Frequently Asked Questions About Greg Isenberg Tiny AI Agent Business Builder
21 answers covering everything from basics to advanced usage.
// Basics
What does 'boring' actually mean in a tiny AI agent business?
'Boring' means the business is unglamorous, specific, and immediately cash-flowable — like flipping expired domains or brokering used restaurant equipment — rather than a flashy billion-dollar startup. Boring is a feature, not a bug: these niches have less competition, clear buyers, and predictable spreads. The methodology optimizes for speed to first dollar, so boring wins over exciting-but-unproven.
What is the Feed → Asset → Trigger → Buyer → Monetization chain?
It's the five-node chain every viable tiny AI agent business must complete: a constantly-updating public data feed surfaces a mispriced or neglected asset; a trigger event (price drop, shutdown, rank decline, hiring signal) flags it as actionable now; an obvious buyer with money exists; and a clear liquidity point (flip, broker, retainer, relaunch) closes the loop. If any node is missing, the idea isn't ready.
What does 'agents are the new SaaS' mean?
It's a monetization philosophy shift from selling software seats to selling outcomes. Instead of a per-seat model where users interpret a dashboard themselves, you sell an AI agent delivering a specific outcome on a recurring basis — like a $9.99/month competitive intelligence brief. The customer pays for the daily intelligence or deal card, not access to a tool.
What is the 'one-liner first' rule?
Before touching any tool, you compress your idea into a single sentence naming the data feed, asset type, scoring criteria, delivery channel, and end buyer. This one-liner becomes your first message to the AI agent and forces clarity before building. Skipping it produces a generic tool instead of a targeted business asset — it's the mandatory first step of the methodology.
// How To
How do I write a good one-liner for my agent?
Use this format: 'Monitor [feed source] for [asset type] that meet [scoring criteria], flag when [trigger event], deliver ranked list to [channel] so I can [monetization method] to [buyer type].' Example: 'Monitor expired domain drops for domains with DR 20+, clean backlinks, no adult history, under $200 — deliver 10 picks daily to Slack — flip to newsletter operators.'
How do I set up a Slack webhook for my agent's deliveries?
Create a dedicated channel for the business idea, then go to api.slack.com/apps, add a new webhook, and paste the URL into your agent config. Never mix multiple agent outputs into one channel — it creates noise that kills the daily action habit. If you're unfamiliar, ask the agent to walk you through it; it produces step-by-step instructions.
How do I fix bugs in my agent's output?
Treat every correction as a conversation, not a code rewrite. Say something like: 'The links in the cold emails have HTML entities bleeding through — fix it so they render clean.' Or: 'Increase the budget ceiling to $2,500.' The agent updates its own logic. You never need to rewrite the original prompt — this conversational loop is the core interaction pattern.
How do I decide which monetization method to pick?
Match the method to your risk tolerance and capital. Choose Flip if you can front acquisition cost and want the full spread. Choose Broker if you want zero inventory risk and are comfortable charging a 15–30% fee. Choose Retainer if you'd rather sell recurring intelligence as a subscription. Choose Relaunch if you can improve a dead asset. Lock in exactly one per agent.
Should I over-specify my agent's configuration upfront?
No. Don't over-engineer the prompt upfront — write the one-liner, answer the agent's clarifying questions concisely, and if you have no preference say 'you pick the best ones' so the agent uses reasonable defaults. Then iterate by talking to it in plain language. Treat it like an AI employee you refine over time, not a spec you must perfect before launch.
// Troubleshooting
My agent went quiet — is it broken?
Not necessarily. When an agent goes quiet, it may be autonomously reconfiguring or rebuilding itself rather than failing. Check the app status before troubleshooting. Also confirm 'prevent sleep' is enabled if you want it running overnight. Only enable the heartbeat — which checks for pending events every 30 seconds — once the business generates revenue, since it continuously consumes tokens.
Why is my agent producing a generic tool instead of a business?
