How to Build a Multi-Step LinkedIn SDR Sequence (Connect → DM → Follow-up)
· how-to
A LinkedIn AI SDR agent is only as good as its sequence design. Before you touch a workflow builder or MCP prompt, map the three layers every high-performing outbound program uses: connect → value DM → follow-up bump. This tutorial walks through timing, copy strategy, merge tokens, and stop conditions — then shows how to configure each layer in Item 12 outbound agents.
The 3-layer SDR sequence architecture
- Layer 1 — Connection request: one-line relevance hook, no pitch (Day 0)
- Layer 2 — Value DM after accept: single question or insight, not a calendar link (Day 2–3 post-accept)
- Layer 3 — Follow-up bump: short nudge or resource offer, then stop (Day 7–10 if no reply)
Recommended timing table
| Tool | Category | Best for | Price | Key strength |
|---|---|---|---|---|
| Connection request | Layer 1 | Day 0 — send on enroll | Wait: none | Max 300 chars; reference company or mutual context |
| Post-accept DM | Layer 2 | 2–3 days after connection accepted | Wait: 48–72 hours | Ask one question; avoid pitch deck links |
| Follow-up bump | Layer 3 | 7–10 days after Layer 2 if no reply | Wait: 5–7 days | Offer resource or polite close; max 1 bump |
Step 1: Configure the connection request step
In Item 12 Agents, add a Send connection request step as the first block. Use static copy for the note when your ICP is narrow and messaging is locked — or switch to AI prompt mode so each contact gets a draft from their profile and company context. Merge tokens like {{first_name}} and {{company}} keep templates readable. Set enrollment to skip contacts already connected.
Step 2: Add a wait step before the value DM
Insert a Wait step between connection and DM — 48–72 hours is the sweet spot for B2B. The wait runs only after the connection is accepted; contacts who never accept never reach the DM step. Enable stop-on-reply so any inbound message halts the sequence and tags the contact for manual follow-up.
Step 3: Write the value DM (static vs AI prompt mode)
- Static template — best when legal or brand requires exact wording on first touch
- AI prompt mode — best when personalization depth varies by prospect (title, company news, recent post)
- Merge tokens — {{first_name}}, {{company}}, {{title}} for baseline personalization in both modes
- Prompt example — "Write a 2-sentence DM thanking them for connecting. Reference their role at {{company}} and ask whether [pain point] is a Q3 priority. No links."
Step 4: Add conditions and stop rules
- Skip if already connected — avoid duplicate invites
- Stop if tagged replied — halt sequence when contact responds
- Stop if pipeline stage is Meeting or later — do not bump active deals
- Max attempts — cap follow-up bumps at one to protect account health
Step 5: Track metrics that matter
Benchmark connection accept rate (25–45% for strong ICP lists), reply rate on Layer 2 DMs (8–15%), and meetings booked per 100 enrollments. Item 12 logs inbox replies beside contact records — compare reply tags against enroll date to tune wait durations and copy. If accept rate is low, fix the connection note before adding more follow-up steps.
Next in series
Once your sequence logic is set, build it in the visual agent builder or add human review guardrails before going live. Continue with How to Build a LinkedIn AI SDR Agent in Item 12 (Visual Builder) and How to Build a Human-in-the-Loop AI SDR for LinkedIn.
How many steps before LinkedIn limits kick in?
Keep sequences to 3–4 touchpoints over 2–3 weeks. Item 12 daily caps and review queues add safety; volume without personalization raises restriction risk regardless of tool.
Best length for a first LinkedIn DM?
Two to four sentences — under 400 characters. One question beats three paragraphs. AI prompt mode in Item 12 can enforce brevity in the step prompt.
Can I build this sequence via MCP instead of the UI?
Yes — Claude can call workflow_create with the same step order. See How to Build a LinkedIn AI SDR Agent with Claude and Item 12 MCP.