Black Lake founder Yuxiang Zhou staged a rare internal experiment. He put an AI coaching agent against his company's top human sales trainer and let the reps decide. His most experienced salespeople chose the AI, Fortune reports.

The setup matters more than the verdict. Few sales leaders run a controlled A/B test (a head-to-head comparison with two groups) inside their own team; even fewer publish the result. Zhou's test hints machine coaching may already match the human alternative for veteran closers. The dimensions reps feel most: speed, recall, and consistency.

What's actually being tested

A coaching agent is software that scores sales calls, flags missed questions, and prescribes next steps between live deals. It reviews every rep's recorded calls and sends notes the same day. A top human coach hears a fraction of those calls and delivers feedback in scheduled one-on-ones (recurring manager-rep sessions).

For senior reps, the trade-off tilts in the AI's favor. Rookies need a human to teach the basics of discovery (the call stage where reps uncover the buyer's problem) and objection handling (rebutting buyer pushback). Veterans already know the playbook; they want a faster loop.

What the methodology needs to show

The Fortune headline reports the outcome but not the sample, deal stages, or win-rate delta (percentage change in deals closed). Three numbers need answering before any operator copies the playbook:

• How many reps were in the test?
• Which deal stages saw the biggest shift?
• Was the AI measured on call scores, closed-won rates (the percentage of deals resulting in a signed contract), or rep self-report?

Vendor claims are data to interrogate, especially when the vendor is your own company. Zhou's experiment is also a sales asset — Black Lake sells AI tools — so the methodology deserves the same scrutiny any third-party benchmark gets.

What changes for sellers and leaders

For reps, the takeaway is narrow: if you're already at quota (meeting your assigned sales target), an AI coach can replace the parts of your manager's job you don't actually need. For managers, the implication cuts deeper.

The most experienced sellers — the ones you usually protect from process changes — are also the first group to opt out of human coaching when the alternative is fast and consistent.

For mid-sized teams of roughly 20 to 80 reps, this points to a real cost question. A senior human coach costs six figures fully loaded (including salary, benefits, and overhead); an AI seat costs a fraction of that. If senior reps genuinely prefer the AI, the math shifts quickly.

The next question

Which parts of the human coach's job did the AI replace, and which did it leave behind? Strategy, deal-by-deal coaching on a stuck account, and career mentorship aren't the same job as scoring a call and flagging a missed question. Until Zhou or Black Lake publishes the full breakdown, treat the headline as a directional signal. Don't run it as a playbook yet.