Frequently Asked Questions About Nick Orgo Solo AI Agent Business Playbook
25 answers covering everything from basics to advanced usage.
// Basics
What is the Executive Abstraction Layer in the AI agent business playbook?
The Executive Abstraction Layer is the universal set of pain points shared by decision-makers across all industries: too many emails, too many meetings, too many follow-ups, too many open loops, and context scattered across too many projects. You lead every sales conversation with these universal pains before layering in vertical-specific solutions. This approach works because every executive relates to these problems regardless of their industry.
What is a Hermes agent and how is it different from OpenClaw?
Hermes is Nick's preferred agent harness — it's self-evolving, model-agnostic (you can switch underlying models without changing infrastructure), and has more stable gateways than OpenClaw. OpenClaw is an earlier harness now considered commoditized with less reliable gateway connections. Hermes also supports long-horizon tasks sent via Telegram, making it ideal for async management of client agents from anywhere.
What is Composio and why is it essential for AI agent deployment?
Composio is a single MCP connector that links your AI agents to thousands of apps including Gmail, Slack, Notion, and GitHub. It handles authentication and tool calling, removing the biggest security and setup friction in agent deployment. Every client agent gets Composio installed as part of the core toolchain — it's non-negotiable because it's what allows agents to actually interact with the client's existing business tools.
Do I need to know how to code to follow this playbook?
No — the playbook is designed for non-engineers. You use natural-language instructions to command Claude Code, Codex, or your existing Hermes agent to deploy and configure new agents. The agents-build-agents principle means you describe what you want in plain language and the agent handles the technical setup. You should have basic comfort navigating tools like Trello, Obsidian, and Telegram, and understanding how to connect apps via Composio. But you don't need to write code from scratch.
Why shouldn't I start with healthcare or finance for my AI agent business?
Healthcare and finance have heavy regulatory requirements (HIPAA, SOC 2, financial compliance) that slow early momentum and add legal risk. The playbook recommends starting with lower-friction verticals — marketing agencies, law firms, insurance agencies, manufacturers, wholesalers, real estate agencies — where you can prove the model, build delivery systems, and generate revenue before tackling regulated industries. Once you have cash flow, systems, and experience, you can expand into healthcare or finance with proper compliance infrastructure.
How much does the tech stack cost to run per client?
The exact costs vary by provider, but the core stack — Orgo VM, Hermes agent, Composio, Agent Mail, Obsidian — typically runs well below $5,000/month per client. Most clients need only 1-3 agents. The unlimited offer works because real usage is far less than what the client perceives. Your margin is protected by the gap between the flat monthly fee and actual infrastructure costs. Never share these unit economics with clients — it breaks the digital employee framing.
// How To
How do I onboard a new AI agent client in under 48 hours?
Create an Orgo workspace and spin up a cloud VM. Use your master Hermes agent or Claude Code to deploy a new Hermes agent inside the VM. Install the core toolchain: Composio for app connections, Agent Mail for a named email, and an Obsidian vault pre-loaded with the client's people, projects, and workflows. Focus the first 48 hours on solving one or two highest-pain tasks — typically inbox triage or follow-up automation. Use template Obsidian vaults for your niche to accelerate setup.
How do I create content that generates warm leads for my AI agent business?
Post content demonstrating agent capabilities to your target vertical before you have paying clients. Show real agent outputs solving industry-specific problems — demand letter drafts for law firms, lead follow-up sequences for real estate, content pipeline management for agencies. The goal is that when a prospect books a Calendly call, they already know who you are and what you sell. Use AI to assist with research and editing. This is the primary warm-lead engine and the highest-leverage customer acquisition activity.
How do I set up the Obsidian vault for a client's AI agent?
Create a markdown-based second brain organized around the client's people, projects, workflows, and decisions. Include key contacts with roles, active projects with statuses, SOPs for recurring tasks, and meeting notes. Sync Granola meeting notes into the vault automatically. The more structured and complete the vault, the more the agent feels like it truly knows the business. Build template vaults for your niche so you can pre-populate structure and customize per client in minutes, not hours.
