Financial Intelligence — Intuit × OpenAI & Anthropic
Defining how Intuit shows up in AI-native platforms — securely, contextually, with action. The reframe that took me from "Build Intuit in ChatGPT" to "Build financial intelligence where users already are." Now live in both ChatGPT and Claude.
- Role
- Principal Product Designer, Gen AI
- Partners
- OpenAI · Anthropic + IDX
- XFN
- Legal · Security · Engineering · Product · OpenAI · Anthropic
- Surface
- QuickBooks · TurboTax · Credit Karma · Mailchimp
- Outcome
- $100M+ partnership · 4 products unified
The problem
Users turn to AI to make sense of their financial data — outside Intuit.
The threat wasn't that AI got smarter. It was that users were leaving trusted financial tools to interpret their own data in tools that lacked context, accuracy, and trust.
Three pressures stacked:
- Workflows fragmented across QuickBooks, TurboTax, Credit Karma, and Mailchimp
- Users wanted to act, not just see — but actions stayed locked inside each product
- A trust gap: who sees the data, how it's used, what it does for me
That meant designing inside three hard constraints simultaneously: data security (explicit consent, scoped permissions, full user control), data freshness (stale answers in financial contexts aren't just wrong — they're dangerous), and context (generic LLM answers are noise; grounded answers are leverage).
The reframe
The reframe that changed the direction.
The hypothesis I argued for — and how it reshaped the system, the trust model, and the partnership.
Original framing
"Build Intuit in ChatGPT."
Ship a product. Users see a brand. The design optimizes for Intuit's presence — not for the user's financial life. One platform, one bet.
My reframe
"Build financial intelligence where users already are."
Solve the actual problem. Users get answers in context, with their real data, on a platform they already trust — whether that's ChatGPT or Claude. Intuit's presence follows naturally.
From product extension → intelligence layer that any AI-native surface can call.
From feature consent → data-flow consent. Users authorize where their data goes, not just what they enable.
From integration deal → strategic platform partnership. The deal got bigger because the design got sharper.
The system
A four-layer system.
Bringing Intuit's ecosystem into AI-native platforms — securely, contextually, with action.
Hover a layer to explore
- 04 Action Layer Trigger actions directly. Done-for-you workflows.
- 03 Cross-Product Orchestration Unified access. AI handles context switching across QuickBooks, TurboTax, Credit Karma, Mailchimp.
- 02 Financial Intelligence Core Context-aware insights across income, tax, business health.
- 01 Trust & Access Secure linking. Scoped permissions. Full user control.
Design principle
Structured outputs over long-form.
I shifted LLM responses from prose to actionable structure — so users could decide and act, not just read. 50+ unique structured-data concepts. One shared system across products.
Key insights at a glance. Replace paragraphs with the answer.
Trends and benchmarks. The "vs." that makes data legible.
Scannable, prioritized. The user's eye lands on what matters first.
Recommendations + next steps. The interface where reading becomes doing.
Orchestration
One system. Four products. Multi-step flows.
Each business unit had its own design system. I abstracted the complexity into a single coherent experience that holds context across steps.
Cash flow, expenses, customer signals — surfaced where the question is asked.
Complex tax questions become refund projections instantly. Live experts when stakes get higher.
Personal finance signals that move from passive monitoring to active recommendation.
Marketing actions tied to financial context. Connecting the customer-side and the business-side.
Each business unit had its own answer. I abstracted the complexity until the user only saw one.
In action — TurboTax in ChatGPT & Claude
From confusion → clarity → expert-backed action in a single flow.
-
01
Real-time tax estimates
Complex tax questions become refund projections instantly — answered in the surface where the user is already asking.
-
02
Personalized insights
Leverages Intuit's tax engine for accuracy. The model isn't guessing; it's calling a system that knows.
-
03
Live experts on demand
Hand-off to a human when stakes get higher. AI for breadth, humans for the load-bearing decisions.
Current chapter
Designed to be portable.
The architecture was never built for one platform. ChatGPT proved the thesis — Claude proves the layer.
Shipped — Intuit × OpenAI
$100M+ partnership.
Foundation laid in ChatGPT. The proof that financial intelligence as a layer beats financial product as a feature.
Shipped — Intuit × Anthropic
Live data integration in Claude.
Financial Intelligence extended to Claude. Same architecture, new surface — the layer is portable across any AI-native platform.
Impact
From integration deal to intelligence layer.
One more thing.
We won a Webby.
Actually, we won two.
Voted best by both judges and the public — out of 13,000+ entries from 70+ countries. 30th Annual Webby Awards · 2026.
Next case study