How to Build a LinkedIn AI SDR Agent in Item 12 (Visual Builder)
· how-to
You do not need Claude, MCP, or code to build a LinkedIn AI SDR agent. Item 12's visual workflow builder lets founders and solo SDRs design outbound sequences, generate personalized DMs with AI prompt mode, and queue every send for human approval — all from the web app. This tutorial walks through creating your first agent from scratch.
What an AI SDR agent is in Item 12
An outbound agent is a visual workflow: contacts enroll → steps run in order (connect, wait, DM) → each LinkedIn action queues in the review inbox → you approve → Item 12 sends. Nothing auto-fires. AI prompt mode drafts unique copy per contact from your step instructions; static mode uses fixed templates with merge tokens.
Step 1: Create a new outbound agent
- Open Item 12 → Agents → Create outbound agent
- Name it (e.g. "Enterprise SDR Q3")
- Set target list or tag — only matching contacts can enroll
- Choose enrollment rules: manual enroll, on import, or when connection accepted
Step 2: Add connection request + wait steps
Drag a Send connection request block as step 1. Write a short note or enable AI prompt mode with instructions like "One sentence referencing {{company}} — no pitch." Add a Wait block (72 hours) before the DM step so you only message after accept.
Step 3: Configure the DM with AI prompt mode
- Add a Send message block after the wait
- Toggle AI prompt mode on the step
- Enter your prompt: tone, length, question to ask, what to avoid (links, calendar)
- Optional static fallback — if AI fails, send a safe template with {{first_name}} and {{company}}
Step 4: Enable human review before anything sends
Review queue is on by default for all outbound agents. Every connection note and DM appears in Review before send. Managers can edit copy, reject bad drafts, or approve batches. This is your LinkedIn account safety layer — do not disable it until you have tested copy on 20+ contacts.
Step 5: Enroll a test contact and approve
- Pick one contact from your target list → Enroll in agent
- Open Review queue — see pending connection request or DM draft
- Edit if needed → Approve → Item 12 sends via your connected LinkedIn account
- Check mirrored inbox thread on the contact record to confirm delivery
Static vs AI-generated copy: when to use each
Use static templates when compliance requires exact wording. Use AI prompt mode when personalization depth matters — different titles, industries, or recent activity per prospect. Most teams use AI for Layer 2 DMs and static for connection notes until copy is proven.
Next in series
Add human-in-the-loop governance for your team, or let Claude build agents via MCP. Continue with How to Build a Human-in-the-Loop AI SDR for LinkedIn and How to Build a LinkedIn AI SDR Agent with Claude and Item 12 MCP.
Does Item 12 auto-send LinkedIn messages?
No — every outbound from agents queues for review first. You approve before anything goes live.
What merge tokens are available?
Common tokens include {{first_name}}, {{last_name}}, {{company}}, and {{title}} — pulled from the contact record at draft time.
Are there rate limits on agents?
Item 12 enforces daily caps per workspace. Keep enrollments gradual when testing a new sequence.