COO — Chief Operating Officer

The Chief Operating Officer (COO) runs on the Sonnet (workhorse) model tier at 5 credits per call. The COO is the practical, process- oriented efficiency expert: workflow optimization, the full hiring pipeline (job design, market research, posting, candidate screening), vendor management, quality monitoring, documentation (SOPs, runbooks), and process improvement.

Day in the life

The COO works against operational signals — workflow latency, vendor performance shifts, hiring requisitions, documentation gaps, and process metric drift. Workflow optimization runs a structured assessment: current-state, identified bottlenecks, recommended improvements with expected time/cost savings, and an implementation sequence. The workflow_optimized event fires when the analysis publishes.

The hiring pipeline is gated to Professional+ via the HIRING_ASSISTANT feature gate and runs three sequential sub-tasks: generate the job description (role summary, responsibilities, required and preferred qualifications, working conditions); research market compensation (salary ranges, benefits expectations, demand level, hiring timeline); and create the ready-to-post listing with mandatory AI disclosure. The Quality Gate validates the final job posting before it emits hiring_pipeline_completed.

Candidate screening scores each applicant 0-100 against the job requirements with strengths, gaps, an interview recommendation (yes/no/maybe), and a ranked shortlist. Vendor outreach (negotiate, contact, email) honors the 48-hour contact-spacing gate using vendor:{vendor_name} as the contact identifier when no explicit ID is provided.

Quality monitoring uses the monitor_quality tool to scan output quality across business functions and flags deteriorating areas. Documentation generation produces SOPs, runbooks, and process docs with version + date stamps, audience-appropriate language, edge cases, and troubleshooting sections. Process improvement lands a root-cause analysis, prioritized recommendations, expected metric impact, and explicit risk assessment.

Tools

Tools are registered via register_operations_tools in src/agents/operations/tools.py.

• _generate_job_description — Sub-handler that drafts the JD section of the hiring pipeline.
• _research_compensation — Sub-handler that pulls market rate ranges, benefits, demand, and timeline.
• _create_job_posting — Sub-handler that combines JD + compensation research into a ready-to-post listing with AI disclosure.
• monitor_quality (tool) — Scans output quality across business functions.
• ContactTimelineService.check_spacing / record_touch — Vendor outreach cooldown.
• quality_gate — Validates every external COO output (job postings, vendor communications, SOPs, process plans) before delivery.
• AI_DISCLOSURE constant — Appended to job postings and other external comms.

Decision patterns

These are the dispatch entries inside COOAgent.execute in src/agents/operations/agent.py.

• workflow_optimization — Triggered by latency/throughput drift; produces a structured improvement plan and emits workflow_optimized.
• hiring — Triggered by hiring requisitions; gated to Professional+ via HIRING_ASSISTANT; runs JD + market research + posting in sequence.
• candidate_screening — Triggered when applications arrive; produces scored candidates with shortlist.
• vendor_management — Triggered for evaluation, contact, negotiation; outreach actions honor the 48-hour cooldown.
• quality_monitoring — Triggered by the quality cadence; uses the monitor_quality tool.
• documentation — Triggered when SOPs/runbooks are needed; versioned and dated.
• process_improvement — Triggered when metrics drift outside targets; RCA + recommendations + impact + risk.

Escalation criteria

• Tenant plan lacks HIRING_ASSISTANT — hiring pipeline returns feature_gated.
• Vendor within the 48-hour cooldown — outreach deferred, not blocked.
• Quality Gate flagged the output (job posting, SOP, vendor email, process improvement plan) and auto-revise failed — content blocked.
• Action would exceed the COO’s daily budget cap.
• Action affects more than escalation_blast_radius_threshold (e.g. fires vendor renegotiation across more than the approved vendor count).
• Process improvement recommendation requires capital expenditure or organizational restructuring beyond the chairman-approved scope.

Example outputs

[Workflow optimization — onboarding sequence]

CURRENT STATE
- 9 manual steps, 3 different tools.
- Median time-to-first-value: 6.2 days.
- 22% drop-off between step 4 (KYC) and step 5 (integration setup).

BOTTLENECKS
1. KYC document upload requires email round-trip with support.
2. Integration setup blocks on Slack invite (manual).

RECOMMENDATIONS
1. Replace email-based KYC with in-app uploader (saves 1.4d median).
2. Auto-Slack-invite on KYC completion (saves 0.6d median).

EXPECTED IMPACT: time-to-first-value 6.2d → 4.2d; drop-off 22% → 11%.

IMPLEMENTATION
- Week 1: ship in-app uploader (CTO owns).
- Week 2: wire Slack auto-invite on KYC complete.

[Hiring pipeline — Senior Product Manager (full-time, remote)]

JOB DESCRIPTION
Senior PM owning the platform's onboarding and trust surfaces. Reports to
the founder. Three years of product leadership at a B2B SaaS
required; experience shipping AI-native products preferred…

MARKET RESEARCH
Salary range: $165K–$195K base + 0.4–0.7% equity (US remote, 2026 Q2).
Benefits expectations: full remote, $1,500 home-office stipend, 401(k)
match. Demand: high (search volume +18% YoY). Estimated hiring
timeline: 5–7 weeks.

— Posted on behalf of {founder_name} by their UltimateTeam.ai COO agent.

{
  "job_id": "req_2026_pm_001",
  "candidates_screened": 47,
  "interview_recommended": 6,
  "shortlist": [
    {"id": "c-12", "score": 92, "recommendation": "yes"},
    {"id": "c-04", "score": 88, "recommendation": "yes"},
    {"id": "c-31", "score": 81, "recommendation": "maybe"}
  ]
}

---

Sourced from src/agents/operations/agent.py and src/agents/operations/capabilities.py. Last reviewed: 2026-04-25.