Designing a clearer path to trusted financial insights.

jAI is an AI-powered financial assistant within the Jumpp app. The landing screen was intended to introduce users to personalized financial insights and encourage them to connect their bank account.

The landing screen had to introduce a new AI interaction without undermining the familiar financial tools users already trusted, such as Spends and Budget.

Jumpp users, reached through a pool of 25 participants. They came to the screen already using Spends and Budget, the tracker tools jAI had to sit beside.

Product designer on Jumpp. I got the opportunity to conduct UX research with real users, and I ran the sprint from the questions to the proposed design response: research, strategy, and interaction design.

We had a short window to learn from real users while the product was still moving, and no research program to run it through. Instead of waiting for the “perfect” study, we used the resources available to us: the existing screen, lightweight Google Forms, and a shared findings sheet.

• Constraints

  • Limited time to recruit, test, and synthesize.
  • No dedicated usability-research platform.
  • A small, available participant pool (25 users).
  • No time to build a fully instrumented prototype.

• What we did instead

  We used the existing jAI screen, created lightweight forms, asked first-impression questions, and consolidated the responses into a shared report. We treated the output as signals and hypotheses to validate later, not a verdict.

We showed participants the existing jAI screen, asked first-impression questions, and pooled the answers in one report. Six findings came out of it.

1. 01

   Finding 01

   Users approached the screen as a tracker first.

   Spends and Budget were the dominant first actions (76%). Only a small number of participants said they would interact directly with jAI.

   Design Implication

   jAI should complement the tracker mental model instead of competing with it. The experience needs to explain when jAI is useful after or alongside Spends, Budget, and Goals.

2. 02

   Finding 02

   The screen caught attention without orienting anyone.

   The jAI animation and "Explore" were the most visible elements. Motion attracts the eye without explaining what jAI can do or what the user gets from it.

   Design Implication

   Retain motion only as a supporting brand or feedback element. Pair it with a concise value statement, example prompt, or visible insight outcome.

3. 03

   Finding 03

   The path to personalized insight was ambiguous.

   Users saw the CTA. They had no mental model of the sequence behind it: discover jAI, understand what it can reveal, preview the type of insight, and connect data to personalize that insight.

   Design Implication

   The interface needs a visible value path alongside the AI presence. A user should be able to say what they will get before we ask them to link data.

4. 04

   Finding 04

   Linking comprehension was stronger than linking confidence.

   Participants understood the likely outcome (account linking), but comfort stayed moderate. Understanding the outcome did not make them confident about it.

   Design Implication

   Improving the explanation of the linking outcome is necessary but insufficient. The flow must communicate data boundaries, user control, and the practical benefit of linking.

5. 05

   Finding 05 & 06

   Trust signals matter, and digital comfort doesn't remove the need for oversight.

   Security, privacy, data misuse, and trust in AI were recurring concerns. Trust signals were most useful when they answered a specific concern about data access or control right at the decision point. Even digitally confident users required visibility and explanation.

• 1

  Create three clear entry points

  Support the mental models already present on the screen: Track (Spends/Budget), Plan (Goals), and Understand (jAI insights).

• 2

  Concrete value path for CTA

  Communicate the outcome ("Get personalized insights"), answering what kind of insight, why data is needed, and what happens next.

• 3

  No-commitment sample insight

  Use Sample Data View or an example insight to demonstrate value before requesting access. Distinguish it from standard trackers.

• 4

  Design linking around control

  Explain only accurate capabilities at the decision moment. Clarify why linking is needed, what is accessed, and how to revoke it.

• 5

  Animation as support, not explanation

  Keep animation if it contributes to personality, but provide stronger orientation cues through copy and clear hierarchy.

Going in, I would have made jAI more prominent and called it solved. Participants showed me they already had a way to start, and that jAI had to earn both its usefulness and their trust before they would link an account.

Reflection & Limitations

The study moved the problem from discoverability to trust and mental models. It has clear limits:

• We need to observe participants complete real tasks rather than only reporting intent.
• The complete linking flow must be tested with a safe prototype.
• Future rounds should standardize question wording, preserve respondent-level data, and test the same instrument across all variants.