An AI finance chatbot that earns trust before it asks

Made financial guidance feel safe from the very first question.

theAI.Money began as a follow-up to We FIRE, our team’s earlier financial planning platform. We FIRE gave users dashboards and manual data entry: useful, but demanding for someone with little financial experience.

The team wanted to lower that barrier. Instead of asking users to organize their finances first, theAI.Money uses AI conversation as the way in: ask a question, get guidance, no setup required.

The solution at a glance

Four ways in, one outcome

However much the user has shared, the assistant answers first, asks only for what would change the guidance, and confirms before using anything.

AGuest, no dataChoose how to shareEnter details or connect a bankBSigned in, no dataConnect only what’s neededOne card, nothing elseCReturningConfirm stored details“Are these still right?”DPartial dataUse what’s knownAsk for the one missing detail

Same outcomeGuidance they can act onWhere the $200 goes, the sources, and a next step

People were already asking AI about money. They just didn’t trust it with their numbers.

By 2024 the pattern was visible. Many Americans were using AI to help manage their finances, yet far fewer trusted it as much as a human advisor. People were willing to ask AI about money, but handing over specific financial details raised real concerns about privacy and how that data might be used.

Sources: BMO Real Financial Progress Index (Ipsos, Jul 2024, n=2,501); Experian (Atomik Research, Oct 2024, n=2,011).

For the chatbot, this trust gap showed up at a critical moment: getting started. The product needed personal context to give relevant guidance, but asking for financial details too early could make users hesitate. We needed to understand what people were comfortable sharing, what they expected in return, and what would help them feel in control.

So the challenge wasn’t building smarter AI. It was making people comfortable enough to use it.

Ran three research streams to learn what makes newcomers willing to share.

User journey: where engagement breaks down

Concerned → hopeful → overwhelmed → discouraged. The drop after the first answers is where the opportunity sits.

!

The first AI response decided whether users continued

Asking AI a financial question and sharing personal details were separate decisions. Participants wanted relevant guidance, a manageable next step, and a clear reason to provide more context. That led me to the tools themselves: how did existing assistants give initial value, request information and help users move forward?

From 12 responses to one template

I compared the 12 tools’ responses to the same car-loan payoff scenario, mapping how they framed the problem, used context, explained options and suggested a next step. The comparison became an ideal response template: understand the question, make the options clear, then suggest a manageable next step.

Found the problem wasn’t a lack of financial advice.

Financial newcomers needed a first step that felt safe enough to try, clear enough to evaluate, and concrete enough to act on.

Mapped how one response template adapts to every way a user could arrive.

The interaction map covers entry, optional onboarding and four account conditions. Each path adapts the conversation to the context already available, while following the same principle: provide value first, explain every request, and leave the user in control.

Entry & optional onboardingUsers can ask a question immediately; sign-up and onboarding appear only when useful.

Guest · no stored dataOffers an initial answer without requiring sign-up or personal financial information.

Signed in · no financial dataLets the user begin without completing a financial profile.

Signed in · partial dataUses available context and asks only for the missing information that would change the guidance.

Returning · relevant data storedShows what’s already stored and asks users to confirm or update it before giving guidance. Nothing is reused without their approval.

Designed four starting points that converge on one shared outcome.

Each path collects or confirms context differently, then arrives at the same personalized guidance once the required information is available. Three decisions carry the trust-first idea across all four paths.

Decision 01 · Path B

Preview what will be shared, before connecting

Before a card is connected, the assistant lists exactly what it will read. Card balance and purchase APR are marked Requested; transaction history is Not requested. Entering details by hand stays one tap away.

From researchUsers hesitated when a tool asked for financial details before showing how they would improve the guidance.

Decision 02 · Path D

Ask only for the one detail that’s missing

Balance and APR are already known, so the assistant asks for the monthly budget alone and says what it will be used for. Keep it general is always offered, so sharing never becomes a gate.

From researchThe strongest AI responses used the context already available and gave every follow-up question a purpose.

Decision 03 · Paths A, B, D

Show how each detail will be used, then confirm

Before guidance is calculated, a table pairs every input with its purpose: balance sets the starting debt, APR the interest cost, budget the payment pace. Users confirm or change a detail. Nothing is reused without their approval.

From researchProduct reviews showed users need to see why information is requested, and stay free to change course.

Learned that trust has to come before the ask.

Users weren’t refusing AI. They were waiting for it to earn their numbers.

Next On

ProjectWorkfx.AI ProjectPet Bridge ProjectPuyue MCS All workBack to projects