THE NUMBER: $4 — what it costs to unmask a pseudonymous employee's online identity. The practical obscurity era is over.
There's a pattern in how technology actually lands. It doesn't arrive with a press release. It arrives while you're watching something else.
AI agents arrived the same way. Not in the lab. Not in the pilot. In production — in your procurement system, your customer success queue, your IT ticket backlog. Making decisions. Touching live systems. Sending emails on your behalf without a human in the loop.
Three things happened in the last 72 hours that look like separate stories. They're not.
ETH Zurich and Anthropic published a paper showing agents can link pseudonymous Reddit and Hacker News accounts to real LinkedIn profiles at 67% accuracy, for $1 to $4 per person. The entire experiment cost under $2,000. The protection your employees had wasn't architectural — it was economic. The cost to breach it made mass deanonymization impractical. That calculation just changed.
At SaaStr, Jason Lemkin replaced 10 SDRs and AEs with 20 AI agents managed by 1.2 humans. Revenue held flat. Volume went from 7,000 emails to 70,000. The inbound agent closed over $1M in its first 90 days. His conclusion, January 2026: done hiring humans for SDR and AE work below enterprise tier.
Meanwhile, Gartner's projecting that embedded AI agents go from under 5% of enterprise applications to 40% by end of 2026. An 800% jump in one year. The same report notes that 40% of those deployments will fail by 2027 — not because the models failed, but because governance did.
We've watched this pattern before. Mobile hit enterprises the same way. The App Store opened, developers built on top of it, and a lot of them deployed features that violated terms they'd never read. The governance gap cost some of them their entire business. Not through malice. Through inattention. Enterprises running agents without documented accountability structures are in the same position today. The difference: Apple's terms cost you your app. An agent making autonomous procurement decisions without audit trails could cost you something significantly larger.
Here's what the Lemkin story actually is — it's not a job displacement story. It's a billing model story. The reps doing research, email drafting, lead qualification, and CRM hygiene were billing for task execution and calling it relationship work. Agents replaced the tasks. The reps who were actually doing relationship work will be more valuable. The ones in the middle are already replaced. The CFO just hasn't noticed yet.
The math isn't subtle when you run it: 10 people versus 1.2 people plus a platform subscription, same revenue, 10x volume. That doesn't require a board presentation. It requires a spreadsheet.
Three things to do this week. Pull a list of employees in sensitive roles and assume a hostile actor runs the ETH Zurich pipeline against them for $4 a profile — what's the exposure? Ask your CTO whether you can produce a complete action log for every agent running in production for the last 30 days. If the answer is "mostly," you have a governance gap. And run the SaaStr math on your own sales team.
The agents aren't approaching. They're operating. The gap isn't between capability and deployment. It's between deployment and accountability.
Build the audit trail before you need it.
-Anthony & Harry
On the site today: The full breakdown — deanonymization at $4 a profile, the governance gap inside enterprise agent deployments, and what SaaStr's sales team replacement actually tells you about your CFO's next question → Full Story
Should Read and share: The AI Race Is a Physics Problem - The AI race isn’t a software race. It’s a physics race, an organizational race, and a clock speed race, and most companies are losing all three.