How do I price my AI agent service — is $5,000/month too high?
$5,000/month is the recommended price point because it positions your service as a digital employee, not a tool subscription. A human employee doing the same work costs $4,000-6,000/month plus benefits, management overhead, and training. Frame the price against the cost of hiring, not against ChatGPT subscriptions. The unlimited offer structure (unlimited agents, usage, support) removes buyer friction. Most clients only need 1-3 agents, so your actual costs stay well below the price.
// Troubleshooting
What do I do when a client's AI agent gateway crashes?
This is why you set up watchdogs and observability before anything breaks. Your agent's gateway watchdog should auto-restore crashed connections — Hermes is more stable than OpenClaw here. Configure agents to email you from their Agent Mail address when any cron job, skill, or gateway fails. You resolve the issue before the client notices. If a gateway crash isn't auto-restored, use your master Hermes agent via Telegram or Orgo MCP to remotely access the client's VM and fix it.
What if my AI agent client keeps asking for more features than I can deliver?
Enforce scope control via your Trello Kanban board. Limit delivery to one to two requests per 48-hour window. The board structure — Backlog, To-Do, Doing, Done — makes the queue visible to the client. When clients see the pipeline, they naturally prioritize. Send Loom videos showing completed work as delivery receipts. If a client consistently pushes beyond limits, it's a pricing conversation — not a fulfillment problem. Consider tiering to a higher price point for enterprise-level demands.
What if I can't find a niche that works for my AI agent business?
Apply Diverge Then Converge with a strict time limit. Test 2-3 verticals for 30-60 days maximum. Post niche-specific content, run discovery calls, and track which vertical generates the strongest inbound pull, fastest close rate, and easiest delivery. If none hit, you may be targeting industries with low pain or high skepticism. Pivot to the recommended starting verticals: marketing agencies, law firms, insurance agencies, manufacturers, wholesalers, or real estate agencies. Stop pivoting after 60 days and commit.
My AI agent feels generic to the client — how do I fix this?
The agent feels generic because the Obsidian vault is underdeveloped. Load it with structured context: the client's key people and their roles, active projects with current statuses, decision history, communication preferences, SOPs, and industry-specific terminology. Sync meeting notes from Granola automatically. The vault is what transforms a general-purpose agent into one that feels like it truly knows the business. Review and update the vault weekly as part of your ongoing delivery.
// Comparisons
How does the solo AI agent business compare to a traditional AI automation agency?
A traditional AI automation agency sells custom projects, requires a team of developers, and has variable scope and pricing. The Orgo playbook sells a fixed-price productized service ($5K/month recurring), runs solo with no employees, and scales through agents managing agents. Traditional agencies face project-based revenue volatility and hiring challenges. The solo model produces predictable MRR, maintains high margins because clients rarely need more than 1-3 agents, and uses vertical specificity as a moat instead of team size.
How is this different from just reselling ChatGPT or Claude to businesses?
Reselling ChatGPT or Claude gives clients a generic tool they don't know how to use and will eventually buy directly. The Orgo playbook sells a vertically specific digital employee that knows the client's business, integrates with their tools via Composio, has its own email via Agent Mail, and stores persistent context in an Obsidian vault. You manage everything — the client never touches tokens, models, or infrastructure. This creates a service moat, not a reseller margin that gets compressed to zero.
Should I use Hermes or OpenClaw as my agent harness?
Use Hermes. It's Nick's preferred harness because it's self-evolving, model-agnostic (switch underlying models without changing infrastructure), and has significantly more stable gateways than OpenClaw. OpenClaw is an earlier harness now considered commoditized. Hermes also supports long-horizon tasks via Telegram, which is critical for the solo operator workflow. The gateway stability difference alone justifies the choice — fewer crashes means fewer client-visible disruptions and less firefighting.
