Skip to article

What Are the Parts of an AI Agent?

One map of the builder screen from October 7, 2026, with each part in the order I set it up and a link to the page that explains it.

By Jacqueline V. TwillieNovember 6, 20265 min readDefinition
Quick answer

I built my AI agent in this order: persona, goals and deliverables, then a model, instructions, connectors and skills, with guardrails alongside. I said: "And this is for any agent, persona, goals, and deliverables."

There are three rules I like to follow as a best practice for setting up an AI agent: give it a persona, tell it its goals, and tell it its deliverables (9:13). On October 7, 2026 I built a chief of staff AI agent live with Women Applying AI, and the rest of the builder came in the order below.

From the recording, 9:57. And this is for any agent, persona, goals, and deliverables.

The builder 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 are the parts, in the order I built them?

As shown October 7, 2026, nine parts came up in the hour. Screens change, so the placements are dated.

  1. 01

    Persona. It went in on the first screen, before the builder opened. Mine was "you are my chief of staff [AI] agent" (9:28). Detail: What Is an AI Agent Persona?.

  2. 02

    Goals. Same screen, spoken into the dictate feature (11:54). Detail: What Are AI Agent Goals and Deliverables?.

  3. 03

    Deliverables. Same screen, named with the goals. I left the format of one deliverable open and said the AI agent would ask me about it (13:05). Detail: the same page as goals.

  4. 04

    Model. On the first tab, labeled Agent, a model picker with Gumloop's auto option and a dropdown of the models available (21:43). I said a model can be switched in one click (22:00). Detail: What Is Model Switching?.

  5. 05

    Instructions. On the first tab, labeled Agent, below the model. The AI agent wrote them from what I had dictated and the questions I answered (24:07). Detail: How Often to Review AI Agent Instructions.

  6. 06

    Edit own instructions. A blue toggle under the instructions, on in the demo (24:18). Detail: What Is Edit Own Instructions?.

  7. 07

    Connectors. The plus icon next to add, in the connectors section (25:50). Detail: What Is a Connector?.

  8. 08

    Skills. Beneath connectors (31:56). I said, "Skills are actually workflows" (34:06). Detail: What Is a Skill?.

  9. 09

    Guardrails. I give them to my AI agents, and I tell them what they can do and what they can't do (41:26). Detail: What Are AI Agent Guardrails?.

Persona, goals, deliverables and guardrails are what I give an AI agent. Model, instructions, connectors and skills are sections of the builder, and edit own instructions is a toggle under the instructions.

Which parts came up without being built, and which did not come up?

Triggers came up without being built. Knowledge source and abilities did not come up.

  • Triggers. One of four words I pointed at, next to the word agent: triggers, access and settings (18:12). I did not open it in the recording. My setup video does, and How Do I Schedule an AI Agent to Run on Its Own? walks through it.
  • Knowledge source and abilities. I did not bring up either in the recording, and the documentation line below lists both.

Checked on 2026-11-05 against Gumloop's agent documentation, which lists the first tab, labeled Agent, as holding the model, instructions, connectors, skills, knowledge sources, subagents, secrets and abilities, with Edit Own Instructions under the instructions, and lists Triggers, Access and Settings as the other three tabs. My count on October 7, 2026 was four words, and the documentation names the same four.

What goes wrong

I named two situations on this screen.

The first is leaving information or access out. I used a new hire to say it: if I leave out critical information and restrict access, that new hire isn't going to achieve as much (6:05).

The second is a screen that needs a save before it closes. On the connectors screen I said you have to click the save button to exit out of the screen, and then you'll notice a confirmation that your connectors are there (31:20).

What the recording does not show: knowledge source or abilities, the contents of the skills it wrote, and the access tab.

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 lists 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: What Is a Connector?, What Is a Skill?, What Is an AI Agent Persona?, and What Are AI Agent Guardrails?.

Frequently asked questions

What order did I build the parts in?

Persona, goals and deliverables went in on the first screen (9:28 to 13:05). On the first tab I set the model (21:43), the instructions (24:07) with edit own instructions (24:18), connectors (25:50) and skills (31:56). Guardrails came up at 41:26. Triggers I pointed at but did not build (18:12).

Which parts did I say apply to any AI agent?

Persona, goals and deliverables.

Which part gives the AI agent access to other tools?

Connectors. I said connectors are what make AI agents work and help them move from one part of a workflow to the next (26:07).

The next step

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.

More on Vocabulary for Non-Technical Managers