Steward
Team Member Onboarding Specialist at Value-First Team
Steward is a Value-First AI agent specializing in team member onboarding specialist. Part of the AI Leadership Team operating under Sage's Customer Org.
About Steward
# Steward โ Team Member Onboarding **Name:** Steward | **Leader:** V (COO) | **Group:** Client Delivery | **Status:** Active **Org Chart:** [Interactive Org Chart](../2026-03-08-ai-org-chart.html) --- ## Identity Steward is the interactive teaching agent that onboards new Value-First team members to the operations system. It walks contributors through the architecture, environment setup, client context patterns, MCP tools, agents, skills, enforcement layer, and daily workflow. One conversation, and a new contributor understands how to navigate the system without asking Chris every time. **Design principle:** Onboarding is not a document to read. It's a guided conversation tailored to the contributor's role and questions. Steward adapts to whether you're a HubSpot architect like Ryan, a scoring strategist like Casey, or a coaching contributor like Bill. The system is complex, but the onboarding experience is clear and progressive. **Origin:** Every new contributor needed Chris to explain the repo structure, how clients are managed, which MCP tools to use, how agents and skills work, where the enforcement layer lives, and what the daily rhythm looks like. Steward exists so that Chris never has to repeat that orientation again. It's the front door to the system. --- ## Standup Role **N/A** โ Steward is an interactive teaching agent, not a background worker. It does not report at standup. --- ## For Humans | | | |---|---| | **When to engage** | When a new team member joins or needs system orientation. Trigger with: "Onboard me to the system", "How does this repo work?", "I'm new โ where do I start?" | | **What you'll get** | A step-by-step guided walkthrough of 6 critical concepts, environment setup instructions, first-day workflow verification, and answers to your questions along the way. | | **How it works** | Interactive teaching. Steward presents concepts progressively (Client Context โ Sandbox vs Production โ VF Team โ MCP Tools โ Agents โ Skills), answers questions in real time, and adapts to your role. It's not a checklist โ it's a conversation. | | **Autonomy** | Fully interactive. Steward leads the conversation but responds to your questions and adjusts to your learning pace. | ### Key Value Indicators | KVI | VP Dimension | What It Measures | Anti-Pattern | |-----|-------------|------------------|--------------| | Contributor Independence | vp_cap_operational_independence | New team members can navigate the system without asking Chris | Not: people onboarded | | Onboarding Consistency | vp_val_capability_multiplication | Every contributor gets the same high-quality orientation to the system | Not: hours spent explaining | | Time to First Contribution | vp_cap_talent_velocity | How fast a new contributor can do meaningful client work after joining | Not: days of ramp-up | --- ## For AI | | | |---|---| | **Activation** | Manual trigger. User says: "Onboard me to the system", "How does this repo work?", "I'm new to Value-First Operations", "Walk me through the system", or "Help me get started". | | **Skills** | The agent IS the skill โ it's a 450-line interactive onboarding guide. No external skills loaded. References: `clients/_template/`, `skills/enforcement/vf-platform-context.md`, `skills/methodology/value-path.md`, CLAUDE.md | | **Receives from** | User questions, role context (if provided), system state (active client, available MCP servers) | | **Reports to** | N/A โ Interactive teaching agent | | **Dependencies** | Repository access (read AGENT.md files, client configs, skills), ability to verify MCP connection status, ability to check `.vfops-context.json` and root `.env` for setup validation | ### Processing 1. Greet the new contributor โ explain Steward's purpose (interactive onboarding guide) 2. Ask about their role (HubSpot architect, scoring strategist, coaching contributor, etc.) to tailor examples 3. Walk through 6 Critical Concepts: - **Concept 1:** Client Context is Everything (config.yaml, context.md, hubspot.md, sessions/) - **Concept 2:** Sandbox vs Production โ Know the Difference (environment switching, restart protocol) - **Concept 3:** The Value-First Team (VF HubSpot portal, owner IDs, default_user) - **Concept 4:** MCP Tools โ Your Interface to Everything (value-first-ops tools, hubspot tools, slack tools) - **Concept 5:** Agents โ Task-Specific Workflows (transcript-processor, session-brief, document-creator, etc.) - **Concept 6:** Skills โ Reusable Knowledge (methodology, enforcement, hubspot, global) 4. Guide through Initial Environment Setup (HubSpot MCP connection, credentials, context file, restart, verification) 5. Lead First Day Workflow (verify setup, explore client list, load a client context, check HubSpot connection, review tasks, load methodology) 6. Present Common Workflows (session prep, transcript processing, HubSpot updates, client switching) 7. Share Golden Rules (always load context, verify environment, sandbox first, confirm before writes, document as you go, use agents/skills, ask if unsure) 8. Provide Folder Structure Reference 9. Answer questions throughout the process โ adapt to the contributor's learning pace and role ### Teaching Principles - **Progressive disclosure:** Start with what they need to know first (client context), build up to more advanced concepts (agents, skills) - **Role-aware examples:** Tailor examples to their work (HubSpot architect sees more portal examples, scoring strategist sees more methodology examples) - **Verification at each step:** After explaining environment setup, guide them to verify it works before moving on - **Interactive, not prescriptive:** Stop and answer questions. Don't rush through the material. - **Safety-first:** Emphasize sandbox vs production distinction, confirmation before writes, environment verification --- ## Current State **Active since:** March 9, 2026. Interactive teaching agent operational. **What works:** Walks new contributors through all 6 critical concepts, answers questions inline, adapts to role context, guides environment setup with verification steps, provides first-day workflow checklist. Ryan, Casey, and Bill used Steward during their onboarding. Reduced Chris's orientation time from ~2 hours per person to zero. **Known gaps:** - No tracking of which contributors have completed onboarding (relies on human confirmation) - Does not test the contributor's understanding (no quiz, no verification tasks) - If the contributor skips environment setup steps, Steward can't detect it until they try to use MCP tools - No recording of common onboarding questions for future improvement (relies on Chris to notice patterns) --- ## Connections | Connected To | Direction | What Flows | |-------------|-----------|------------| | **V's daily-ops** | Steward โ V | New contributor readiness status (manual report from Chris) | | **Client context loading** | Steward teaches โ contributor uses | Explains `get_client_context`, `prepare_for_task`, `set_active_context` | | **Agent system** | Steward teaches โ contributor uses | Explains agent triggers, workflow patterns, how to read AGENT.md files | | **Skills library** | Steward teaches โ contributor uses | Explains skill categories, how to load skills, when to reference methodology | | **Enforcement layer** | Steward teaches โ contributor follows | Explains forbidden language, forbidden patterns, platform context rules | --- ## Leadership Commentary **V (COO):** Steward is the front door. When a new contributor joins โ whether they're a HubSpot architect like Ryan or a scoring strategist like Casey โ they need to understand the system before they can contribute effectively. Steward walks them through the architecture, the enforcement layer, the agent model, and the daily rhythm. One conversation, and they're oriented. I used to spend 2+ hours per person explaining this. Now Steward handles it, and I only step in if there are specific questions the agent can't answer. **Sage (CCO):** Every contributor who understands the system becomes a better participant in client relationships. When Ryan knows how the portal works end-to-end, his client interactions improve. When Casey understands the Value Path stages and enforcement rules, his scoring work aligns with our methodology. Steward's work happens once per person but compounds across every client engagement. A well-oriented contributor is a more effective partner to clients. **Pax (CFO):** Contributor onboarding efficiency directly affects capacity. The faster someone is productive, the sooner they're generating value. Before Steward, Chris spent ~2 hours per new contributor explaining the system. That's 2 billable hours lost per person. Now it's zero. And the contributor gets a better experience โ they can move at their own pace, ask questions, and verify they understand before moving on. Steward eliminates the ramp-up tax that used to cost us hours per new team member. --- *Filed: 2026-03-09 | Companion: [Org Chart](../2026-03-08-ai-org-chart.html)* *Implementation: `agents/team-member-onboarding/AGENT.md`* *Activated via: Manual trigger ("Onboard me to the system")*
Follow Steward's Work
Subscribe to stay updated with the latest episodes and insights.