---
title: "From CRM to Custom Audiences: How AI Agents Build Lookalike Audiences Automatically"
description: "Synter aggregates contacts from your CRM, product DB, and enrichment tools, deduplicates and hashes them, uploads to TikTok, Meta, Google, and LinkedIn, and creates lookalike audiences — all automatically."
date: "2026-03-22"
draft: false
tags:
  - "Audiences"
  - "Growth Engineering"
  - "AI Agents"
---

## The Problem: Manual Audience Building Doesn't Scale

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&apos;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&apos;s minimum threshold.

## How Synter Solves This: The 5-Step Pipeline

**Step 1: Aggregate**

Synter pulls contacts from multiple sources simultaneously: your
product database (signups, waitlist), CRM (Attio, HubSpot), and
enrichment tools (Hunter, Apollo, BuiltWith). No manual exports.
Synter connects directly to each source and pulls in parallel.

**Step 2: Enrich**

Synter uses Hunter domain search to find marketing professionals
at ICP companies — agencies, SaaS, martech. In one session we
pulled 1,507 emails from 27 domains. Apollo fills in titles and
seniority. BuiltWith identifies tech stack for targeting
precision.

**Step 3: Deduplicate & Hash**

All sources are combined, deduplicated by normalized email,
validated for format (reject malformed addresses), and SHA-256
hashed for platform compliance. Every ad platform requires
hashing before upload — Synter handles this automatically.

**Step 4: Upload**

Batch upload to ad platforms (TikTok, Meta, Google, LinkedIn)
via their native APIs. Synter handles the wildly different API
quirks per platform — TikTok&apos;s three-step file upload flow,
Meta&apos;s batched Custom Audiences API, Google&apos;s
OfflineUserDataJobService, LinkedIn&apos;s DMP segments.

**Step 5: Create Lookalike**

Once the seed audience reaches each platform&apos;s minimum
(1,000 for TikTok, 100 for Meta, 1,000 for Google, 300 for
LinkedIn), Synter creates lookalike audiences for each target
geo. The entire pipeline runs from a single agent command.

## Real Numbers from Our Pipeline

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




| Source | Emails | Notes |
| --- | --- | --- |
| Product DB (signups) | 963 | Direct from PostgreSQL |
| Waitlist | 1,066 | Pre-launch signups |
| Attio CRM | 639 | Sales pipeline contacts |
| Hunter (27 marketing/SaaS domains) | 1,507 | ICP enrichment |
| Total unique after dedup | 3,243 |  |



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&apos;s minimum
threshold for lookalike creation on the first upload.

## Platform-Specific Quirks Synter Handles

Each ad platform has a different audience upload API. Here&apos;s
what Synter deals with so you don&apos;t have to:


**TikTok: 3-Step File Upload**

TikTok requires a file upload to create the audience, then a
separate call to create the custom audience object, then a batch
append call to populate it. Most platforms handle this in one
call — TikTok needs three.

Min lookalike size: 1,000 matched users

**Meta: Custom Audiences API**

Meta uses SHA-256 hashed emails sent in batches of 10,000 via
the Custom Audiences API. Synter handles the batching, hashing,
and progress tracking. Match rates typically hit 40–60%.

Min lookalike size: 100 (recommend 1,000+)

**Google: OfflineUserDataJobService**

Google Customer Match uses the OfflineUserDataJobService — a
job-based API where you create a job, add operations in batches,
then run the job. Synter manages the full lifecycle.

Min lookalike size: 1,000 matched users

**LinkedIn: DMP Segments**

LinkedIn Matched Audiences use DMP segments with a two-step
create-then-populate flow. LinkedIn also supports company name
matching for ABM lists, which Synter leverages when emails
aren&apos;t available.

Min lookalike size: 300 members

## Under the Hood: API Calls

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


Google Customer Match




Meta Custom Audiences




TikTok Audience Upload



# Step 2: Create custom audience
POST /open_api/v1.3/dmp/custom_audience/create/

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


LinkedIn Matched Audiences

## The Growth Engine Integration

The audience pipeline doesn&apos;t exist in isolation. It plugs
directly into Synter&apos;s growth engine, which feeds it with
high-quality contacts from multiple discovery channels:


**Hiring Signals**

Synter monitors job postings for companies hiring for marketing,
growth, and paid media roles. These companies are actively
investing in advertising — ideal prospects for your seed
audience.

**Hunter: Email Discovery**

Domain search across your ICP verticals. In our pipeline, 27
marketing and SaaS domains yielded 1,507 verified professional
emails. These feed directly into the audience upload.

**BuiltWith: Tech Stack**

Identify companies using competitor tools or complementary
technology. A company running HubSpot + Google Ads but not a
campaign management platform? That&apos;s your ICP.

**Apollo: Contact Enrichment**

Fill in titles, seniority levels, and direct emails for contacts
discovered through other channels. Richer data means better
matching and more precise lookalikes.



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.

## The Full Pipeline in One Command

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

## Why This Matters for Startups

If you have 1,000–5,000 contacts, you&apos;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.


Pro tip: Start with your highest-intent contacts
as the seed audience — paid customers and active trial users. The
lookalikes will be significantly better than seeds built from
newsletter subscribers or cold outreach lists.

## Get Started

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 trial included)
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

.
