Anar Volset's research separates six supply-side and four demand-side shocks, beyond the common fears that customers will vibe-code apps or LLMs will absorb them. Examples include platform absorption, agency-built custom software, shrinking seat counts, and disappearing customer industries. Rob Walling says most founders have considered only two.
Rob Walling lists unique features, complex code, integrations, and UX familiarity as old moats. Anar Volset says AI can figure out complex workflows and browser agents can click through software, so those advantages shrink. High switching costs also weaken when agents, not people, do the work.
Rob Walling describes a platform you integrate with shipping your feature for free. Anar Volset says the spec is now your app: AI can crawl and screenshot it, and platforms move faster than ever. ServiceNow reportedly cut off a lead-management partner to own that feature itself.
Anar Volset says agencies using AI turn $30,000–$50,000 projects into a couple thousand upfront plus $500–$1,000 monthly maintenance and support. That custom, SLA-backed alternative can undercut a mid-market SaaS charging $50,000 a year. It is a supply shock, not just a threat to tiny tools.
Anar Volset cites Zendesk: if AI resolves 80–90% of customer support tickets, a 15-person team becomes two or three people monitoring agents. Per-seat revenue falls 70–80% even if ticket volume grows. HubSpot faces a similar problem when AI reduces the number of writers customers need.
Anar Volset points to translation and transcription agencies whose revenue effectively collapses because AI does the work. The volume of translation and transcription may rise, but software sold to those agencies loses its market. If the buyer's business model dies, demand often goes to zero.
Anar Volset separates tools that recommend actions from tools authorized to move money, like Stripe Connect or Reambi. He also cites Gusto filing taxes that users never audit. When customers must check every action, replacing the software is easier; when it is trusted plumbing, it sticks.
Anar Volset says HIPAA, SOC 2, eSign validity, and ABA/CEU accreditation and certifications take months and carry real liability. Rob Walling cites SignWell: eSign looks easy until a million-dollar contract is disputed. Founders likely will not vibe-code a compliance replacement just to save $500 a month.
Anar Volset cites school apps that parents hate but schools will not move thousands of families. Slack is similar: switching means coordinating a whole organization. The moat is the pain of collective migration, not product quality. AI can clone features but not the install base.
Rob Walling cites Peter Levels paying for ScrapingBee because AI cannot manage proxies and CAPTCHAs. Anar Volset compares email deliverability: one customer cannot hire 25 people to fight spam and IP issues. The cost only makes sense amortized across many customers, so shared infrastructure remains sticky.
Anar Volset says agents will directly interact with software. Instead of locking them out like Amazon did with Meta's bot, offer an MCP and charge for agent access. He calls it a new kind of SEO: getting agents, not humans, to actually select your solution.
Anar Volset urges founders to become the worker: if customers hire employees to click your tool, sell the completed outcome instead. Labor budgets dwarf software budgets, so a $75 monthly tool could charge $750. Existing customer trust gives bootstrappers a strong edge over AI-native startups.
Anar Volset argues a learning loop where an AI worker acts—resolving tickets, issuing refunds—and learns from outcomes, then transfers that learning across customers, is the one moat likely to stay ahead of expanding model capability. Static data is not enough; the system must keep learning.
