01 / Understand
Start with what the system can already know.
Kate opts in once. Her profile begins to assemble from available context, making progress visible while keeping the permission explicit.
AI behind the work.
Expert help when it matters.
An iOS design provocation using on-device machine learning to understand a tax situation, capture documents, anticipate issues, and bring in a human expert at the right moment.
Consumer AI / MLKate’s return,
already in motion.
already in motion.
Context gathers quietly. The right expert appears when judgment matters.
Make the relationship tangible.
Design was asked to create an iOS provocation for TurboTax Live. The prototype showed how AI could gather context, prepare a return, and bring a tax professional into the moments where judgment mattered.
Concept prototype / near-term capability plus a 12–18 month vision / not a claim that every concept shipped.
One experience, four moments for intelligence.
The concept followed one customer, Kate, through a connected tax experience. Four product bets gave the story structure and kept the AI tied to a recognizable customer need.
- 01First use
Use available context to reduce setup and create a more confident start.
- 02Document capture
Classify and extract tax information while the customer scans.
- 03Anticipatory intervention
Find issues early, prepare the detail, and route the decision to the right person.
- 04Year-round services
Extend help beyond filing season with timely, contextual guidance.
03 / Human + machineAI does the groundwork.
AI does the groundwork.
The expert owns the judgment.
The interface shows a clear handoff. The system gathers context, drafts the return, and flags what needs attention. A named tax professional reviews the consequential change and explains it to Kate.
Customer contextProfile · Documents · Prior year
Machine preparationClassify · Draft · Flag
Human judgmentReview · Explain · Recommend
Kate decidesAccept · Adjust · Ask live
A four-minute product story, built to be experienced.
The original prototype begins at the iPhone home screen and follows Kate from first-use permission through document capture, expert intervention, filing, and year-round guidance.
Original interactive prototype04:00
Recorded from the original build. Sound is available through the player controls.
Work out the system before polishing the screens.
The experience was mapped end to end, then annotated at the moment-to-moment level: what appears, what changes automatically, where an expert enters, and what remains for the customer to decide.
From background intelligence
to visible human help.
Each moment has a different job. Together, they tell one continuous story without turning the technology into the protagonist.
02 / Capture
Turn documents into useful progress.
Computer vision finds the relevant forms, classifies them, and carries the information into the return. The interface explains what is happening instead of hiding the automation.
03 / Intervene
Bring in a person at the consequential moment.
The system finds an issue and prepares the evidence. Eve, a named CPA, reviews the situation, explains the change, and remains available for a live conversation.
04 / Continue
Keep the relationship useful after filing.
The experience looks beyond the current return. Refund detail, a next-year estimate, timely notifications, and a live tax model turn a seasonal transaction into ongoing guidance.
The strongest AI experience may feel like better service.
This concept did not ask the customer to manage an AI tool. It used intelligence to remove setup, prepare information, and make expert help more timely and specific.
The technology recedes. The relationship gets stronger.
KeepContext, continuity, and clear progress.
ProtectPermission, explanation, and human accountability.
MeasureWhether people reach a better decision with less effort.
Next story / Platform strategyEnterprise-Wide