Uploaded August 2026 | Updated September 2026, 3 weeks ago
Nine months after Aliisa Rosenthal started begging for enterprise features, OpenAI finally shipped them. By then almost every company on the list had the same answer: you never got back to me, so I bought Microsoft Copilot. Rosenthal joined OpenAI at a couple of million in revenue and helped take the enterprise side to several billion, and the lesson drawn is that the sequence was wrong at nearly every step. They built for the loudest voices: a fast, expensive, aggressively upmarket product. Four months later came self serve, which grew so much faster it cannibalized the enterprise business and left reps competing with their own funnel. Self serve should have shipped first, with the enterprise tier built from what it revealed.
The advice to founders inverts the usual order: build the machine, then hire the humans. Automate what you can, find where it breaks, and add people at those bottlenecks rather than staffing up first. Capture more at signup, phone number included, and reply to everyone even if automatically. Never send a buyer away with homework; you lose control of the cycle. Treat pilots as a trap, a second sales process you run for free: reserve them for the largest deals and negotiate around them with a reference call, an eval on a slice of data, or a 90 day opt out that puts the clock on the customer. Automate security questionnaires and trust portals, where deals quietly die. Sixty dollars per user per month proved too high; a low base fee plus usage dropped the barrier enough to spread company wide.
Speaker info:
- linkedin.com/in/aliisa-rosenthal
- acrewcapital.com/team-members/aliisa-rosenthal
Timestamps:
0:00 - From a couple million to several billion at OpenAI
1:03 - Build the machine before you build the team
2:05 - Nine months without enterprise features
3:08 - Self serve launched late and cannibalized enterprise
4:24 - Ten thousand inbound a day, and what to do about it
5:52 - Do not give your buyer homework
6:53 - Pilot hell, and how to get out of it
8:35 - Automating security so deals stop stalling
9:26 - Pricing too high, then fixing it
11:08 - When to finally hire salespeople
16:08 - Q&A: forward deployed engineers and design partners
Nine months after Aliisa Rosenthal started begging for enterprise features, OpenAI finally shipped them. By then almost every company on the list had the same answer: you never got back to me, so I bought Microsoft Copilot. Rosenthal joined OpenAI at a couple of million in revenue and helped take the enterprise side to several billion, and the lesson drawn is that the sequence was wrong at nearly every step. They built for the loudest voices: a fast, expensive, aggressively upmarket product. Four months later came self serve, which grew so much faster it cannibalized the enterprise business and left reps competing with their own funnel. Self serve should have shipped first, with the enterprise tier built from what it revealed.
The advice to founders inverts the usual order: build the machine, then hire the humans. Automate what you can, find where it breaks, and add people at those bottlenecks rather than staffing up first. Capture more at signup, phone number included, and reply to everyone even if automatically. Never send a buyer away with homework; you lose control of the cycle. Treat pilots as a trap, a second sales process you run for free: reserve them for the largest deals and negotiate around them with a reference call, an eval on a slice of data, or a 90 day opt out that puts the clock on the customer. Automate security questionnaires and trust portals, where deals quietly die. Sixty dollars per user per month proved too high; a low base fee plus usage dropped the barrier enough to spread company wide.
Speaker info:
- linkedin.com/in/aliisa-rosenthal
- acrewcapital.com/team-members/aliisa-rosenthal
Timestamps:
0:00 - From a couple million to several billion at OpenAI
1:03 - Build the machine before you build the team
2:05 - Nine months without enterprise features
3:08 - Self serve launched late and cannibalized enterprise
4:24 - Ten thousand inbound a day, and what to do about it
5:52 - Do not give your buyer homework
6:53 - Pilot hell, and how to get out of it
8:35 - Automating security so deals stop stalling
9:26 - Pricing too high, then fixing it
11:08 - When to finally hire salespeople
16:08 - Q&A: forward deployed engineers and design partners
