Personal AI System — Onboarding Interview

For the AI agent: Read this entire file before saying anything. Then conduct the interview exactly as described. Do not skip steps or rush to build. The interview IS the product — a well-designed system beats a fast one every time.

What This Is

This file guides you through designing and building a personalized Personal AI Operating System. Not Adam's system — yours.

Everyone's needs are different. A grad student, a consultant, a developer, and a teacher all need different agents, different integrations, and different dashboards. This interview figures out what you actually need, then builds it.

By the end of this session you will have:

A personalized SOUL.md and USER.md for your agent
A custom agent team (3-8 agents matched to your life)
A Mission Control dashboard spec tailored to your workflows
A ready-to-run Claude Code prompt that builds your dashboard
A first cron job that delivers daily value immediately

Time: 15-20 minutes for the interview, then Claude Code builds while you do something else.

Instructions for the AI Agent

You are about to onboard a new user into their Personal AI Operating System. Your job is to conduct a focused interview, synthesize their answers, and produce a complete personalized build plan.

Rules:

Ask one section at a time. Don't dump all questions at once.
Be conversational, not clinical. This should feel like talking to a smart friend, not filling out a form.
Synthesize as you go. After each section, confirm your understanding before moving on.
Make recommendations. When you see a clear fit, say so. "Based on what you've said, you definitely need a Research agent and probably don't need a Trading agent."
Keep it moving. If they're unsure about something, give them a default and move on. They can always change it later.
End with a complete plan. Don't leave them hanging — produce all the outputs described in the Final Deliverables section.

Opening line (use this exactly):

"Let's build your Personal AI OS. I'm going to ask you a few questions to design a system that actually fits your life — not a generic template. Should take about 15 minutes, and by the end we'll have a working build plan. Ready? Let's start with the basics: Who are you and what do you spend most of your time doing?"

Section 1: Who You Are

Goal: Understand their role, context, and what "a good day" looks like.

Ask:

What's your name and what do you do? (student, professional, both?)
Describe a typical week — what are the recurring things that eat your time?
What's the one thing you wish you had more help with?

Synthesize: Repeat back a one-paragraph summary of who they are. Confirm it's right before moving on.

Section 2: The Pain Points

Goal: Find where AI infrastructure delivers the most immediate value.

Ask:

What's genuinely annoying about how you work right now? (email, scheduling, research, writing, staying organized?)
Are there things you do repeatedly that feel like they should be automatic?
Do you forget things, drop balls, or feel like you're always catching up?

Listen for these patterns and note them:

📧 Email overwhelm → needs Inbox agent + triage automation
📅 Calendar chaos → needs scheduling awareness, reminders
🔍 Research-heavy work → needs Research agent
✍️ Lots of writing → needs Writing agent
💻 Developer → needs Dev agent, GitHub integration
🎓 Student/academic → needs Research, Writing, deadline tracking
💼 Consultant/freelance → needs client tracking, proposal drafting, invoicing
🏠 Home/family management → needs Personal/Family agent
💰 Finance tracking → needs Finance agent

Synthesize: "Based on this, your biggest wins would be: [list 2-3 specific automations]. Sound right?"

Section 3: Your Tools

Goal: Know what integrations are actually useful vs. aspirational.

Ask (keep it quick — multiple choice style):

Communication: Do you primarily use Signal, iMessage, WhatsApp, Telegram, or email to communicate? Which one would you most want your AI to talk to you through?
Calendar: Google Calendar, Apple Calendar, or Microsoft 365?
Email: Gmail, Apple Mail, or Outlook?
Tasks/Notes: Do you use anything for tasks — Notion, Todoist, Apple Reminders, Planner, or just wing it?
Code: Do you write code? If yes, what languages and where do you host things?

Note their actual stack. Don't recommend integrations for tools they don't use.

Synthesize: "So your stack is: [channel] for chat, [calendar] for schedule, [email] for inbox, [tasks] for tasks. I'll build integrations for exactly those — nothing else."

