Merlin Engage setup happens entirely inside Merlin in your left navigation.
Before you start
You need at least one Merlin Qualify agent set up, since Engage agents must be linked to one.
Your firm needs at least one SMS number configured. Merlin Engage sends and receives texts through this number, and it won't run without one.
1: Create the agent
From the left navigation, go to Merlin → Engage → Create New Agent.
Give the agent a name
This is an internal-only name, used exclusively by your team. This name is not ever used by the agent with your leads.
Link the agent to a Merlin Qualify profile
This is required, it's how Engage knows what information to look for.
Optional: Analyze My Firm's Tone
This scans your firm's past text messages to suggest a tone and starting instructions. It's optional: Engage works well without it, and skipping it saves setup time.
2: Identity & Tone
Agent Identity
Set an Identity Name
You can name it like a real team member (e.g., "Artie") or something generic. This is the name it'll use to introduce itself to leads.
Optional: Identity Role
This is used as another identifying option for the agent, like "Intake Coordinator."
Add Supervisors
These are the users who'll be notified if a message needs approval or a conversation needs to be escalated.
Communication Tone
Choose Professional, Balanced, Casual, or Empathetic. Use Custom Instructions to further refine how the agent should behave, and Never Say Rules to block specific words or phrases.
3: Engagement Settings
When to Engage
By default, Engage automatically activates when its linked Qualify profile scores a matter as Inconclusive. You can also turn on a toggle to activate on Low Confidence scores, which generally a good idea, since a low-confidence score usually means information is still missing.
Conversation behavior
Operating mode
Choose how much autonomy the agent has:
Manual Review
Every message is drafted and held for a supervisor to approve before it sends. This is the default, and we recommend starting here for your first few tries so you can see exactly what the agent will say.
Autopilot
Messages send automatically, no human review needed. Most firms move to Autopilot once they're comfortable with the messages Engage is generating.
AI Disclosure
Set your AI Disclosure preference: either Escalate to a Human immediately or Acknowledge & Confirm before moving on with the conversation. If you choose Acknowledge and Confirm, you have the option for a custom confirmation message:
Cadence Settings
Initial outreach and replies during an active conversation always go out immediately. For example, a lead who came in at midnight gets an immediate response.
Follow-ups to an unresponsive lead are the only part of the cadence that's time-restricted. Set how often a follow-up goes out (e.g., every 24 hours) and how long to keep trying before giving up (e.g., 48 hours).
Quiet hours apply only to those follow-ups. Think of it as a delivery window for nudges, not for live conversation. Turn these on if you don't want unresponsive-lead follow-ups going out overnight or on weekends:
Note: recurring follow-ups sent by text count against your firm's total SMS volume.
Engagement Summary
Determine if you want the agent to add a conversation summary to the matter's notes after engagement ends
Notifications
Supervisors will be notified when a conversation completed, was escalated to your team, or stopped because an exit condition was met
When to Stop Engaging
Matter-based exits
Set whether stage or status changes should stop the engagement
Conversation Limit
A soft cap on total messages; the agent wraps up gracefully rather than cutting off abruptly)
4: Test the agent
Before turning your agent on, use the Test Agent tab to simulate a conversation without sending real messages. Pick a matter that's currently Inconclusive in Merlin Qualify, and walk through the exchange. You can approve or revise each drafted reply. This is a safe way to see exactly what a live conversation will look like before any lead sees it and without affecting your SMS count.
Test mode never updates your actual matter data, so it's safe to run on any real matter.
5: Activate
Once you're happy with how the agent performs in testing, turn it Active from the upper right corner:
