Why this is a special window of opportunity

In a long post on X, Nav Toor points out that the AI automation services market currently has a supply-demand gap that is widening rapidly. The structure of this gap is simple: on one side, millions of small business owners have realized AI can save them time, but after trying tools like Claude and receiving answers that are too generic, they give up; on the other side, a small group of builders who master sub-agent architectures, MCP server configuration, and skill file design are landing projects at $3,000 to $15,000 each, with clients queuing up.

The essence of this window is a skill premium. When a technology has spread to the point that “everyone knows it exists” but has not yet reached the point that “everyone can independently use it to create value,” those who master the intermediate-layer skills can earn outsized returns. Toor believes this premium won’t last long, because supply is growing quickly. The article’s core argument: the next 90 days are the critical period for building a personal brand and vertical expertise, and after 18 months the market will become crowded.

Defining and bounding AI automation services

The AI automation services Toor describes are not simply “using AI to write copy” or “having AI reply to emails.” Their core feature is: converting 5 to 15 high-frequency, predictable, fixed-process tasks inside a business into automated systems that run without human intervention, through sub-agent architectures and external tool integrations.

These tasks typically include: research report generation, client report compilation, content production, customer email handling, document preparation, inventory updates, meeting notes organization, lead information enrichment, and so on. What they share in common: highly repetitive, clearly defined processes, but each business has subtle, personalized needs.

The difference from traditional SaaS tools is that AI automation services are customized deliverables: each client’s context files, sub-agent configurations, and MCP toolchain are tailored to their specific business scenario. This makes it closer to a consulting delivery model, but with marginal costs far below traditional consulting.

The six stages of the 90-day action roadmap

Toor breaks the whole process into six stages, each with clear time boundaries and deliverables.

Stage 1: Build your toolbox (Days 1-14)

The goal of this stage is to truly master four core skills rather than skim the surface.

Context file design is the starting point. The output quality of every automation system depends on how deeply the system understands the client’s business scenario, audience characteristics, brand tone, and quality standards. Toor suggests practicing by building context files for hypothetical businesses in three different industries: a real estate agency, a marketing consultancy, and a DTC e-commerce brand.

Sub-agent architecture is the engine. Unlike a single general-purpose agent, Toor emphasizes designing dedicated sub-agents for each specific task: code reviewer, weekly report generator, cold email writer, meeting notes organizer, bug reproducer. Each sub-agent needs to be tested and iterated with 10 different inputs before it reaches a deliverable reliability standard.

MCP server configuration is the watershed that upgrades a chatbot into a system. Toor lists five tool integrations you must master: Tavily (web search), Google Drive (file storage), Gmail (email), Slack (communication), and at least one CRM. The practice goal is to build an end-to-end workflow: search the web for information, save the summary to Drive, email it to yourself, post a notification in Slack, and log it to the CRM at the same time.

Scheduled automation design handles the highest-value scenario: unattended operation. Daily briefings, weekly client reports, monthly invoice processing — all of these require understanding scheduling mechanisms, timezone edge cases, and graceful degradation strategies when data is missing.

Stage 2: Pick a vertical (Days 14-21)

Toor is emphatic about “don’t try to serve every business.” He offers four selection criteria: you understand the industry’s workflows well enough, clients can afford to pay $3,000 to $5,000, there are at least 5 repeatable tasks that can be automated, and clients are not technical enough to do it themselves.

The article lists eight verticals that are already generating real revenue: real estate agencies, marketing agencies, e-commerce brands, professional services firms, content creators, financial advisors, medical clinics, and law firms. Each has 5 to 10 high-value automation scenarios waiting to be developed. Toor’s warning is blunt: “Choosing two verticals means zero referrals.”

Stage 3: Build your first case study (Days 21-30)

This is the most counterintuitive step in the entire roadmap: do automation for free for your first client. Toor believes the value of this “free” deal far exceeds money: you get a production system actually running in the real world, a case study with concrete numbers, and a testimonial from a real client.

Ways to find that first free client include: posting in vertical community groups, reaching out through your personal network, or sending cold outreach directly to 10 businesses. Toor’s conversion expectation is 1 yes per 10 messages. The key is to measure everything: how long the task took manually, how long after automation, what the quality difference is, how many hours are saved per week.

Stage 4: Package and price (Days 30-35)

Toor is explicitly against billing by the hour. His pricing logic is based on value delivery: if an automation system saves 10 hours per week, at an internal cost of $75/hour that’s $3,000 of value per month. Charging a one-time build fee of $3,000 to $5,000 is an obvious ROI for the client.

For complex multi-agent systems, custom MCP servers, or cross-platform integrations, the price range rises to $5,000 to $15,000. The delivery package includes: a 2-hour discovery session, 3 to 5 custom sub-agents, MCP server configuration, a custom context file, a two-week testing and optimization period, maintenance documentation, and 30 days of post-launch support.

Stage 5: Find paying clients (Days 35-60)

Toor lists three channels that work. Content marketing is about showing before/after: “I built an automation system for a marketing agency that cut client reports from 3 hours to 3 minutes. Here’s exactly how.” One post a week, and after four months you’ll have organic traffic. Direct outreach targets 3 to 5 messages a day; 100 messages produce 10 calls, converting into 2 to 3 closed deals. Referrals close at rates of 50% to 70% — one good referral is worth 10 cold emails.

Stage 6: Scale to $10K/month (Days 60-90)

Once you have 3 to 5 paying clients, patterns naturally emerge. Every marketing agency needs the same 3 automations; every e-commerce brand has the same 5 pain points. This is the moat: productize the common patterns, replacing from-scratch builds with templates.

Toor also suggests adding a monthly maintenance service at $500 to $1,000/month, covering automation optimization, adding new features, and ensuring systems keep running stably after model updates. The math is simple: 5 maintenance clients × $750/month = $3,750 in recurring revenue; 2 new builds per month × $5,000 = $10,000 in project revenue; total $13,750/month, or $165,000/year.

Tool stack and cost structure

Toor emphasizes that the tool costs for this business are extremely low. The entire stack needs only five tools: Claude Code or Cowork (sub-agent runtime environment), Notion or Obsidian (context files and case study management), Cal.com (scheduling discovery calls), Stripe (invoicing and payments), and Loom (async client updates). Total monthly cost is under $200.

This stands in sharp contrast to the cost structure of traditional software development or consulting businesses. No server fees, no team salaries, no office rent. The only cost is time — and after the first two stages, time investment is massively compressed through templates.

The three most common failure modes

Toor notes that 90% of those who try will fall into three traps.

Tool perfectionism: spending 8 weeks learning instead of doing. Toor’s position is clear: two weeks of practice is enough; your first paying client will teach you more than two months of theory.

Underpricing: charging $500 because you’re “still learning.” Toor believes this actually kills deals, because clients trust the professionalism signaled by a high price. Charge a minimum of $3,000.

No vertical focus: trying to serve every business and ultimately serving no one well. Pick one industry, go deep, then expand.

Why this window is closing

Toor’s “uncomfortable truth” lies in the market dynamics themselves. The current supply of builders who can deliver AI automation systems is tiny — almost negligible relative to demand. That explains why clients accept small portfolios, why competence alone is enough, and why the premium can hold.

But 18 months from now things will be different. More people will master these skills, more templates will be open-sourced, and more SaaS products will wrap these capabilities. The market will then shift to price competition — and those who built vertical authority early will hold pricing power.

The article closes with a blunt assertion: most people will read this, think about it for a few weeks, and do nothing. Those who book their first discovery call this week will own a $10,000/month business before summer ends.

References

• Nav Toor, “The full playbook for becoming an AI automation builder in 90 days”, X (Twitter), 2025. Original link