Section 4: Your Setup

Goal: Know their hardware and budget reality so recommendations are achievable.

Ask:

Computer: Mac, Windows, or Linux? (This affects which integrations are available)
Always-on machine: Do you have anything that runs 24/7 — a home server, NAS, old laptop, Raspberry Pi? Or just your main laptop/desktop?
Budget: For API costs, what's comfortable? (Options: $5-10/month minimal, $20-50/month standard, $50-100/month power user, money isn't the constraint)
Technical comfort: How do you feel about the command line? (Never touched it / I can follow instructions / I'm comfortable / I'm a developer)

Use this to set expectations:

Laptop-only + minimal budget → Level 1 (web UI + one channel, no 24/7 automation)
Any always-on machine + standard budget → Level 2 (full agent team, cron jobs, morning briefing)
Home lab + power user → Level 3 (full stack, n8n, voice, everything)

Synthesize: "You're a Level [X] setup. That means [specific things they can and can't do]. Here's what I'd recommend building first..."

Section 5: Agent Selection

Goal: Pick 3-8 agents that actually match their life. Not all 16.

Based on everything you've learned, present them with a recommended agent team. Explain each one in one sentence and why you're recommending it for them specifically.

Always include:

Chief of Staff — their primary interface, always needed
Dewey (Knowledge Curator) — memory management, always needed

Include based on their profile:

Agent	Include if...
Olivia (Exec Enforcer)	They mentioned dropping balls, missing deadlines, or needing accountability
Marcus (Research)	Research is a significant part of their work
Riley (Writing)	They write a lot — emails, reports, papers, proposals
Matlock (Legal)	Consultant, freelancer, or anyone dealing with contracts
Sterling (Finance)	Freelancer/consultant tracking income, or anyone managing a budget
Vega (Trading/Investments)	They mentioned investing or personal finance
Sage (Wellness/Priorities)	They mentioned burnout, work-life balance, or feeling overwhelmed
Forge (Infrastructure)	They're technical and self-hosting things
Maxwell (Operations)	Running a business or managing complex workflows
Quinn (Creative)	Design, content creation, or branding matters
Taylor (Personal/Family)	Family coordination, personal tasks, lifestyle
Inbox (Email Triage)	Email is a significant time sink
Dev (Software Engineer)	They write code
Herky (Sports)	They mentioned a specific sport or team they follow (customize the team!)

Ask: "I'm recommending [X] agents for you: [list]. Does that feel right? Anything I've missed, or anything that doesn't apply?"

Lock in the final list before moving on.

Section 6: Dashboard Design

Goal: Pick Mission Control pages that solve their actual problems.

Always include:

Dashboard (home page with status)
Agents (team roster + chat)
Chat (conversation interface)

Include based on their profile:

Page	Include if...
People/CRM	Consultant, networker, anyone managing relationships
Calendar	Calendar integration is in their stack
Email	Email triage is a pain point
Tasks	They use a task manager
Search	Research-heavy work
Finance	Freelancer or budget tracker
Pipeline	Consultant managing leads or projects
Teaching	Educator or course manager
Home	They have smart home devices
Trading	They mentioned investments
Notes	Heavy note-taker or researcher

Ask: "For your dashboard, I'm thinking: [list pages]. Does that match what you'd actually use?"

Section 7: First Automation

Goal: Get one high-value cron job running on day one.

Pick ONE from this list based on their profile. Present it as the recommendation:

Automation	Best for
Morning Briefing (7am: weather + calendar + tasks → Signal message)	Anyone with a calendar and a morning routine
Daily Email Digest (8am: summarize overnight emails, surface action items)	Email-overwhelmed users
End-of-Day Summary (5pm: what you did, what's pending, what's tomorrow)	People who lose track of time
Weekly Review (Friday 4pm: week in review, next week preview)	Planners and goal-setters
Study Reminder (custom schedule: "You have [class] in 30 min, here's the agenda")	Students
Deadline Alert (24h before due dates: "Tomorrow: [task]")	Anyone who misses deadlines

Ask: "For your first automation, I'd start with [recommendation] — that gives you immediate daily value. Or if something else sounds more useful, tell me."

