Why This Role, Why Now
GTM Data Strategy & Operations stood up from scratch with no predecessor. Today the function runs on three offshore contractors and zero FTEs, managed by a single leader who is simultaneously building the agentic infrastructure, operating it in production, and driving major initiatives (hierarchy redesign, data quality assessment, vendor optimization).
The operating model is deliberately agentic AI–first: a multi-agent pipeline (Cartographer, Sentinel, Resolver, Reporting) handles detection, enrichment, hierarchy mapping, and conflict resolution at scale. This is not a future-state vision, these agents are live and processing enterprise account families in production today.
The problem: one person cannot build, operate, and extend this system while also managing strategic workstreams. The function currently covers only core Tier‑1 fields. Dozens of account, contact, and lead signals remain unaddressed. Every pipeline run, every failure diagnosis, and every offshore handoff flows through a single point of failure.
This role is the first onshore execution hire for an agent operator who can keep the system running, improve it, and extend detection and resolution coverage as GTM leadership prioritizes new data elements.
Role Summary
Sit between AI systems and GTM data. Operate, tune, and extend our agentic data quality pipeline (detection, enrichment, hierarchy mapping, conflict resolution) so it runs reliably, improves continuously, and expands to cover more of the data landscape. Own the handoff between automated output and human review, managing quality and throughput with our offshore team. You don’t build agents from scratch, but you run them, evaluate their output with GTM data judgment, and make them better.
Core Responsibilities
Agent Pipeline Operations
- Run and monitor production pipeline sessions (Cartographer, Sentinel, Resolver) across scheduled cadences; diagnose and resolve failures (API errors, session timeouts, data anomalies) without escalating to the function lead.
- Execute pipeline runs in Claude Claude and tmux; manage long-running batch processes; interpret logs and output to confirm data integrity before downstream handoff.
- Maintain pipeline orchestration scripts and configuration; extend agent coverage as new data elements are prioritized by GTM leadership.
Agent Tuning & Improvement
- Refine detection rules, prompt logic, and confidence thresholds based on output analysis and false-positive/negative patterns.
- Evaluate agent accuracy by segment (Enterprise vs. MM/SMB) and recommend rule or workflow changes backed by evidence.
- Run bake-offs (vendor vs. AI enrichment) to optimize cost, coverage, and accuracy; document results for decision-making.
Sentinel → Offshore Resolution Loop
- Own the handoff between Sentinel detection output and Concentrix triage queues; define queue structure, priority tiers, and resolution instructions.
- Monitor offshore resolution quality and throughput; refine detection rules based on patterns surfaced through triage.
- Close the feedback loop: track resolution outcomes back to agent configuration to reduce recurring false positives and improve detection precision.
Data Quality & Enrichment Operations
- Maintain ops-only staging fields; manage the promote-to-production flow with audit controls.
- Design and run AI-assisted enrichment workflows (Clay + LLM prompts) with evidence links and confidence thresholds.
- Monitor fill-rate, sampled accuracy, freshness, and cost-per-record by source and segment; surface vendor performance issues and recommend changes.
- Keep data dictionaries, SOPs, and runbooks current as agents and processes evolve.
Cross-Functional Partnership
- GTM Systems (SFDC): field configuration, permission sets, automation, flows.
- Data Engineering: source availability, ID mapping, lineage (no pipeline coding).
- Reporting: define metrics and acceptance criteria; partner on dashboard requirements.
What to Expect
This is a triage environment, not a steady-state one. The function is young, the data has known gaps, and the work is to stabilize and extend, not maintain and optimize. You’ll be building the plane while flying it, alongside a small team that operates with high autonomy and a bias toward measurable outcomes. If ambiguity and mess energize you, this is the right fit.
Success Metrics (6–12 Months)
Pipeline Reliability
- Scheduled pipeline runs execute without function-lead intervention; failure-to-resolution cycle time under 24 hours for non-blocking issues.
- Agent coverage extended to new data elements as prioritized (measured by number of signals under active detection).
Detection & Resolution Quality
- Sentinel detection precision and recall improve quarter over quarter, tracked by segment.
- Concentrix resolution queue throughput and accuracy meet defined acceptance thresholds.
- False-positive rate decreases through feedback-loop refinement.
Data Quality Outcomes
- Tier-1 field fill-rates: Country ≥95%; Vertical ≥90% at ≥85% sampled accuracy; Revenue bands ≥90%.
- Hierarchy coverage 65–80%+ across target segments.
- Enterprise cost-per-record reduction of 30–40% via AI-first + selective vendor usage.
Qualifications
Required
- 3–6 years in Data Ops, Sales Ops, or GTM Ops with hands-on data quality ownership for account and contact data.
- Proficiency with Snowflake (SQL for querying, analysis, validation) and SFDC (object model, field configuration, data flows).
- Working experience with Claude Code or comparable LLM-based tooling in an operational (not just experimental) context.
- Experience designing and running AI-assisted enrichment workflows (e.g., Clay + LLM prompts) and evaluating accuracy/coverage.
- Comfort operating in a command-line environment: tmux, shell scripts, log analysis, batch process monitoring.
- Process design mindset with a bias toward measurable outcomes; strong written communication.
Strong Plus
- Experience with account/contact data vendors (D&B, ZoomInfo, Clearbit, StoreLeads) and waterfall enrichment logic.
- Python for QA scripting, sampling, or light automation.
- Familiarity with prompt engineering, confidence scoring, and AI guardrails (evidence capture, versioned prompts, QA sampling gates).
Tool Stack
- Core: Snowflake (SQL), SFDC, Claude Code, Clay
- Pipeline: Shell orchestration, Cartographer / Sentinel / Resolver agents
- Enrichment: D&B, ZoomInfo, Clearbit, StoreLeads, LLM prompts
- Nice to Have: Python, SOQL, prompt engineering frameworks
- AI Guardrails (Expected Practice): Confidence floors, evidence capture, versioned prompts, 10% QA sampling gates, audit-on-promote, drift alerts, and privacy/compliance checks. This role is expected to uphold and improve these practices, not just follow them.