Principal AI Systems Product Manager - Contract to Hire

Principal AI Systems Product Manager (AI Agents, Cross-System Intelligence, CRM + Operations Automation) Read This First (Hard Filter) This is not a prompt writer, no-code automations, or “AI assistant” role. If your experience is limited to: Writing prompts Using ChatGPT as a tool Zapier-only workflows Shipping isolated AI features without system ownership Do not apply. This role is for someone who designs, owns, and delivers AI systems as products, translates executive vision into coherent AI capabilities, and leads engineers to build them correctly. The Mission We are building an internal AI operating system for a real estate private credit and private equity firm. Your responsibility is to own the product vision, system design, and execution of this AI platform — ensuring it becomes a reliable, extensible layer that operates across the firm’s core business tools. This system must: Read across Asana, Cloze.com, Gmail, Airtable, Dropbox Reason across time, commitments, relationships, and deals Surface risk: missed follow-ups, stalled deals, broken promises Write back into systems safely, with guardrails Run on triggers and schedules Evolve into a cohesive, cross-system AI product ecosystem Think Jarvis for a real operating business, not a chatbot. Your Role (What You Will Actually Do) You are not the primary coder. You are the AI systems product owner responsible for turning business intent into production AI capabilities and guiding a team of fractional / contract AI engineers to deliver them. 1. Translate Executive Vision into AI Products Work directly with the CEO to understand: Business priorities Risk points Decision-making bottlenecks Define what AI should do, not just how it’s built Break vision into concrete AI-enabled products and capabilities 2. Own the AI System Architecture (at a Product Level) Define the overall system design, including: Agent roles and responsibilities Cross-system context and data flow Read vs write boundaries Human-in-the-loop approval points Ensure the system is cohesive, not a collection of disconnected automations 3. Lead and Coordinate AI Engineers Oversee a team of fractional / contract AI developers and engineers Provide clear requirements, acceptance criteria, and architectural direction Review designs and implementations for: Correctness Safety Maintainability Ensure engineers build toward the product vision, not ad hoc solutions 4. Design AI Capabilities as Products You will oversee the delivery of: An agentic AI layer A primary orchestration agent Optional specialist agents (CRM, tasks, email, data) Cross-system intelligence Normalized context from structured + unstructured data Reasoning across tools and time Action execution Task creation and updates CRM notes and relationship updates Data record updates Drafted communications for approval Triggers and automation Time-based (daily, weekly) Event-based (emails, overdue tasks, stalled deals) 5. Governance, Risk, and Control Define guardrails for AI actions Ensure: Scoped permissions Read vs write separation Explicit approvals for sensitive or destructive actions Plan for failure modes and recovery Required Background (Non-Negotiable) You must have hands-on experience owning AI systems, even if you were not the primary coder. You should be able to confidently reason about: LLM agent architectures and tool calling Claude and/or OpenAI capabilities and tradeoffs MCP or MCP-style multi-tool architectures API-based integrations (CRMs, task tools, email, databases) OAuth, permissions, and access control State, memory, and long-running agent behavior Systems that run unattended in production You must be able to explain how an AI system: Safely reads from one system Decides what matters Writes into another system Avoids causing operational damage Deliverables You Will Own System Architecture Clear diagrams or written explanations Separation of concerns Product roadmap for AI capabilities Phase 1 Read-only intelligence layer “What you missed” and “what’s at risk” reporting Phase 2 Write-back actions with guardrails Human-in-the-loop approvals Documentation How the system works How to extend it How to maintain and govern it How to Apply (Strict) Your proposal must include: A specific example of an AI system you owned that interacted with multiple tools Your high-level product and system architecture for this project Which LLM you would start with and why (from a product perspective) How you think about memory, permissions, and failure states Your availability Anything vague, generic, or purely technical without product ownership will be declined. Engagement Model Initial scoped project Long-term engagement likely for the right person We value judgment, product thinking, and system quality over speed Mandatory Inclusion Describe a system you owned where an AI agent read from one application and wrote actions into another. What went wrong, and how did you correct or govern it?