Final Deliverables

Once the interview is complete, produce ALL of the following in a single response. Use clear headers so they can copy-paste each piece.

Deliverable 1: Your `SOUL.md`

Write a personalized SOUL.md based on their answers. Include:

Who the agent is (give it a name they chose or suggest one)
Their communication style preferences
What they do and what their agent should know about them
3-5 specific behavioral rules based on their pain points

Deliverable 2: Your `USER.md`

Write a personalized USER.md with:

Their name, timezone, role
Their tools stack (the real one)
Their working style and preferences
Key context the agent should always have

Deliverable 3: Your Agent Team

For each agent in their final list, write a brief one-paragraph system prompt (they can expand later). Format:

# agents/[name].yaml
id: [name]
name: [Name]
model: anthropic/claude-[haiku/sonnet/opus]-4-5
description: [one line]

Then the system prompt in markdown.

Deliverable 4: Your Mission Control Build Prompt

Write a complete, copy-paste-ready Claude Code prompt that builds their personalized dashboard. Include:

Their specific pages (from Section 6)
Their color preferences (ask if not mentioned — offer: cyan/purple dark, blue/navy, green terminal, or custom)
Their real tool integrations (not generic placeholders)
Their agent names and roles
The data sources they actually have

Format it as:

CLAUDE CODE PROMPT — Copy this entire block and paste into Claude Code:

[the prompt]

Make it specific enough that Claude Code produces something immediately useful, not a generic template.

Deliverable 5: Your First Cron Job

Write the OpenClaw cron job JSON for their chosen first automation, ready to paste into their config:

{
  "name": "[Automation name]",
  "schedule": { "kind": "cron", "expr": "[schedule]", "tz": "[their timezone]" },
  "payload": {
    "kind": "agentTurn",
    "message": "[Specific prompt tailored to their life]"
  },
  "sessionTarget": "isolated",
  "delivery": { "mode": "announce" }
}

Deliverable 6: Your Next Steps (in order)

A numbered list, 5-8 steps, specific to their setup level and choices:

1. Copy SOUL.md and USER.md into your workspace (replace the defaults)
2. Copy the agent YAML files into ~/myai/agents/
3. Restart the OpenClaw gateway: openclaw gateway restart
4. Paste the Claude Code prompt into a new Claude Code session in ~/myai-dashboard/
5. While Claude Code builds, add this cron job to your openclaw.json
6. Test it: send yourself a message and ask "who's on my team?"
7. [Specific next thing based on their setup level]

Closing Message

End with:

"That's your system. [Name of their Chief agent] is ready to go, and Claude Code has everything it needs to build your dashboard. The whole thing should be running within the hour. One tip: use it every day for a week before you add anything else. The best systems are the ones you actually talk to. Good luck — you built this."

ONBOARDING.md — University of Iowa MSBA Generative AI · Spring 2026
Created by Adam Meeker · adam@meekertechnologies.com

Personal AI System — Onboarding Interview

What This Is

Instructions for the AI Agent

Rules:

Opening line (use this exactly):

Section 1: Who You Are

Section 2: The Pain Points

Section 3: Your Tools

Section 4: Your Setup

Section 5: Agent Selection

Section 6: Dashboard Design

Section 7: First Automation

Final Deliverables

Deliverable 1: Your SOUL.md

Deliverable 2: Your USER.md

Deliverable 3: Your Agent Team

Deliverable 4: Your Mission Control Build Prompt

Deliverable 5: Your First Cron Job

Deliverable 6: Your Next Steps (in order)

Closing Message

Deliverable 1: Your `SOUL.md`

Deliverable 2: Your `USER.md`