Set up an Instagram AI agent in minutes
Turn keyword-triggered DMs into guided conversations. Your agent can build rapport, ask the right questions, collect lead data, qualify interest, and push the chat toward the right next step automatically.
AI Agent Setup
From comment trigger to guided conversation
Conversation hook
"Hey, hope the eating guide helps. What's been your biggest frustration with getting fit lately?"
Objective
Build rapport around the user's fitness journey, qualify willingness to invest $50/week, and route them to either a booking link or the $29/month community.
Required fields
fitnessPainPoint, willingnessToInvest, offerPresented, nextStepConfirmed
Compact instructions
One question at a time. Keep replies to 1-2 casual sentences. Never pressure. Offer coaching first, community second.
Handoff rule
Hand off if they ask for a human, get frustrated, or the conversation loops without progress.
Test mode
Run test chats with the agent before switching it live so you can verify the hook, tone, and routing.
Setup with a conversation, not a complex builder
Configuring an AI assistant usually means writing long prompts from scratch, guessing the right structure, and repeatedly editing fields until the behavior feels right.
Here, you can start with the Setup Assistant instead. You describe what the agent should do in plain language, and the assistant drafts the key pieces for you: the hook, instructions, objective, required fields, and handoff guidance.
Then you review the draft, make manual edits if you want, and test the conversation before switching the agent to Live.
Hook
The first question added after your automatic DM to get the user talking.
Instructions
How the agent should behave, speak, and pace the conversation.
Objective
What the conversation is trying to achieve before it is considered complete.
Required fields + handoff
What data must be captured and when the thread should be moved to a human.
What these agents actually do
This is built for guided conversations: rapport-building, qualification, routing, data capture, and moving people to the right next step.
Lead the conversation
- Build rapport before asking for the next step
- Guide users toward a call, application, quote request, offer, or sale
- Collect structured fields like name, email, budget, timeline, or product interest
- Qualify intent before your team steps in
- Route people to different offers depending on what fits them
- Keep the conversation moving with optional reminders
- Switch to manual handoff when the agent reaches a limit or a closer is needed
Collect usable data
If your objective needs structure, define the exact fields to capture. The agent works toward completing them inside the conversation instead of leaving you with vague chats to sort manually later.
Examples
Email for a lead magnet, budget for qualification, team size for B2B routing, preferred offer, deadline, or any other field that matters before handoff, qualification, or conversion.
CSV export
Download completed objective data whenever you want.
Webhook delivery
Send structured completions into your CRM or backend in real time.
How the hook actually works
The hook is the bridge between your automatic DM and the AI conversation.
1. Keyword triggers the DM
Someone comments your keyword. CommentKeyword replies publicly and sends the automatic DM you already configured.
2. The hook gets them talking
Right after that DM, you can add a short hook question. Its job is not to do the whole conversation. Its job is to get the user to reply in the direction you want.
3. AI only kicks in after they answer
Once the user responds to the hook, the AI agent continues the conversation using your instructions, objective, reminders, and handoff rules.
From idea to live agent
The workflow is linear: describe the job, review the draft, test it, assign it, and go live.
Tell the assistant what you want
Describe who the agent should talk to, what it should ask, and what a successful outcome looks like.
Review the generated draft
The setup assistant drafts the hook, instructions, objective, required fields, and handoff guidance for you to review.
Test and refine
Run test conversations before going live so you can refine tone, pacing, qualification logic, and routing.
Assign it to keywords
Attach the live agent to the comment keywords that trigger your automatic reply and DM flow.
Go live
Once it behaves the way you want, switch it live and let it handle real conversations under your rules.
Example: Fitness Budget Qualifier
1. User comments to get the guide
CommentKeyword replies publicly and sends the first DM with the guide plus the fitness-frustration hook.
2. The first job is rapport
The agent keeps it short, asks one question at a time, and explores their frustration, how long they have been dealing with it, and what result they want.
3. Then it qualifies and routes
If they can commit to $50/week, it sends the booking link. If not, it offers the $29/month Skool community instead.
4. Completion triggers action
Once the right next step is confirmed, export the result, fire a webhook, or move the thread to manual either because the objective was completed or because handoff logic triggered first.
Test, remind, and hand off when needed
Inside the AI setup area, you can run test conversations against the current draft of your hook, instructions, objective, and handoff rules.
That lets you validate whether the agent builds rapport the way you want, asks the right follow-up questions, and routes users to the right next step before you switch it to Live.
If the behavior is not quite right, you can reopen the Setup Assistant on the same agent, describe what you want to change, and let it help you refine the draft instead of starting over manually.
After the agent is live, you can also configure reminders if the user goes quiet or if Instagram's 24-hour automated reply window is about to close, giving the conversation another chance to restart.
Also available
CSV export, webhook delivery, and moving the conversation to manual either when the objective completes or when your handoff logic says a human should step in.
Build the DM flow once. Let the agent run it.
CommentKeyword gives you the trigger, the first DM, and the AI layer that can build rapport, qualify intent, capture data, and move interested users toward a real next step.