How to Run a LinkedIn Outreach Campaign Targeting AI Engineers Who Hate Model Versioning (2026)

Quick answer: Once your list of AI engineers who hate model versioning is ready inside Origami, you can launch a targeted LinkedIn campaign using Origami's built-in sequencer—no exporting, no tool-switching. In this guide, I'll walk you through the exact steps I've used to turn enriched contacts into conversations, including a 3-touch sequence you can copy-paste and send directly from Origami.

Your list is built. You've followed the how to build a list of Selling to AI Engineers Who Hate Model Versioning guide, used Origami's AI agent to find and enrich leads with verified emails, phone numbers, titles, and company details, and you're staring at a pool of potential prospects. Now comes the part that separates a random batch of cold messages from a campaign that books meetings: refining your list for LinkedIn, crafting a sequence that speaks to their real pain, and sending it all from a single platform. Here's how to do it, step by step.

Step 1: Refine your AI engineer list for LinkedIn outreach

Before you fire off a single connection request, you need to slice your list into segments that match who you want to reach and how you want to talk to them. In Origami, you aren't stuck with a static CSV; the enriched data lets you filter, tag, and qualify leads without leaving the platform.

What "qualified" looks like for AI engineers who hate model versioning

Not every ML-related title struggles with versioning, but the pain is concentrated in roles that actually touch experiments and production models. Here's the segment you want:

• Job titles: Machine Learning Engineer, AI Engineer, MLOps Engineer, Data Scientist (who builds models, not just dashboards), Applied Scientist, Research Engineer, Deep Learning Engineer, or even Head of AI. Steer clear of pure Data Analysts or ML recruiters.
• Company size: Startups with 20–200 employees often have the worst versioning chaos because they lack dedicated infrastructure; mid-market companies (200–1,000) are growing fast and need scalable solutions; large enterprises have complex governance headaches. All three are good, but the messaging will differ—I'll show you how to handle that in the sequence.
• Industry signals: Look for companies in autonomous vehicles, fintech, healthcare AI, SaaS with ML features, any place where models go to production. In Origami, you can filter by "Industries" or use the AI to tag leads whose company website mentions terms like "model registry," "ML pipeline," or "experiment tracking."
• Buying trigger indicators: Engineers who recently changed jobs (last 12 months) or work at companies with recent funding rounds are often open to new tools. Origami enriches contact details and can pull funding data if you've chained the right sources.

How to segment the list inside Origami

I usually export nothing. Instead, I create smart segments right on the list view:

1. Role buckets: Use the filter bar to group by the "title" field. Save a view called "IC ML Engineers" (individual contributors) and another called "ML Leadership" (heads of AI, directors). You'll sequence them separately because a manager's versioning pain is more about team productivity, while an individual engineer cares about the daily friction.
2. Company size tiers: If you've enriched company employee count, create segments for <50, 50–200, 200–1,000, and 1,000+. Adjust the urgency and examples in your outreach accordingly.
3. High-intent signals: I tag leads whose enriched profile shows they use tools like MLflow, DVC, or Weights & Biases but might still hate the overhead—those are perfect. Origami's AI can automatically apply tags like "model-versioning-pain" based on tech-stack keywords if you ask it.

Once you have at least 150–200 leads in a segment, it's worth launching a sequence. For your first campaign, I recommend targeting the "IC ML Engineers at 50–200 employee companies" segment—they feel the pain most acutely and are likely to respond to a tool that removes manual busywork.

Step 2: Write the LinkedIn outreach sequence

Two ways to build your sequence in Origami

When you hit "Sequence" on your segmented list, Origami gives you two options:

1. Paste your own templates: You write the messaging, define the delays, and Origami sends each touch automatically. You can use variables like {first_name}, {company}, and {title} that fill from the enriched profile.
2. Let the AI agent write it for you: If you want, you can ask Origami's agent to generate a personalized 3-touch LinkedIn sequence based on each lead's actual role, company, and industry. It reads the enriched data and writes messages that feel human—no generic copy-paste.

Below, I'm giving you the exact 3-touch sequence I've used to engage AI engineers who hate model versioning. You can paste these templates right into the sequencer, tweak a few details, and go. Each message is direct, under 100 words, and references the real frustration these engineers feel every day.

The full 3-touch sequence for AI engineers who hate model versioning

Touch 1 – Day 1: Connection request note (300-character limit, keep it tight)

Hi {first_name}, saw you're an AI engineer at {company}. The model versioning circus is real—mixing up experiments, lost lineage. I built a lightweight tool that auto-versions every run with zero config. Worth connecting?

