From CRM to Custom Audiences: How AI Agents Build Lookalike Audiences Automatically

Building lookalike audiences manually is a multi-hour grind. You export CSVs from your CRM, clean the email column, hash them with SHA-256, upload to each platform separately, wait for matching, then create lookalikes. For a startup with 1,000–5,000 contacts, this takes hours per platform.

Multiply that by four platforms (Google, Meta, TikTok, LinkedIn) and you're spending an entire day on what should be a pipeline step. And the moment your contact list changes, the audiences are stale.

Synter eliminates the entire manual workflow. Its AI agents aggregate contacts from every source you have, deduplicate them, hash for compliance, upload to every platform in parallel, and create lookalikes once the seed audience reaches each platform's minimum threshold.

Here's what Synter pulled together in a single session for our own audience building:

TikTok matched approximately 20% of the seed audience (expected to increase as their identity graph processes the data over 24–48 hours). Meta typically matches 40–60% of email lists. With 3,243 unique emails, we cleared every platform's minimum threshold for lookalike creation on the first upload.

Each ad platform has a different audience upload API. Here's what Synter deals with so you don't have to:

Here's what Synter executes behind the scenes. These are the actual API patterns the agent uses for each platform:

Google Customer Match

# Google Ads API — OfflineUserDataJobService
# Synter handles this via google_ads_create_customer_match.py

POST /tools/run
{
  "script_name": "google_ads_create_customer_match",
  "platform": "GOOGLE",
  "args": [
    "--audience-name", "ICP Seed - Q1 2026",
    "--emails", "sha256_hash_1,sha256_hash_2,...",
    "--membership-lifespan", "540"
  ]
}

# The script:
# 1. Creates UserList with CRM_BASED membership type
# 2. Creates OfflineUserDataJob linked to the list
# 3. Adds hashed emails in batches of 100,000
# 4. Runs the job and polls for completion

Meta Custom Audiences

# Meta Graph API — Custom Audiences
POST /{ad_account_id}/customaudiences
{
  "name": "ICP Seed - Q1 2026",
  "subtype": "CUSTOM",
  "customer_file_source": "USER_PROVIDED_ONLY"
}

# Then populate with SHA-256 hashed emails:
POST /{custom_audience_id}/users
{
  "payload": {
    "schema": ["EMAIL_SHA256"],
    "data": [
      ["a1b2c3..."],  // pre-hashed
      ["d4e5f6..."]
    ]
  }
}

# Synter batches in groups of 10,000 for large lists

TikTok Audience Upload

# TikTok Ads API — 3-step audience creation

# Step 1: Upload hashed email file
POST /open_api/v1.3/dmp/custom_audience/file/upload/
Content-Type: multipart/form-data
file: emails_sha256.csv
file_signature: <md5_hash>

# Step 2: Create custom audience
POST /open_api/v1.3/dmp/custom_audience/create/
{
  "advertiser_id": "...",
  "custom_audience_name": "ICP Seed - Q1 2026",
  "audience_type": "customer_file",
  "file_paths": ["<path_from_step_1>"]
}

# Step 3: Wait for processing (polling)
# TikTok processes matches asynchronously — ~20% initial match rate

LinkedIn Matched Audiences

# LinkedIn Marketing API — DMP Segments
# Step 1: Create audience
POST /rest/dmpSegments
{
  "name": "ICP Seed - Q1 2026",
  "sourcePlatform": "FIRST_PARTY",
  "type": "USER",
  "destinations": [
    {"destination": "LINKEDIN"}
  ]
}

# Step 2: Add members (SHA-256 hashed emails)
POST /rest/dmpSegments/{segment_id}/users
{
  "elements": [
    {"userIds": [{"idType": "SHA256_EMAIL", "idValue": "a1b2c3..."}]},
    {"userIds": [{"idType": "SHA256_EMAIL", "idValue": "d4e5f6..."}]}
  ]
}

The audience pipeline doesn't exist in isolation. It plugs directly into Synter's growth engine, which feeds it with high-quality contacts from multiple discovery channels:

The output of the growth engine — a deduplicated, enriched contact list — becomes the input for the audience pipeline. Synter orchestrates both sides, so new contacts discovered today are in your ad platform audiences tomorrow.

Here's what a single agent session looks like. You tell Synter what you want, and it handles every step:

Agent prompt:
"Build a seed audience from our signups, waitlist, Attio CRM,
and Hunter enrichment for marketing/SaaS domains. Upload to
TikTok, Meta, Google, and LinkedIn. Create lookalikes for US."

Synter executes:
1. Query product DB → 963 emails
2. Query waitlist → 1,066 emails
3. Pull Attio contacts → 639 emails
4. Hunter domain search (27 domains) → 1,507 emails
5. Deduplicate → 3,243 unique emails
6. Validate email format → 3,201 valid
7. SHA-256 hash all emails
8. Upload to TikTok (3-step flow) → audience created
9. Upload to Meta (Custom Audiences API) → audience created
10. Upload to Google (OfflineUserDataJobService) → audience created
11. Upload to LinkedIn (DMP Segments) → audience created
12. Create lookalike on Meta (1% US) → ready
13. Create lookalike on Google (similar audiences) → ready
14. Check TikTok min threshold (1,000) → pending match processing
15. Check LinkedIn min threshold (300) → lookalike created

Total time: ~4 minutes
Manual equivalent: ~6 hours across 4 platforms

If you have 1,000–5,000 contacts, you're sitting on the most valuable targeting data you own. First-party data consistently outperforms interest-based and demographic targeting because these people already know your brand.

The problem was never the data — it was the operational overhead of getting it into ad platforms. Synter reduces that from hours of manual work to a single command, and keeps the audiences in sync as your contact list grows.

The audience pipeline is live today. Connect your data sources, tell Synter to build your audiences, and have lookalikes running across four platforms in under 10 minutes.

1. Create a Synter account (free, includes 1,000 credits)
2. Connect your ad platforms at Settings > Credentials
3. Tell the agent: “Build a seed audience from my signups and CRM, upload to Meta and Google, and create lookalikes for US.”

Questions? Reach out at joel@synterai.com.