ref/AgenticSwimlanes.com
menu
PROCESS / SALESLast verified: October 2026

AI Agent Sales Pipeline Workflow: Cited Swimlane (2026)

Sales pipelines split cleanly into agent-suited steps (data lookup, enrichment, scoring, drafting) and human-suited steps (judgement on fit, relationship management, the actual conversation). The canonical swimlane keeps every step before the send in the agent lane, places the send-decision in the human lane, and lets a low-score path divert into a nurture sequence with no human touch.

SALESAI AGENTSALES REP (HUMAN)CRM / EMAILNew leadEnrichfirmographicsScore intentDraft outreach×confidence?Review andapprovelowAdd to nurturesequenceSend emailLog activity
Sales pipeline swimlane. Agent lane handles enrichment, scoring, and draft generation as bpmn:serviceTask steps. Sales-rep lane carries the review-and-approve bpmn:userTask. CRM/email lane carries the send and the activity log.Source: pattern observed in Anthropic customer stories, claude.com / customers (accessed October 2026); Clay AI agent documentation, university.clay.com (accessed October 2026); 11x product pages, 11x.ai (accessed October 2026). The review-and-approve user-task is this reference's recommended shape on reversibility grounds, not a step either vendor documents. Shape conformance per OMG BPMN 2.0 §10.3, §10.6.

One named case study

Clay is the clearest published example of the agent lane above. In Anthropic's Clay case study (accessed October 2026), Claygent, described as the platform's AI research agent, “can scrape the public web to find a prospect's thought leadership: podcast appearances, blog posts, media mentions” and then “generates a highly personalized email that references the lead's thought leadership”. That is enrichment, scoring input and drafting, all inside the agent lane, which is exactly the span the swimlane above assigns to the agent.

The gate is where the published record and the recommended shape part company, and it is worth being precise about it. 11x markets Alice as an outbound digital worker that “engages prospects across channels, handles replies, drives qualified meetings and builds pipeline”, running on what the product page calls autopilot (accessed October 2026). Neither 11x nor the Clay case study publishes a human approval step before outreach is sent, and neither publishes a confidence threshold that would trigger one. The review-and-approve user-task in the diagram above is therefore the shape this reference recommends, on the reversibility argument below, and not a step either vendor documents. Any page that attributes a send-time human gate to these products is describing something their documentation does not say.

Where the human gates sit

One soft gate (review and approve before send) on the high-confidence path. The low-confidence path bypasses the human gate by routing into a nurture sequence (no immediate outreach). The gate is reversible (an unsent draft can be edited or discarded), so a soft gate is appropriate; irreversibility appears only on send.

Where the handoffs sit

Three handoffs: agent enrichment to agent scorer (sequence flow within agent lane), agent draft to sales rep (sequence flow into human lane), and approved draft to email system (sequence flow into the system lane). All within a single pool. There is no cross-pool message flow in the simplified diagram; in production, handoffs to external enrichment-data providers are message flows out of pool boundary.

Workforce-impact note

The defensible measure is hours-per-rep saved on outbound preparation rather than rep-replacement. Published case studies report drafting and enrichment time falling by an order of magnitude; the rep-as-relationship role remains. For the methodology, see aijobimpactcalculator.com.

Related pages