Can I run this business with local Mac Minis instead of cloud VMs?
No. The playbook explicitly warns against local Mac Minis. Cloud VMs are remotely manageable, sandboxed, deletable, and scalable. With Mac Minis, you can't fix problems remotely, a hardware failure takes down all agents on that machine (blast radius), and you can't manage them via Orgo MCP from Telegram while walking or sleeping. Cloud VMs per agent per client is the architecture that makes the one-person model work at scale.
// Advanced
How do I manage 8-10 AI agent clients as a solo operator?
Your master Hermes agent, connected to all client Orgo workspaces via Orgo MCP, acts as your operations manager. From Telegram, send long-horizon tasks like 'fix the gateway on Client X's VM' or 'upgrade Client Y's Obsidian vault' while you're walking, sleeping, or in meetings. Monitor all VMs with a single query. Each client's Trello board limits requests to 1-2 per 48-hour window. Template Obsidian vaults and the agents-build-agents principle keep onboarding under 48 hours per new client.
How do I use the Orgo MCP to manage all my client agents remotely?
The Orgo MCP is a connector that lets your personal Hermes agent access, configure, and fix all client agents across all Orgo workspaces. From Telegram, you issue natural-language commands to your master agent, which then executes tasks inside specific client VMs. You can query status across all workspaces ('how many VMs do I have?'), deploy fixes, upgrade configurations, and add new skills — all without logging into individual dashboards. This is the engine behind managing the entire business from your phone.
What is the agents-build-agents workflow in practice?
In practice, you open your terminal or Telegram and give your existing Hermes agent a natural-language instruction like 'deploy a new Hermes agent in Client X's Orgo workspace, install Composio, set up Agent Mail as aria@clientdomain.com, and create an Obsidian vault from the insurance agency template.' Your agent uses Perplexity MCP for current Hermes docs, Context 7 for GitHub docs, and Exa AI for real-time search. It spawns sub-agents if needed. The new client agent is configured without you manually touching the VM.
How do I transition from the Diverge phase to the Converge phase for niche selection?
Set a hard deadline of 30-60 days for the Diverge phase. During this period, run the same Unlimited Offer into 2-3 verticals via content and Calendly. Track three signals: inbound response rate, close rate on discovery calls, and referral frequency. The vertical with the strongest market pull — meaning prospects come to you and refer others — is your convergence target. Then go sub-niche by geography or specialization, re-angle all content, and build template vaults for that vertical. Stop testing other industries entirely.
What vertical-specific agent skills should I build first for law firms?
For law firms, start with inbox triage and demand letter drafting — these solve the highest-pain tasks for partner-level attorneys. Next, add case status follow-up emails, deposition scheduling, and client update automation. Each skill is added one per 48-hour window via Trello. The Obsidian vault should include case types, client names, firm SOPs, and court deadlines. The agent's named email (e.g., aria@firmname.com) sends and receives on behalf of the firm, reinforcing the digital employee positioning.
What vertical-specific agent skills should I build first for real estate agencies?
For real estate agencies, start with lead follow-up sequences — agents that respond to new inquiries within minutes and nurture leads through automated but personalized email sequences. Next, add listing status updates, open house scheduling coordination, and transaction milestone tracking. The Obsidian vault should include current listings, buyer/seller profiles, market data, and agency SOPs. Agent Mail gives the agent a named persona that communicates with leads and clients, making it feel like a dedicated team member.
Is the $5K/month AI agent business model sustainable long-term?
Yes, because the moat is vertical specificity and accumulated client context — not the underlying AI models. As models commoditize, generic AI tool resellers lose value. But an agent that knows a law firm's 200 active cases, or an insurance agency's renewal pipeline, becomes more valuable over time. The Obsidian vault deepens weekly. Switching costs are high because the agent's context is irreplaceable. The unlimited offer structure also creates sticky recurring revenue with low churn when the agent demonstrably improves the client's business outcomes.