Outreach is the sales-engagement platform for enterprise SDR teams. The AI workflow that survives at 50+ reps is narrow: agents triage sequence replies, tier accounts against Salesforce, and write post-call summaries through the Outreach REST API. Outreach Kaia transcribes the meeting. Salesforce stays system of record. The SDR keeps reply judgment. The handoff to the AE carries the research brief and the next-step decision, not just a calendar invite. Everything below assumes that split.
The five-step workflow
1. Tier accounts before sequencing. Pull the target list from Salesforce, score it with a ChatGPT or Claude prompt against ICP fit, firmographic signal, and intent (Common Room, 6sense). Write the tier back to a Salesforce custom field. Only Tier 1 and Tier 2 get hand-built openers. The Apollo prospecting playbook covers the equivalent flow for mid-market teams.
2. Build sequences in Outreach, draft openers with an agent. Sequence steps, send windows, and A/B variants live in Outreach. The agent drafts step-1 personalization from the research brief, the SDR edits, then the SDR triggers sequenceStates.create through the API. Volume-first is a trap. The Bridge Group SDR Metrics & Compensation Report, which benchmarks 434 B2B companies, shows reply rates fall off a cliff past 200 contacts per rep per week. The email sequencing breakdown covers tone and cadence rules.
3. Triage replies with an agent, not a folder rule. Pull mailings and prospects updated in the last hour. Classify each reply: positive, objection, referral, unsubscribe, out-of-office. Positive replies get a draft response and a calendar link. Objections route to the SDR with the relevant talking point pulled from win/loss notes. Unsubscribes go straight to suppression.
4. Capture every call with Outreach Kaia. Kaia (or Gong, if that is the recorded source of truth) transcribes the discovery call. An agent extracts pain, decision criteria, competitors mentioned, next step, and timeline. That structured summary is what the AE actually needs. See meeting prep workflows for the inbound discovery equivalent.
5. Hand off to the AE with state, not vibes. When an SDR books a meeting, the agent assembles a handoff packet: research brief, sequence history, reply transcript, Kaia summary, and a recommended opening question. The AE reads three paragraphs instead of scrolling Salesforce activity.
Worked example: an SDR sequence on a Tier 1 account
A 200-rep SDR org runs Outreach against Salesforce. An agent watches replies every fifteen minutes. A VP of RevOps replies to step 3 with "send me the security overview." The agent classifies it as a positive reply, drafts a response with the SOC 2 link and three discovery-call slots, and surfaces both in the SDR's Outreach inbox. The SDR sends in 90 seconds. The meeting books. Kaia transcribes. The agent writes the post-call summary. The AE opens the handoff packet before the follow-up.
The persistent-state pain
Here is what breaks at scale. The research brief lives in a Google Doc. The reply rationale lives in the SDR's head. The Kaia summary lives in Outreach. The next-step decision lives in a Slack message the AE missed. Salesforce shows "Meeting Booked" and nothing else. When the AE opens the account, half the context is gone.
One way to solve this is a workspace like Dock that holds what the agent interprets around the deal: the prospect-research brief, the personalization rationale, the reply triage decision, and the post-call summary. Each row carries a salesforce_account_id and an outreach_prospect_id pointer. Salesforce stays the system of record for stage and forecast. Outreach stays the system of record for sequence state. Dock holds the interpretive layer the AE actually reads.
Why it matters
Enterprise SDR ops fail at the handoff, not at the sequence. The agent's job is to compress what the SDR learned into a shape the AE can act on. That requires a persistent home for interpretation with a stable agent identity writing into it.
The full pillar is here: AI sales prospecting that actually compounds.
FAQ
Does Outreach have a native AI assistant? Yes. Outreach Kaia handles meeting transcription and conversation intelligence. For sequence triage, account tiering, and post-call summarization, teams layer ChatGPT or Claude on top of the Outreach REST API.
Should AI send replies automatically? No. Drafts only. The SDR reviews and sends. The reply rate cost of a bad auto-send is higher than the time saved.
Where do post-call notes belong: Salesforce, Outreach, or somewhere else? The structured fields belong in Salesforce. The narrative summary belongs where the AE will actually read it before the next call. That is usually not Salesforce.
How is this different from the Apollo workflow? Apollo is data-plus-engagement for mid-market. Outreach is engagement-only at enterprise scale with deeper Salesforce sync, stricter compliance, and Kaia for conversation capture. The handoff rigor matters more because the AE pool is larger.