you press the up arrow (16:45) | A regular part of the build process; the chat is the starting point, and more controls appear once you press it | Don't be alarmed | Dictating goals | | A section grays out after you press save (21:21) | The confirmation that the save was successful | I stated no further step | The session hub | | A screen will not let you leave (21:12, 31:20) | Some screens advance on their own and others hold you until you advance out of them; the connectors screen is one where save is how you exit | Press save, then check the connectors section for the confirmation | What is a connector | | The AI agent asks for access you did not set up, or keeps asking for one you paused (37:49, 38:10) | I had not connected my calendar in the connectors section, and had named Slack in my first prompt before pausing it | Log in with one click when the AI agent asks; tell the AI agent you are pausing a connection so it stops prompting | Which connectors an AI agent needs | | A connected tool stops working after you change its password (45:28) | The password change cut off the access | Return to Gumloop and reauthorize the permission | Control what an AI agent can do in Gmail | | Delete shows an error (48:11) | Sometimes the delete already went through | Refresh; if that does not confirm the delete, go back to that box, the one next to the three dots | How to delete an AI agent | | The self-updating toggle has a different name than the one you read about (24:18) | See the next section | I stated no fix; I am not claiming the names are the same control | What is edit own instructions |
The gray-out also appears in my setup video, where I say it typically grays out the area once the platform saves it.
Why does the same toggle have different names?
Four places name the toggle that lets the AI agent update its own instructions. Across them the names are three, because the documentation repeats two of them:
| Where | The name | | --- | --- | | The recording, October 7, 2026 | "edit own instructions," a blue toggle that was on | | My setup video | "allow self updates" | | The session page | Self-Improving Instructions | | Gumloop's documentation | A section titled Self-improving instructions, and a setting it calls Edit Own Instructions |
Checked on 2026-11-05 against Gumloop's Agents documentation, which uses the recording's name and the session page's name.
On the toggle I said:
From the recording, 24:18. You'll notice there's something that says edit own instructions with a blue toggle button that is on. What that does is it allows the agent to learn from your interaction inside of the chat feature.
I am not claiming these names are one control. A screen check in your own account would settle that, and the recording does not show one.
What goes wrong
Three moments in the session where a build stalls or breaks, in my words.
A screen that holds you until you save. On the connectors screen I said:
From the recording, 31:20. This is one of those screens inside of Gumloop that you actually have to click the save button to exit out of the screen, okay?
A connected tool that stops after a password change. I said:
From the recording, 45:48. So whenever you change your password in a system connected to Gumloop, you have to come back to Gumloop and reauthorize permissions. It won't automatically update once those passwords are changed.
A delete error (48:11). I said to refresh the screen first, because the delete sometimes already went through. If the refresh doesn't confirm it, I said to go back to that box, the one next to the three dots. The longer steps are on How Do You Delete an AI Agent?.
What the recording does not show: any step past the eight rows above. Where a row says I stated no fix, I stated none.
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 tables 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. Screen steps are dated as shown October 7, 2026.
Build Your First AI Agent in One Hour: the full session with the clips, the prompts and the transcript.
Related reading: How Do You Delete an AI Agent?, How Do You Control What an AI Agent Can Do in Your Gmail?, What Is a Connector?.
Frequently asked questions
Does the screen change after the up arrow mean something went wrong?
I said it is a regular part of the build process and not to be alarmed. The chat is there to get started, and more controls appear once you press the up arrow.
How do I know a save worked?
I said the section grays out when you press save, and that confirms the save was successful.
What happens to a connected tool when I change its password?
I said the platform can no longer reach that system until you come back and reauthorize the permission. It does not update on its own after a password change.
Is the self-updating toggle called the same thing everywhere?
No. The recording, my setup video, the session page and the documentation use different names for it. I am not claiming they are the same control.
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.