Greg Isenberg AI Opportunity Scanner

// When should I use the Greg Isenberg AI Opportunity Scanner?

Use this skill when you are evaluating a new business idea, deciding whether to pivot an existing product, or trying to identify which AI-era opportunities are most relevant to your current audience, niche, or skill set. Also useful when stress-testing a SaaS business model against the outcome-based pricing shift.

// What information do I need before running the AI Opportunity Scanner?

• Your idea or business conceptrequired
  What you are building or considering building. Can be vague — the skill will help sharpen it.
• Your current audience or distributionrequired
  Size and type of existing audience, email list, social following, or customer base. If none, state that explicitly.
• Your niche or domain expertise
  The vertical, industry, or subject matter you have an unfair advantage in.
• Your current pricing model (if applicable)
  Whether you are seat-based, usage-based, or outcome-based today.
• Your team size and structure
  How many humans are currently involved in the business.

// What are the core principles behind Greg Isenberg's AI opportunity frameworks?

// How do you apply the Greg Isenberg AI Opportunity Scanner step by step?

1. 1

   Audit your distribution before touching the idea

   Ask: Do I have an email list, social audience, or existing customers? If yes, this is your unfair advantage — the New Timeline only works at full speed when you have someone to sell to on day one. If no, decide upfront whether you will build distribution first or pay for it (ads), knowing paid acquisition cuts into your Micro Monopoly Math margins.

2. 2

   Identify your niche and apply the Vertical AI filter

   List the industry or workflow you know. Score it against Isenberg's Boring Gold Mine Verticals checklist: Does it run on phone calls and faxes? Does it involve 30-year-old processes? Is it sub-niched inside a large category (insurance, legal, logistics, eldercare, accounting, construction, government)? The more boring and the more niche, the better. Avoid high-red-tape verticals (government procurement) to start. Pick a wedge in a subniche, not the whole category.

3. 3

   Classify your idea as Vertical SaaS or Vertical AI

   Ask: Am I selling a tool humans operate (Vertical SaaS, capped TAM, IT budget), or am I replacing work humans are paid to do (Vertical AI, labor P&L, 10x TAM)? If Vertical AI, plan to sell outcomes and results, not seats. If still SaaS, pressure-test whether you can pivot the pricing model to outcome-based before building.

4. 4

   Choose your pricing model using the Seat→Usage→Outcome framework

   Default to outcome-based (pay per result delivered) wherever agents are doing the work. Examples: $1.50 per resolved support ticket, per qualified lead generated, per contract drafted. Only use seat-based if your product is genuinely a human-operated tool. Usage-based is acceptable for infrastructure plays. Ask: What result does the customer care about? Price that.

5. 5

   Run the Micro Monopoly Math on your idea

   Fill in: (a) How large is my reachable niche audience? (b) What can I charge per month given agents are doing the labor? (c) What does 100 customers at that price equal in monthly revenue? (d) What are my agent-run operating costs? If the profit for one person is meaningful, proceed. If not, adjust price point or target a higher-value vertical.

6. 6

   Apply the 1-Hour Company Stack to validate before building

   Before full build: get or validate the idea (use a tool like IdeaBrowser or equivalent), vibe code / agent-code a minimal version with a tool like Claude Code or Codex, attach a payment layer, and attempt to get one paying customer. Treat this as an experiment in your experiment machine, not a six-month commitment. If you can't get one customer in a day or two, reassess.

7. 7

   Design the Ghost Team Org Chart

   Map which business functions can be handled by agents: sales outreach, content creation, customer support, market monitoring, analytics. Assign named agent roles. Identify the minimum human judgment required (creative direction, relationship escalations, ethical review). Structure the business so you check in every few days — the Ambient Business standard. Document what each agent can access using the Agent Permission Stack (files, email, calendar, bank accounts, third-party sharing) and apply the principle of least privilege.