You likely skipped the one-liner step and jumped straight into building. Without a compressed brief naming the feed, asset, scoring criteria, channel, and buyer, the agent has no target and defaults to a generic scraper. Go back, write the one-liner, and paste it as your first message. The specificity of the brief determines the specificity of the output.
Why aren't my deal cards leading to any sales?
Usually because there's no obvious, reachable buyer, or the spread is too small to matter. Reapply the three quick-screening questions: Is there urgency? Is there enough spread between acquisition cost and resale value? Who pays first? If you can't name a specific buyer before committing capital, the Feed → Asset → Trigger chain is worthless. Also check for output bugs like broken links.
My cold emails have HTML entities in them — how do I stop this?
This is a common data bug caught during manual review — exactly why you should never automate outreach before validating a full output cycle. Tell the agent in plain language: 'The cold email drafts show HTML entities like & instead of clean characters — fix the encoding so emails render properly.' The agent corrects its own logic. Always inspect the first 10–15 deal cards before scaling.
// Comparisons
How does this differ from just using ChatGPT to brainstorm business ideas?
Brainstorming with ChatGPT gives you ideas; the Tiny AI Agent Business Builder gives you a running, revenue-producing asset. It goes beyond ideation to a five-node validation chain, a live data feed, structured deal cards, a delivery channel, and a locked monetization method. The deliverable is a recurring daily brief you can act on — not a list of concepts you still have to execute manually.
How does the broker model compare to the flip model?
The flip model requires you to front acquisition capital and hold the asset until resale, capturing the full spread but carrying inventory risk. The broker model — used in the Local Liquidation pattern — connects seller and buyer for a 15–30% fee with zero inventory and zero capital at risk. Flip has higher upside per deal; broker has lower risk and faster velocity.
How is this different from traditional dropshipping or arbitrage?
Traditional arbitrage requires you to manually hunt listings and calculate spreads; here an AI agent monitors public feeds continuously and surfaces ranked deal cards with spread math already done. It's also broader than physical goods — it covers domains, software, leads, and intelligence briefs. The recurring daily brief and outcome-based monetization ('agents are the new SaaS') set it apart from one-off arbitrage.
// Advanced
When should I enable the heartbeat feature?
Only after the business generates revenue. The heartbeat checks for pending events every 30 seconds to keep the agent continuously active, but it consumes tokens continuously and becomes a cost sink without cashflow to justify it. Before revenue, run scheduled daily briefs instead. Enable 'prevent sleep' for overnight runs, but reserve the heartbeat for validated, paying businesses.
Can I run multiple tiny AI agent businesses at once?
Yes, but build separate agents and separate delivery channels for each idea. Never mix multiple business ideas into a single Slack channel — it creates noise that kills your ability to act quickly on individual deal cards and destroys the daily habit. Also build a separate agent per monetization method; mixing flip, broker, and retainer logic in one agent creates confusion.
What are the three quick-screening questions before scaling an agent?
Before enabling autonomous runs, verify: (1) Is there urgency? — the asset gets claimed by others if you don't act fast. (2) Is there spread? — the gap between acquisition cost and resale or broker value must be large enough to be worth acting on. (3) Who pays first? — you can identify a specific, reachable buyer before committing capital or time. All three must be yes.
How do I turn a tiny AI agent business into a recurring subscription product?
Pick the Retainer liquidity point and productize the daily brief itself. Configure the agent to deliver a one-page intelligence brief — like a competitive intelligence summary for a specific vertical — then vibe-code a landing page based on a reference style and sell access for something like $9.99/month. This is the 'agents are the new SaaS' outcome-based model: customers pay for the brief, not a dashboard.
Why does the agent need to run on a schedule to count as a business?
Because the value is in the recurring daily brief, not a single scrape. A one-time output is a research tool; a scheduled agent delivering fresh deal cards each morning is a business. If the agent doesn't run on a fixed schedule, you can't build the daily action habit or reliably generate cashflow. Set it to deliver a ranked brief at a fixed daily time.