Why it works: It names the pain without jargon, mentions a solution that sounds effortless, and ends with a low-friction ask.

Touch 2 – Day 3: Follow-up message (send as a regular message after they accept your connection)

Hey {first_name}, thanks for connecting. Curious: how much time do you waste just tracking which model version did what? Teams using our platform cut that by hours each week. I can share a 2-minute video showing how it grabs every hyperparameter, dataset hash, and artifact automatically—no extra work. Worth a quick look?

Why it works: It quantifies the pain, offers social proof without naming a competitor, and dangles a no-effort demo.

Touch 3 – Day 7: Soft close (final message, still no pitch)

Wanted to leave you with one example: a team of four AI engineers like yours went from "which version is production?" to deploying with full lineage in days. If model versioning still makes you cringe, I'd be happy to do a 5-minute walkthrough—no pitch, just product. Let me know a day that works, or I can send a Loom.

Why it works: It humanizes with a concrete story, frames the ask as a quick look, and gives them an off-ramp (a Loom) that feels even easier.

Customizing for different segments

• For ML leadership: in Touch 2, swap "how much time do you waste" to "how much time does your team lose on manual version tracking?" In Touch 3, mention scaling pain: "a team of fifteen cut model release delays by 60%."
• For enterprise engineers: highlight governance and audit trails: "auto-captures everything needed for SOC2 audits."
• For startup engineers: emphasize speed and simplicity: "zero config, works with your existing notebooks."

You can paste these variants as separate templates in Origami and assign the right one to each segment.

Delays

Set Touch 1 (connection request) immediately on Day 1. Touch 2 fires on Day 3 after connection accepted. Touch 3 fires on Day 7. Origami's sequencer automatically handles the timing; you just pick the intervals.

Step 3: Launch the sequence directly from Origami

Here's where the platform really shines. You don't export the list, don't import it into another tool, don't deal with LinkedIn automation nightmares. Everything happens inside Origami.

1. Go to your prospect list and select the segment you refined (e.g., "IC ML Engineers 50-200").
2. Click "New Sequence" and choose "Paste my own templates."
3. Paste the three messages into the respective touch inputs—Touch 1 (connection request), Touch 2, Touch 3.
4. Set delays: Connection request sent now, follow-up after 3 days, final message after 7 days total. You can adjust these for your campaign.
5. Hit "Launch Sequence."

The built-in LinkedIn sequencer then sends connection requests and follow-up messages automatically, respecting the delays you configured. Sending is free on all paid plans—you only pay for the credits you use to enrich leads, not for the outreach itself. No extra per-message fees, no volume caps beyond LinkedIn's normal limits (which Origami manages responsibly).

What you'll see in the dashboard

As the sequence runs, the same dashboard that showed you the list now shows live activity:

• Connection acceptance rate: How many accepted your request.
• Messages opened, clicked, and replies: See exactly who engaged.
• Automatic un-enrollment: If a lead replies to any touch, Origami removes them from the rest of the sequence—no accidental "just following up" after they said "yes."
• Full prospect context: While looking at a reply, you can see their enriched profile (title, company size, tech stack) right next to the conversation, so you remember why you reached out and can respond intelligently.

What response rates to expect

With a list of AI engineers who genuinely hate model versioning and a sequence that names that pain, here's what I typically see in 2026:

• Connection acceptance: 30–40% for highly targeted, personalized notes.
• Reply rate (on Touch 2 or 3): 12–18%.
• Meeting booked: 5–8% of the original list.

These numbers assume your sender profile looks credible (a title like "Founder" or "ML Solutions Consultant" works), the list is tight, and you're sending to 200+ leads per segment. If you're below these thresholds after 100 sends, iterate.

When to iterate on the messaging vs. iterate on the list

• Low connection acceptance? Your note might be too generic, or your LinkedIn profile doesn't look like someone an engineer would connect with. Add more personalization in the note (Origami can pull recent job changes or mutual connections) or improve your headline.
• Low reply rate but healthy connections? Test different angles. Keep one sequence as is, create a variant that emphasizes "reproducibility" instead of "time savings," and A/B test on different halves of the same segment. Origami lets you run multiple sequences on the same list.
• High replies but low meeting bookings? Your soft close might be too pushy. Try the Loom-first approach: Touch 3 simply offers a 2-minute screen recording with no next-step—some will book a call after watching.
• All metrics low? Go back to the list. Run Origami's AI on your ideal customer description again with more specific filters, or exclude companies that appear too enterprise (some engineers there have versioning locked down by internal platforms).