8. 8

   Assess the SaaS Graveyard risk if you are an existing product

   Run your product against the graveyard checklist: Is it a generic CRM? Basic analytics dashboard? Template marketplace? Scheduling tool? Basic chatbot/support? If yes to any, evaluate pivot options: (a) become a vertical workflow tool, (b) add proprietary data or network moat, (c) pivot to agent company that sells outcomes. Generic + no moat + no pivot = graveyard candidate.

9. 9

   Locate yourself in the Scarcity Flip and position accordingly

   Determine which premium tier your offering occupies: Human-made (no AI — use as explicit brand signal, like organic certification), AI-assisted but human-led (premium positioning, human taste + AI speed), fully AI service (commodity, compete on price), or race to zero. For new businesses, aim for human-made or AI-assisted but human-led to avoid the commodity trap.

10. 10

    Evaluate Founder-Agent Fit honestly

    Ask: Can I orchestrate a fleet of agents toward this specific business goal? Do I understand enough about this niche to direct agent output the way a film director directs actors? If yes, this is your unfair advantage. If no, either deepen niche knowledge or pick a niche where you genuinely have that directorial judgment. Founder-agent fit is the new founder-market fit.

11. 11

    Implement Agent Attack Surface hygiene

    Before deploying agents with real access: audit the Agent Permission Stack (what can it access, remember, do, share?). Apply minimum necessary permissions. Schedule quarterly agent cleanses — review and revoke permissions the way you review app access on your phone. Brief yourself on prompt injection and poisoned context window risks, especially if agents browse the web or accept external inputs.

12. 12

    Decide build-in-public strategy and set the flywheel

    If your audience is your customer: build in public. Share what you are building, let the community vote on features, ship updates in days. This creates the co-builder flywheel — users become distribution, trust compounds, and community becomes a moat against forks. If your audience is not your customer: be more selective about what you share to limit competitive signal while still building distribution.

// What are real-world examples of the AI Opportunity Scanner in action?

// What mistakes should I avoid when using the AI Opportunity Scanner?

• Building without distribution first — the New Timeline only compresses to hours if you have someone to sell to; without an audience, finding customers remains the hard part and the whole time advantage evaporates.
• Building in a top-level vertical category (all of insurance, all of legal) instead of picking a wedge in a subniche — the YC-backed players will own the big categories; the opportunity is in the subniches they ignore.
• Staying seat-based when you should be outcome-based — selling software licenses when agents are doing human work means you're capturing IT budget instead of the 10x larger labor P&L.
• Treating vibe-coded output as finished product without iterating — the 1-Hour Company Stack produces a starting point, not a done business; iteration by lunch is part of the process, not optional.
• Ignoring the Agent Attack Surface — giving agents broad permissions (email, bank accounts, code modification) without a permission audit or quarterly agent cleanse creates prompt injection and context poisoning exposure that can cause real damage.
• Waiting for things to settle down — the Asymmetric Window is time-bounded (estimated 12 months before competition catches up, 24 months before the window narrows); the new normal is not settling, and waiting forfeits the asymmetry.
• Building a generic tool instead of a vertical workflow tool — generic CRMs, basic dashboards, template marketplaces, and basic chatbots are in the SaaS Graveyard; survival requires vertical specificity and a pivot toward agent-delivered outcomes.
• Underestimating how small an audience you need — the 100 True Fans revision means you do not need thousands of customers; over-scaling the target market leads to under-niching and losing the micro-monopoly position.
• Confusing founder-market fit with founder-agent fit — knowing the market is no longer enough; you must also be able to orchestrate agents in that market like a film director, or your domain knowledge does not translate into operational leverage.

// What do the key terms in Greg Isenberg's AI frameworks mean?

1-Hour Company Stack

The compressed startup process enabled by agent coding tools: grab a validated idea, vibe code a product, build a landing page, attach Stripe, and reach first customers within hours rather than months.

Old Timeline vs. New Timeline

Old: idea → hire devs → MVP by month 3 → first revenue by month 12. New: idea at 9am → built by 9:15 → product by 9:45 → first customer by 10am → iterate by lunch. Distribution is the only bottleneck remaining.

Ambient Businesses

Businesses that run with zero or very low daily human input, powered by agents that monitor markets, identify opportunities, execute tasks, and handle customer service autonomously. The founder checks in every few days.

