Two cards, two independent operators. One is about AI automation discipline. One is about competitive moat strategy. Both published September 10 to 15, both from well-established operator voices.
Workflow codification is the prerequisite for AI inbound sales ROI
Tomasz Tunguz published "When Inbound Sells Itself" on September 15, 2026. He cited a Vercel outcome: 90% of inbound sales development automated, 32x ROI on the sales development agent, headcount reduced from 10 to 1.25 people, at a single-digit thousands annual infrastructure cost.
Codifying the best human workflow before deployment is the critical bottleneck for AI inbound sales agents, not model capability documents the mechanism behind that number. The result did not come from deploying a capable model. It came from first shadowing the best human rep, encoding every step and tool call in their workflow, running the agent in shadow mode until it exceeded 90th-percentile human performance, and only then removing the human.
"We had 90% automation of sales development for inbound. And our support agent that we've home-built handles 93% of all support cases…"
Jeanne DeWitt Grosser, COO of Vercel
The implication Tunguz makes explicit: the bottleneck in AI inbound automation is not model capability. It is workflow documentation quality. A company that has not codified what its best rep actually does cannot capture that practice in agent form. The agent will approximate a generic playbook and produce generic results.
This pairs with People closest to the problem should build their own software, not engineers with certifications from September 16, where Anton Osika argued that the people closest to a problem, not engineers with certifications, should build the software. The same principle appears here from the sales side: the person who knows what the best workflow looks like must document it before the agent can inherit it.
- Tomasz Tunguz, Codifying the best human workflow before deployment is the critical bottleneck for AI inbound sales agents, not model capability. Workflow documentation quality, not model capability, is the limiting factor in AI inbound sales automation.
Counter-positioning is the moat a startup can build before it earns the others
Packy McCormick published "An Ode to Counter-Positioning" on September 10, 2026. The concept originates with Hamilton Helmer's "7 Powers." McCormick applies it to current AI-era companies and argues it is the most accessible power for early-stage startups.
Counter-positioning works as a startup moat because incumbents face profit cannibalization if they copy, not because startups have technical advantages names the mechanism. The startup adopts a business model that, if copied by the incumbent, would cannibalize the incumbent's own profits. The incumbent must choose between matching the startup and preserving current margins. Most incumbents choose margins. That delay is the window in which the startup builds durable advantages.
Three examples McCormick uses. Ramp positioned itself around helping customers spend less. Corporate card incumbents earn from transaction volume. Matching Ramp's model would directly cut their own revenue. Base Power sells electricity rather than batteries, which requires incumbents to restructure their entire business, not just their product. Facebook launched Harvard-only, which contradicted MySpace's growth-at-all-costs approach; MySpace could not slow its own growth to match without disappointing its own investors.
The point McCormick emphasizes is about timing. Counter-positioning requires no scale, no network, and no technical advantage to activate. A startup with a genuinely different business model can access this moat on day one. The risk is treating it as permanent. The window closes when the incumbent decides short-term margin pain is worth eliminating the threat.
- Packy McCormick, Counter-positioning works as a startup moat because incumbents face profit cannibalization if they copy, not because startups have technical advantages. Counter-positioning creates a business-model conflict that prevents incumbents from copying without cannibalizing their own profits, giving startups time to build durable moats.