Podcast
Scaling human expertise with AI
Jay Pandya on AI avatars that deliver personalised, multilingual support around the clock, and the question that governs the whole design: how do you scale a human without losing the human?
How do you scale a human without losing the human touch?
This is the design constraint that makes the difference between an AI deployment people trust and one they route around. Automation that removes the person entirely is easy to build and easy to resent. Automation that extends a specific person's expertise is harder, and it is the one that works.
What an AI avatar is for
The episode works through the practical form: a system that carries a known expert's knowledge, answers in the customer's language, and is available at any hour, while being explicit about what it is and when a human takes over.
- Personalised support that reflects one organisation's actual expertise, not generic answers
- Multilingual delivery without staffing every language separately
- Round-the-clock availability without round-the-clock rostering
- Clear handover to a person when the question warrants one
Pricing it
Jay also gets into the commercial side, which is the part most discussions skip, how you price a solution whose marginal cost is near zero but whose value is tied to the expertise it carries. Cost-plus pricing badly undervalues this category.
From Brisbane to a billion
The conversation closes on the arc from an Australian startup to the ambition MVP1 works towards: technology that reaches a billion lives, built from a country that is not usually where that sentence starts.
Key takeaways
- Design constraint: extend the expert, do not erase them.
- Multilingual, always-available support without proportional headcount.
- Explicit handover points are what make the system trustworthy.
- Near-zero marginal cost breaks cost-plus pricing, price on value carried.