Agent Economy

Isenberg's term for the 2025–2030 era, following the App Store era (2009–2015) and API Economy (2015–2024), in which agents discover and hire other agents on the fly, fixed teams dissolve, and agent-to-agent commerce becomes a major share of total economic activity.

Vertical AI

AI products built for specific industries that tap directly into the labor P&L by replacing work humans are paid to do. Contrasted with Vertical SaaS, which only captures IT budget. Vertical AI has a 10x larger TAM because it replaces headcount.

Vertical SaaS

Software built for a specific industry vertical where humans operate the tool. Captures a fraction of IT spend, typically capped at $10M–$100M outcomes. Being disrupted by Vertical AI.

Outcome-Based Pricing

Pricing model where customers pay per result delivered (e.g., per resolved ticket, per contract drafted) rather than per seat or per usage. The direction Isenberg argues all AI-native businesses should default to.

Boring Gold Mine Verticals

Industries with legacy workflows (phone calls, faxes, 30-year-old processes) that are ripe for Vertical AI disruption: insurance, legal, logistics, eldercare, government, accounting, construction. The more boring and the more niche, the better.

SaaS Graveyard

Category of software products likely to be killed by AI: generic CRMs, basic analytics dashboards, template marketplaces, scheduling tools, and basic customer support chatbots — any generic tool that agents can replace natively.

Scarcity Flip

The shift in what is valuable as AI commoditizes execution. Scarce and premium: creative judgment, human-made crafts, physical experiences, original weird thinking, proprietary data. Commoditized: generic content, basic design, data entry, routine analysis.

Premium Stack

Isenberg's hierarchy of perceived value in the AI age: (1) Human-made / no AI — most premium, like organic certification; (2) AI-assisted but human-led — human taste with AI speed; (3) Fully AI service — commodity; (4) Race to zero — undifferentiated.

100 True Fans

Isenberg's revision of Kevin Kelly's 1,000 True Fans for the AI era. Because agents cut costs dramatically, 100 people paying $500–$1,000/month is a real, high-margin business for one person.

Micro Monopoly Math

The unit economics formula: niche audience (500–5,000) + custom app built in 48 hours + 100 customers at $50/month + agent-run operations ≈ $60K profit per person. The building block of the AI-era holding company.

Founder-Agent Fit

The successor to founder-market fit. The ability to orchestrate a fleet of agents toward a specific business goal, operating like a film director — not executing directly, but getting the best performance out of an AI cast.

Ghost Team Org Chart

A company structure with a small number of humans and a set of named AI agents (sales, content, support, etc.) running operations. The architecture of the AI-native holding company.

Agent Attack Surface

The total set of vulnerabilities created when agents are given access to files, email, calendars, bank accounts, and autonomous decision-making. Includes prompt injection, poisoned context windows, malicious MCP servers, agent-to-agent manipulation, and permission escalation.

Agent Permission Stack

The four-axis audit of what an agent can access (files, email, calendars, bank accounts), remember (conversations, personal data, business data), do (send emails, make purchases, modify code, delete data), and share (with other agents or third parties).

Quarterly Agent Cleanse

Regular hygiene practice of reviewing and revoking agent permissions, analogous to reviewing app access permissions on a phone or in SaaS tools.

Agent Injection

The AI-era equivalent of phishing: tricking an AI agent via hidden instructions embedded in context windows or web content, exploiting the agent's autonomy rather than human judgment. Isenberg argues this has higher damage potential than classic phishing.

Asymmetric Window

Isenberg's term for the current moment: inputs required are minimal (API key, prompts, tweet, small niche audience); outputs possible are outsized (24/7 business, 95% margins, compounding distribution). Estimated 12 months before competition catches up, 24 months before the window narrows significantly.

Co-Builder Flywheel

The virtuous cycle created by building in public when your audience is your customer: community votes on features → you ship in days → users feel ownership → trust and distribution compound → moat against forks.

// FREQUENTLY ASKED QUESTIONS