Designing a strategic vision for Intuit's AI × Human Intelligence Platform

There's a quiet irony at the heart of the AI revolution: the better AI gets at answering questions, the more urgent the human becomes.

At Intuit, I work within the Virtual Expert Platform — managing close to 40,000 experts during peak season. And for years, the internal conversation has been some version of the same question: how do we automate them away?

I believe that's the wrong question. In a world of ubiquitous AI, our experts aren't a liability to optimize out — they're the one thing competitors cannot replicate. The differentiator isn't the model. It's the 40,000 people who bring judgment, nuance, and trust to the moments that matter most.

The question I wanted to answer: how do we make their expertise scale infinitely?

Today, our world-class expertise is locked in people's heads, Slack threads, and one-off docs. Every expert interaction at Intuit contains judgment that disappears the moment the call ends. We came up with Structured Expertise Tokens (SETs) — a way to convert unstructured expert interactions and knowledge into machine-consumable knowledge assets. Not just what an expert said, but what they did across interactions.

Behind every customer question is a routing problem: who has solved this before? The platform builds a living knowledge graph of People × Content × Domains — mapping which experts know what, which issues they've solved, and where the relevant artifacts live. It can answer: "How have we solved this before?" and "What would our best experts do here?"

SETs power Expert Twins — AI models trained on a real expert's actual work and judgment. Every key expert gets a twin that captures their knowledge, patterns, and reasoning, available 24/7 across any surface. Routine questions are handled with clear attribution and guardrails. Nuanced or high-stakes issues route back to the right human — with full context pre-packaged.

Tokenizing expertise means the same SETs flow across every surface — internal tools like IEP and Slack, customer-facing Intuit products, and external AI integrations like ChatGPT. The knowledge isn't siloed to a single product; it becomes infrastructure that scales across the entire ecosystem.

For customers who never interact with a live expert, every solved problem becomes a service adoption moment. An AI-powered answer that references a real expert's judgment can surface a natural next step: "Want to go deeper? Work with Emma for your business taxes." Human intelligence becomes the bridge to human connection.

And because the platform controls how expertise is used — not just captured — it can enforce data guardrails, apply attribution, and route complex cases back to the right human with full context pre-packaged. The goal isn't to replace experts. It's to make their knowledge permanent, portable, and protected.