AviaryAI's outbound strategy uses a three-tier ABM model to segment targets by personalization depth and send volume. The core problem driving this architecture: manual sending was capped at 100 emails/day — statistically insufficient to measure open rates or message effectiveness. The solution is a custom AWS SES automation layer that scales volume while preserving full HubSpot visibility.
This approach was finalized in the [1].
| Tier | Type | Volume | Personalization | Send Method |
|---|---|---|---|---|
| Tier 1 | Personal outreach | ~100/day (scaling) | High — individual strategy docs, custom email sequences | AWS SES (automated) |
| Tier 2 | Semi-personalized bulk | Up to 1,000/week | Medium — templatized with name/company merge | AWS SES |
| Tier 3 | Broad market blast | 1,000+/week | Low — industry-level personalization only | AWS SES |
Decision: Build a custom AWS SES solution rather than using SendGrid or HubSpot's native email tools.
Rationale:
- Avoids paid tool costs (HubSpot bulk email requires Marketing Pro; SendGrid adds another vendor)
- AWS SES can handle millions of emails/day at near-zero marginal cost
- Integrates with HubSpot via webhooks for full activity tracking (sent, opened, clicked)
- Sends can be logged back to HubSpot contact/deal records with full email body included
Key requirement from client: Blessing (AviaryAI's internal stakeholder) reviews all outreach activity in HubSpot. The system must log the full email body — not just open/click events — mirroring what the existing Outlook integration provides. This was the deciding factor in the architecture.
"He would want to have the emails in HubSpot. Not just, did they send an email, did they open the email? He would like to see the body of the email." — Aaron Grossman
Aaron does not review individual emails before send. The expectation is: review the strategy docs and email tone once, approve the flow, then let it run and iterate based on results.
The existing 71k-contact HubSpot database must be cleaned before high-volume campaigns can run. Key issues:
- ~20k contacts are VC firm records (irrelevant to ICP) added in a prior data dump
- Many contacts have broken/missing names from a failed automation attempt
- Email addresses are stale — contacts have changed jobs, leaving invalid addresses that will bounce
Cleanup scope: deduplicate, fill missing names via Clay/Apollo, categorize into Credit Unions / Community Banks / Insurance Companies / Other, run email hygiene validation.
See [2] for full scope.
This tier architecture applies beyond AviaryAI. The core insight:
When a client's ICP is large and well-defined (e.g., "all U.S. credit unions"), a tiered send model lets you run high-touch Tier 1 outreach in parallel with high-volume Tier 3 market coverage — using the same infrastructure.
The AWS SES + HubSpot webhook pattern is reusable for any client where:
- HubSpot is the CRM of record
- Stakeholders need email visibility without a separate tool login
- Send volume exceeds what HubSpot's native email or a paid ESP can handle cost-effectively
See also: [3]