 | Every one of these clips lands the same blow eventually: a role someone spent years building gets quietly outperformed by a system that never clocks out.
Kavak sells used cars across Latin America — a genuinely messy transaction: ~20,000 SKUs to choose from, then financing, insurance, and a trade-in valuation stacked on top. Historically, closing one sale meant routing a customer through 15 separate human specialists across 15 different teams, each holding one piece of the process.
Alejandro Maza Ayala, Kavak's Chief Product & AI Officer, explained on a16z's show how they fixed it — not by making a support bot, but by building a single "mega-expert" agent that holds all 15 specialties at once (financing, insurance, trade-in, advisory) and puts that one agent in front of the customer.
The result: 2.1x the conversion rate of their own human sales team, tripled customer satisfaction. The agent never tires, never forgets a customer's history, and when it makes a mistake, the correction propagates to the other 200,000 agents in the fleet by the next morning — a scale of self-correction no individual human career can match working alone.
It closes on Alejandro flatly stating that the industry assumption — "customers aren't going to want to buy expensive things from AI" — is wrong, and Kavak's numbers are the proof.
When I read the transcript, it felt so eerily similar to the Borg Collective Mind in Star Trek. That's the ultimate evolution. The question we need to ask is, will it serve us, or subjugate us?
If your role is the coordination layer between departments — the person routing a customer between financing, insurance, and everyone else — that's precisely the layer this consolidates first. Worth sitting with, not scrolling past.
Clip credit: a16z — full video on their channel. DM for credit or removal requests.
Drop your take below. [link] [comments] |
Predictive technology: computational social science