Making tax law feel human.

The law was written for regulators. The product was written for no one.

FBAR exists because US persons with foreign financial accounts over $10,000 must report them annually to FinCEN. The legal definitions are precise, the consequences of mistakes are severe, and the audience ranges from experienced investors to first-time filers who had no idea they were even required to file. The existing TurboTax FBAR content read the way it was written, by lawyers and for lawyers, which left a gap between accuracy and usability that the product had no way to close.

Here's how I planned the work, the timeline, and the cross-functional cadence.

Before writing a single word, I built the operating infrastructure. The FBAR project required unusually close collaboration across product design, legal, tax content analysts, and engineering. I established a structured cadence with the product design team that created fast feedback loops and dedicated time for open discussion. I set content principles early so every stakeholder shared a direction before any copy existed. And I built a decision log that proactively documented every content call and its rationale, which reduced back-and-forth and gave legal a record they could reference instead of relitigating decisions.

Fig. 01 Cadence, content principles, decision log. The infrastructure shipped before any copy did.

The content principles I established for FBAR were grounded in three ideas. Transparent meant being clear and direct about what TurboTax could do and where it stopped, building trust through honesty. Confident meant guiding users toward the right direction while showing tax expertise and customer awareness. Resourceful meant providing the right information without being overbearing. Every piece of copy passed through that filter before it went to legal.

Fig. 02 The decision log: each content change annotated, with rationale and legal approval tracked alongside.

I started with discovery and research.

I studied FBAR regulations through primary documentation and SME interviews, dissected existing solutions including the FinCEN BSA e-filing system and external FBAR services, and conducted user interviews and prototype testing. One filer in the interviews paused on the qualification screen, read it three times, and said "I don't know if this is asking me a question or telling me a rule." That was the pattern. The research uncovered three consistent pain points: confusion about FBAR eligibility, difficulty with data entry especially around currency conversion, and anxiety about whether they had filed correctly. FBAR caused anxiety, frustration, and confusion at proportions that went well beyond a typical tax flow.

Fig. 03 Discovery: regulatory documentation, competitive audit, prototype research. The brief for the prompt engineering work that followed.

User research surfaced what people needed at each moment in the flow: reminders and clear tracking, confidence that they had filed correctly, saved bank account information for returning users, and specific currency conversion guidance. None of those needs were being met by the existing content, which is what made the rewrite scope addressable through copy rather than only through new features.

I built a content engineering workflow on Writer.ai.

The core challenge was scale. Dozens of screens, each with legally specific language that couldn't be rewritten casually. Manual rewrites one by one would take months and introduce inconsistency. I built a five-phase workflow on top of Writer.ai with prompt engineering as the engine and human judgment at every decision point that mattered.

Fig. 04 Content engineering: analysis, prompt development, implementation. The five phases below describe what each one produced.

What the prompts looked like.

Here's the redesigned experience, screen by screen.

The work shows up in the screens. Every content decision was annotated with its purpose, hierarchy, and tone so the design team and engineering shared the same context. The content notes were a structured part of the design artifact rather than informal Figma comments, which is what kept the intent intact through to engineering.

The entry point from the main tax flow.

FBAR begins on the post-filing dashboard, alongside the federal and state returns the customer just submitted. The FBAR row is treated as a peer to the rest of the return, with a clear Get started CTA. Surfacing it here lets a Premier filer discover the obligation while they are still in the act of finishing their taxes, instead of finding out months later from a notice.

Entry Welcome dashboard with the FBAR row promoted alongside the federal and state returns. The Get started CTA treats FBAR as a peer obligation, not a buried link.

Before and after: the flow entry.

Fig. 05 Entry flow: from a generic income question to a direct FBAR identification.

Fig. 06 Account disclosure screens: restructured copy with clearer value framing.

The qualification flow.

Three screens carry the qualification logic: the dollar-threshold check, the ownership-or-control confirmation, and the marketing landing page that catches anyone who arrives outside the flow. Every step states the IRS requirement and lets the filer self-identify in their own words, instead of asking them to interpret regulatory language.

Qualification Threshold check, ownership confirmation, and marketing landing. Each screen lets a filer self-identify against the IRS rule in plain language.

The landing page: before and after.

Fig. 07 Landing page to in-product: copy focused on user benefit, with feature description in support of it.

Content notes in context.

Every key screen shipped with annotated content notes that documented the purpose hierarchy and the tone. These were a structured part of the handoff artifact and ensured engineering understood the intent behind every word.

08a FBAR welcome screen with annotations.

08b Error state with content notes.

Even the lightest screens carried notes. The data-transfer transition reads as a moment of trust: the system is doing work the filer would otherwise have to redo by hand. The note documents both jobs the screen has to do, so engineering and motion design ship the same intent.

Transition Even the lightweight transition screen ships with a content note. Two jobs: signal the transfer is happening, and earn credit for the work TurboTax is doing on the filer's behalf.

The "You Need to File a Report" screen required particular care. TurboTax couldn't file the FBAR directly at the time, and that constraint had to be communicated honestly without causing panic or breaking trust. Empathetic honesty at a high-stakes moment is one of the hardest content problems in compliance work, and the screen had to lead with what was true and actionable instead of with an apology.

Fig. 09 "You Need to File a Report." Purpose hierarchy: inform, be sympathetic, be optimistic about the future. Content notes documented tone, voice, and the PON ID and accessibility strings alongside, so the support team had aligned content at every channel.

The filing flow itself.

Three screens carry the actual filing: account entry, review, and sign-and-transmit. Each one keeps the legally immovable terms (FinCEN, Consent to Disclosure, Transmit FBAR) verbatim while letting everything around them speak in plain language. The Review step exists because the rewrite made the data feel light enough to lose track of, so the screen pauses the filer before transmission and shows the inputs in groups they can scan in a single read.

Filing Account entry, review, and sign-and-transmit. Legally immovable terms preserved verbatim; everything else rewritten to plain language. The Review step gives the filer one pre-transmission read of what they're about to sign.

The success state.

Fig. 10 "That's a wrap!" The completion screen reinforces value and sets up next year's filing.

Beyond the product: the email layer.

FBAR doesn't end at the success screen. Three email moments carry the rest of the relationship: a pre-filing nudge that frames the work ahead, a rejection note that holds trust when something goes wrong, and a confirmation that closes the loop with the IRS. Each one follows the same content principles as the in-product flow, so a filer hears one voice across product, email, and support.

Email Three email moments: an invitation, a rejection that holds trust, and a confirmation. Same voice as the product, so the filer never hears a stranger.

Three rewrites show the workflow at the line level.

Each rewrite preserved every required legal term while cutting reading level, eliminating passive voice, and reorienting the copy around the question a user was trying to answer. The output was copy that helped someone make a correct decision instead of copy that demonstrated regulatory expertise.

Here's why this matters now.