Minna Song didn’t have a single ‘aha’ moment about AI. She had a problem — housing and healthcare were drowning in administrative inefficiency — and AI looked like the only solution that could scale. That was 2017, and the bet is paying off. In a recent conversation for our Executive Function series, the EliseAI CEO walked through how her company moved from older models like BERT to today’s generative systems, and why the jump to GPT-4 changed everything.
Song explained that early on, the challenge was less about raw model capability and more about design psychology. In non-technical industries, ‘natural’ meant familiar. EliseAI had to replicate existing workflows so closely that users felt like a task was simply being done faster, not handed off to a mysterious black box. That’s shifting now. As AI literacy grows, the company is moving from mimicking old processes to completely rethinking them. But Song cautions startups still need to ‘toe the line’ when introducing brand-new applications to skeptical audiences.
Voice was the missing piece for years. EliseAI built traction over text channels in housing, but healthcare was a different beast — nearly all communication happens over the phone. Without models like Whisper, Song says entering that industry was impossible. ‘We were yearning to solve phone calls, but before those models, the tech just wasn’t there. It wasn’t even close.’ Now voice AI is operational, and it’s unlocked a whole new vertical.
Success metrics are refreshingly pragmatic. The team benchmarks AI performance against the best human agents, tracking how much of a workflow can be fully automated. In housing, that means occupancy rates and maintenance resolution times. In healthcare, it’s about service quality. Internally, EliseAI eats its own dog food — every department from sales to finance uses AI components to scale. Song argues this gives her team a sharper intuition for what customers actually need when they ask for new features.
