← All Docs

2026-03-07-ai-agent-manager-brainstorm

AI Agent Manager — Brainstorm

Date: 2026-03-07 Business: Squamish AI (squamishai.com) Status: Exploring

What We’re Building

A managed AI agent service under Squamish AI. Aaron builds, runs, and improves AI agents for clients. He covers API costs, engineering, and support. Clients get outputs and a human with judgment at the controls.

Core proposition: You get an army of AI agents with a human you trust making the judgment calls.

Target Customer

  • Primary: Small business owners (solo operators, 5-20 person teams)
  • Geography: Start hyperlocal in Squamish, BC — people Aaron already knows
  • Profile: Needs AI but can’t hire for it. Doesn’t know what an “agent” is and shouldn’t have to. Buys outcomes, not technology.
  • Examples: Local trades, real estate agents, coaches, tourism operators, retail

Service Scope

Broad menu — whatever moves the needle for the client:

  • Content & social: Blog posts, social media drafts, newsletters, SEO content
  • Research & reports: Market research, competitor analysis, customer feedback, weekly business intelligence
  • Operations & admin: Invoice processing, scheduling, data entry, email triage, document generation

The agent type adapts to client needs. The overnight agent infrastructure (already built) is the delivery backbone.

Pricing Model

Tiered packages — simple enough for a local business owner to say yes to.

Draft tiers (to be validated with real prospects):

TierPriceWhat You Get
Starter~$500/mo2-3 agents, weekly outputs, email delivery
Growth~$1,500/mo5-8 agents, daily outputs, priority support, Slack channel
Scale~$3,000+/moUnlimited agents, custom workflows, dedicated time, strategy calls

These are hypotheses. Real pricing comes from delivering to first clients and seeing what they value.

Why This Approach

Hybrid: handshake clients + marketing validation

  1. Start with 1-2 handshake clients — local Squamish businesses Aaron already knows. Free or discounted discovery. Build agents, deliver results, learn what works.
  2. Build the service page as a hypothesis — the landing page on squamishai.com tests messaging and pricing with real prospects. It also generates Aaron’s own excitement and clarity about the offering.
  3. Productize what works — after 2-3 months of real delivery, use the data to define real tiers, write case studies, and formalize the packages.

The page does double duty: marketing research AND sales tool.

Key Decisions

  1. Under Squamish AI, not a new brand — agents are a natural extension of the AI consulting offering
  2. Tiered packages over flat retainer — gives clients a clear upgrade path and Aaron a way to manage scope
  3. Start hyperlocal — Squamish businesses, in-person relationships, word of mouth
  4. All agent types — don’t narrow the menu yet; let client needs dictate
  5. Test with 1-2 clients first — validate before building infrastructure
  6. Landing page = hypothesis — use the page to test messaging, not just sell

Competitive Positioning

  • Not a SaaS platform — you’re not competing with Zapier AI or Make. You’re a human who understands their business.
  • Not a dev shop — you don’t hand over code and walk away. You run the agents and deliver results.
  • Not a consultancy — you don’t write reports about what they should do. You do it.
  • The differentiator is judgment — anyone can spin up agents. The value is a human who knows when to override, when to iterate, and when to say “this isn’t worth automating.”

Existing Infrastructure

The overnight agent skill is already a working prototype:

  • Runs Claude Code in tmux with crash recovery
  • Cron-based restart with circuit breaker
  • Question queue pattern (never blocks, logs with best guess)
  • Workspace isolation per run
  • Dedicated second computer for security

This infrastructure directly supports client agent delivery.

Open Questions

  1. How do you deliver outputs? Email? Slack? A simple dashboard? What format do small biz owners actually want?
  2. How do you scope “unlimited” in the Scale tier? Need guardrails to prevent one client consuming all your time.
  3. What’s the onboarding flow? Discovery call → what happens next? How long before first agent is running?
  4. How do you handle client data/privacy? Especially for operations agents that touch invoices, emails, etc.
  5. What’s your cost structure? API costs per client, time per client, break-even point per tier.
  6. Do you need a contract/SLA? Or is a handshake + monthly billing enough to start?
  7. How do you show ROI? What metrics do you track and report to justify the monthly cost?

Next Steps

  1. Identify 1-2 Squamish businesses to approach for pilot
  2. Build a service page on squamishai.com with draft tiers and messaging
  3. Run discovery sessions with pilot clients
  4. Build and deploy first agents using overnight infrastructure
  5. Document everything — time spent, API costs, client feedback, outcomes
  6. After 2-3 months: refine tiers, write case studies, formalize offering