AI Agent is where you ask Pod for help using the context already available in your workspace. You can ask a question, tag records, add one or more Skills to guide the response, include files when available, use web browsing for live external context, and keep working in the same thread.

Use this article when you want to write better AI Agent prompts and understand the controls in the input box. If you want to create reusable Skills, start with Build a new Skill or Build and manage Skills.

You can open AI Agent from multiple places, depending on what you are working on.

In the Pod web app:

In the Chrome Extension:

In both the web app and Chrome Extension, AI Agent works best when you are clear about the record, workflow, and output you want.

Most AI Agent work starts in the input box. The placeholder tells you that you can ask anything and type / for agent commands.

You can use the input to:

Think of the input box as a workspace for composing the request, not just a chat field.

A good AI Agent prompt gives enough context for Pod to understand the job, but stays focused on the outcome you need.

Strong prompts usually include:

Example:

Review this opportunity before my forecast meeting. Focus on close date movement, executive engagement, recent activity, next steps, and any risks that could affect commit confidence. Return the top 3 risks, the evidence for each, and the next action I should take.

AI Agent can use the record you are already viewing, but you can also tag records directly in the prompt when you want to be more specific.

Use record tags when you want to:

For example, you might ask AI Agent to compare two opportunities, summarize engagement across an account and its open deals, or draft next steps for a specific contact.

Skills are reusable instructions that guide how AI Agent should work. You can add a Skill when you want a saved workflow, prompt style, or response format to shape the answer.

You can add Skills from the input in a few ways:

The picker groups options such as Recently Used, Pod Agents, Personal Agents, and Org Agents. The picker uses the runtime label Agents, but the saved reusable workflows you create and manage are Skills.

You can keep typing after / to filter by name or description. Select a Skill to add it to the prompt.

Some prompts benefit from more than one Skill. For example, you might add one Skill that reviews deal health and another that drafts a customer-facing follow-up.

Use multiple Skills when each one contributes a distinct instruction set. Keep the rest of the prompt clear about the final output you want, so AI Agent knows how to combine the guidance.

If the request gets too broad, split it into separate prompts or turn the workflow into a more focused Skill later.

Turn on web browsing when the next prompt needs live external context, such as company news, market context, industry updates, or information from a prospect's website.

Leave it off when the answer should rely only on Pod's internal deal context. Deal-specific facts should still come from CRM fields, activity history, emails, meetings, transcripts, notes, and other connected data inside Pod.

Web browsing applies to the prompt you are about to send.

In supported surfaces, AI Agent can include files with a prompt. Use attachments when the AI Agent needs to consider a document, screenshot, deck, or other file as part of the current request.

If an upload fails or expires before sending, use the retry or remove controls shown on the attachment.

After you send the prompt, AI Agent runs against the available context and returns an answer in the same thread.

When reviewing the answer, check:

If the answer is close but not quite right, follow up in the same thread instead of starting over.

Use follow-ups to refine the response or move from analysis into action.

You can ask AI Agent to:

Click New agent when you want to clear the current conversation and start fresh.

When AI Agent is still responding, you can draft follow-up prompts in the expanded experience. Pod queues them and sends them after the current response finishes when the run is ready.

This is useful when you already know the next question but do not want to interrupt the current answer.

If the run stops or fails, Pod may ask you to manually send the queued follow-up.

After a response completes, you can manage the latest answer in the thread.

Depending on the run state, you can:

Feedback helps the team improve AI Agent quality over time.

AI Agent can propose actions, such as email drafts, CRM updates, or other tool-driven changes. Pod does not complete those actions until you approve them.

When actions are pending, review the action details in the thread and choose whether to approve or reject them.

For the full action lifecycle, see Manage agent actions.

Use History when you want to reopen a previous AI Agent run on the same record.

History is useful when you want to:

If no Skills appear, your workspace may not have any personal or Org Skills yet. You can still write a prompt directly, or create/import a Skill for repeatable workflows.

If search returns no results, clear the search and try a broader term.

If the Skill list fails to load, use Retry. If it keeps failing, send us a message.

If you are viewing another teammate's run, it may be read-only. Click New agent to start your own conversation.

Need help? Send us a message via the in-app chat or email us at [email protected].

Want to talk to someone? Book a session with one of our specialists.