AviaryAI LinkedIn Automation — Research & Solutions

Overview

As part of the ^[1] ABM program, the 30-day outreach sequence for Tier 1 accounts includes LinkedIn touchpoints (connection requests, messages) alongside email. While email automation via SendGrid is being implemented, LinkedIn automation remains an open problem due to platform restrictions. Both the Asymmetric team and Aaron Grossman are actively researching solutions.

Context

The ^[2] built for AviaryAI's 90 Tier 1 credit union accounts includes a per-account strategy document with a 30-day outreach sequence. Each sequence contains:

• Multiple email touchpoints (automated via SendGrid)
• LinkedIn connection requests and follow-up messages (currently manual)

Manually executing LinkedIn steps across 90 accounts — each contacted 7+ times over 30 days — is not scalable for a single rep (Aaron). Automation is the preferred solution, but LinkedIn's platform actively restricts third-party automation tools.

The Challenge

LinkedIn's terms of service prohibit most forms of automated activity, and the platform employs bot detection that can flag or restrict accounts using automation tools. This creates a meaningful risk for Aaron's personal LinkedIn account, which is the primary outreach vehicle. Any solution must balance:

• Scale: Reducing manual effort across 90+ accounts
• Safety: Avoiding account restriction or bans
• Authenticity: Maintaining the personalized, high-touch feel expected in ABM

Current Status

As of the meeting on 2026-02-11, no solution has been selected. Two parallel research tracks are underway:

Known Considerations

• Platform risk is real. LinkedIn has tightened enforcement. Any tool used should have a track record of safe operation and ideally simulate human-like behavior (randomized delays, daily send limits, etc.).
• Aaron's account is the asset. Since outreach comes from Aaron personally, a ban or restriction would directly damage pipeline activity. Conservative limits are warranted.
• LinkedIn steps may stay manual initially. If no safe automation solution is identified quickly, the contingency is for Aaron (or a hired assistant) to execute LinkedIn steps manually while email runs automatically. This was explicitly discussed as acceptable.

Candidate Tool Categories to Evaluate

The following categories are worth researching, though no specific tools have been selected:

• Chrome extension-based tools (e.g., Dux-Soup, Phantombuster LinkedIn agent) — operate within the browser session, lower detection risk but limited throughput
• Cloud-based LinkedIn automation (e.g., Expandi, Lemlist LinkedIn add-on, Waalaxy) — run outside the browser, higher throughput, variable safety records
• Sales engagement platforms with native LinkedIn steps (e.g., Outreach.io, Salesloft, Apollo) — may offer safer semi-automated LinkedIn tasks with manual approval gates
• HubSpot sequences with LinkedIn reminders — not true automation, but surfaces LinkedIn tasks in a queue for Aaron to execute efficiently

Fallback: Manual Execution with Task Queuing

If automation is delayed or deemed too risky, the team discussed a low-tech fallback:

A "minion" (hired assistant or intern) could be given access to Aaron's email and LinkedIn to execute the outreach steps from a task list. This removes the burden from Aaron while keeping a human in the loop, avoiding platform risk entirely.

Aaron confirmed he is comfortable with this approach if needed.

Next Steps

• [ ] Mark Hope — Complete research on LinkedIn automation tools; share findings and recommendation with Aaron and Justin
• [ ] Aaron Grossman — Poll sales network for tools currently in use; report back to team
• [ ] Team — Align on selected approach (automation tool vs. manual queue) before Tier 1 sequences launch

Related

• ^[3]
• ^[4]
• ^[5]