The AI chat panel is available from two places:

* Click the **Parsa AI** icon (robot) in the left sidebar. * Press `Cmd + K` (Mac) or `Ctrl + K` (Windows/Linux) from anywhere in the app. The panel slides in on the right side of the screen and stays open as you navigate between CRM sections. Parsa AI maintains context of your current view — if you have a contact record open, it knows which contact you're looking at.

***

Conversational Actions

## What You Can Ask ### Search and Retrieve

Find records without using filters or search bars:

* *"Show me all contacts in the Negotiation stage"* * *"Find leads assigned to Marcus that were created this week"* * *"Which accounts are in the Prospect stage with revenue over \$5M?"* * *"What are the last 5 activities on the Acme Corp account?"*

### Create Records and Tasks

Take action without leaving the chat:

* *"Create a task for me to follow up with Jane Doe on Friday"* * *"Add a note to the TechCorp opportunity: they need a custom SLA in the contract"* * *"Create a new lead: Sarah Kim, [sarah@growthco.com](mailto:sarah@growthco.com), source is referral"* * *"Log a call with Marcus Rodriguez — 15 minute discovery call, positive"*

### Get Pipeline Insights

Ask about your business performance:

* *"What's my total open pipeline value?"* * *"How many deals closed this month?"* * *"Which rep has the highest win rate?"* * *"Show me opportunities closing in the next 30 days"* * *"What are our most common reasons for losing deals?"*

### Draft and Send Emails

Generate and send emails without leaving the chat:

* *"Write a follow-up email to John Smith at Acme Corp about their proposal"* * *"Draft a cold outreach email to leads in the Tech industry"* * *"Send a check-in email to all contacts I haven't emailed in 30 days"*

When Parsa drafts an email, it shows you the preview and asks for confirmation before sending.

### Extract Meeting Notes

Paste raw notes and let the AI structure them:

* *"I just had a call with Dana Lee. Here are my notes: \[paste]. Create tasks and update the opportunity."* * *"Extract action items from this meeting transcript and create tasks for the right people."*

***

Prompt Examples

## Example Prompts Here are some ready-to-use prompts to get started:

Show me all new leads from the last 7 days and their sources. Which deals in the Proposal stage have been there for more than 14 days? Write a personalized re-engagement email for a lead who went quiet 3 weeks ago. Create a follow-up task for every contact I emailed today. What's the average deal size for Closed Won opportunities this year? Summarize the last 3 interactions with Acme Corp.

***

Permissions

## How AI Accesses Your CRM Data

Parsa AI queries your live CRM database in real time. It has access to:

* All contacts, leads, accounts, opportunities, cases, and proposals in your workspace. * Activity timelines — emails sent, calls logged, notes added, stage changes. * Task history and current due dates. * Pipeline values, stages, and close dates. The AI does **not** have access to: * Other workspaces (your data is isolated). * The content of files stored in connected Google Drive or OneDrive. * External CRM systems not integrated with ParsaLink.

***

Context

## Keeping a Thread

The chat maintains context within a session. You can ask follow-up questions without repeating context:

> *"Show me deals closing this month."*\ > *"Which of those are in the Proposal stage?"*\ > *"Send me a reminder to call the reps on those deals tomorrow morning."* Start a **New Chat** to clear the context and start a fresh conversation.

When you're viewing a specific contact, lead, or opportunity and open the chat, Parsa AI automatically has that record in context. You can ask "Summarize this contact's recent activity" without naming them. *** ## Persistent Memory (Facts) Beyond the current chat thread, the agent has a **persistent memory** — a list of facts it remembers about you across every conversation. Useful for steering tone, preferences, and recurring context the agent shouldn't have to re-learn each day. You can: * **Tell it to remember something** — "*Remember that I prefer Mondays for prospect calls and Fridays for internal meetings*". The agent saves a fact and applies it next time you ask about scheduling. * **Browse remembered facts** — go to **Settings → AI Memory** to see and edit everything the agent knows about you. * **Tell it to forget** — "*Forget what I said about Mondays for prospect calls*" or delete the fact from the Memory page. Memory is per-user, not per-workspace — your teammates have their own facts. The agent automatically saves facts when they're explicitly stated ("I'm in the Pacific timezone", "we never email on Saturdays"). It doesn't save everything you say — only things it judges likely to matter later. You're always in control via the Memory page. *** ## Booking Across Multiple Calendars The Calendar agent can compare availability across a group of team members and pick a time that works for all of them. Useful when scheduling a customer meeting that needs two reps + an account exec. > *"Book a 45-minute kickoff with Andrew, Amy, and Jacob and the Globex contacts next week, in the afternoon. Make sure all three of them are free."* What it does under the hood: 1. Looks up the three team members' calendars (internal + Google busy time). 2. Finds the intersection of free slots given the duration and your window ("next week, afternoons"). 3. Picks the best slot (closest to now, most aligned with everyone's preferred booking hours). 4. Drafts the calendar invite with all team members + the Globex contacts as attendees. 5. Asks you to confirm before sending. If no slot works, it tells you that explicitly instead of picking a sub-par time silently — and offers the next-closest options for you to choose between. Pair this with "Send the invite from my calendar so reminders look like they're from me" to get the right ownership of the event in the recipient's view. *** ## Replying from a Digest Email If you got a [daily digest](/email/digest) and hit Reply, your message is routed straight to the AI agent under your identity. So: > *Subject: Re: Your Parsa digest* > > *"Reassign Amy's overdue VIP outreach tasks to Jacob, then send Amy a quick thanks for closing the Globex deal."* The agent reads the reply, calls the right tools (`reassign_task`, `send_email`), and replies in-thread with a summary of what it did. Same agent as the in-app chat — just over email. This is why dropping by the chat panel and dropping into a digest reply feel the same: it's literally the same agent + the same toolset.