Capital One AI-powered decision support interface for compliance evaluation
Case Study

AI-Powered Decision Support

Building Capital One’s first human-in-the-loop AI feature for its risk management platform and defining reusable interaction patterns for future AI systems.

Explainable AI AI Decision Support Enterprise Risk Cross-Functional Collaboration
Role UX Designer (Sole Designer)
Timeline Feb 2026 – May 2026
Domain Enterprise Financial Services, AI Risk Management
Tools Figma, Gemini

Overview

I designed and integrated an AI-assisted evaluation tool into Capital One's risk management platform to improve data quality at the point of entry. The work delivered a production decision-support feature and defined reusable AI interaction patterns to support consistency across future features in the platform.

Traditional Workflow

Analysts were responsible for writing complex control descriptions that had to satisfy evolving compliance requirements. Quality issues were often identified only after submission, creating a reactive workflow that required revisions after Reviewer analysis. This delayed validation increased uncertainty, slowed delivery, and allowed quality issues to propagate downstream.

Compact vertical workflow of centered cards, all labels in sentence case and centered within each card: User writes & submits control description (on a wider first card, single line), then Reviewer analyzes description, styled with the same neutral light gray card as every other step, which forks via muted blue-gray connector lines into two side-by-side outcomes at the same visual level, Reviewer approves and Reviewer requests revisions on a wider card so its label fits on one line, with a faint blue-gray dotted loop on the right from Reviewer requests revisions all the way back to User writes & submits control description
Control descriptions were often not reviewed until a downstream issue occurred.

AI-Assisted Workflow

I redesigned the workflow by embedding AI evaluation directly into the authoring experience. Instead of discovering quality issues after submission, analysts receive immediate feedback while writing, allowing them to resolve issues earlier while preserving human decision-making.

Compact vertical workflow matching the Current Workflow's style, all labels centered within each card. The first four stages, User writes control description, AI analyzes description (the only step with a distinct light blue background), User evaluates AI analysis, and User submits description to Reviewer, are grouped inside a subtle rounded container. Below and outside the container, Reviewer analyzes description uses the same neutral light gray card as every other standard step, forking via a muted blue-gray connector into two side-by-side outcomes, Reviewer approves (no outgoing arrow) and Reviewer requests revisions on a wider card extended to align its right edge with the card above, with a faint blue-gray dotted loop running from Reviewer requests revisions straight up outside the container's right edge to the height of the first step, then making a 90-degree left turn into a short horizontal segment ending in a leftward-pointing arrowhead whose tip touches the right border of User writes control description without crossing into the card's interior, representing a return to the start of the AI-assisted workflow
Users could refine descriptions with AI before submitting them to Reviewers.

Design Strategy

The AI interaction model was guided by four design principles. Each principle was translated into concrete interaction patterns that balanced usability, explainability, and responsible AI communication.

Design Principle UI Implementation
Embed AI into the workflow Integrated AI evaluation directly into the authoring experience instead of a separate workflow or chat interface.
Communicate AI responsibly Defined AI confidence, scope, status terminology, and color semantics to communicate AI findings without implying certainty or correctness.
Support informed decision-making Designed per-criterion explainability, attribution, and supporting evidence to help analysts understand and verify AI recommendations.
Establish reusable interaction patterns Created consistent UI patterns that could support future AI capabilities across the platform.

Analysis Placement

The AI evaluation was embedded directly alongside the authoring experience rather than presented as a separate workflow or chat interface. Analysts could review AI feedback while editing the control description, eliminating the need to switch contexts or navigate away from their work.

Split-panel interface shown in a smaller browser window with generous left and right margin: analyst form on the left, AI report with findings on the right.
Authoring and evaluation share the same workspace, so analysts can review AI findings without interrupting the writing process.

Analysis Breakdown

The AI Analysis panel was organized into four sections, separating AI context, evaluation results, supporting attribution, and user feedback into a clear, structured workflow.

An AI Analysis card showing the production AI Analysis Panel, with four bracket annotations outside the panel identifying its major sections, each labeled with a simple heading: AI Context (the AI Confidence indicator, explanation, disclaimer, and Learn more link, with three bullets: Overall AI confidence, Evaluation scope, Disclaimer), Report (the list of five evaluation results, with three bullets: Evaluation criteria, Detection results, Transparency), Attribution (the policy reference section, with two bullets: Supporting references, Source traceability), and User Feedback (the Was this helpful control).

AI Confidence Indicator

Presenting AI confidence as a progress bar and percentage blurred the distinction between confidence and analytical results already shown throughout the interface. High, Medium, and Low communicated the same information while keeping confidence as supporting context.

AI Confidence Indicator Design Options. A neutral side-by-side comparison of two AI Analysis cards in the same card style and visual weight. The left panel, headed Percentage-Based Indicator, shows an 87% confidence value with a graphical progress bar, above Volume and Trend charts. The right panel, headed Qualitative Confidence Label, shows a text-based Medium confidence label in place of the progress bar, followed by the same Volume and Trend charts. The two panels present two different UI approaches without indicating either as preferred; the underlying charts are identical in both.
A qualitative confidence label distinguishes AI confidence from analytical metrics.

Per-Criteria Status Indicators

Control descriptions were evaluated against multiple quality criteria. The AI generated a report showing whether each criterion was detected within the description.

Status Indicator Terminology

Multiple terminology options were evaluated to communicate AI evaluation without implying certainty or correctness.

Option Consideration Result
Pass / Fail Implied a final judgment.
Present / Missing Described the criterion rather than the evaluation.
Found / Not Found Resembled search results.
Detected / Not Detected Framed the result as an evaluation rather than a definitive system status. ✓ Final

Status Indicator Colors

Multiple color options were evaluated using established UI semantics to reinforce AI evaluation without implying success or failure.

Option Consideration Result
Green / Red Implied success and failure.
Blue / Orange Communicated information and warnings. ✓ Final
An AI Analysis card showing the final interface, with five rows marked Detected in blue or Undetected in orange, each followed by a small circular question-mark info icon. Two grouping brackets sit to the right of the card, both spanning the same vertical range as the five status pills, each connected by a short horizontal line to a labeled annotation: Status, with two bullets, Detected / Not Detected terminology and Evaluation-oriented language; and Colors, with two bullets, Blue indicates informational results and Orange indicates warnings.

Per-Criteria Explainability

User research showed that AI findings alone were insufficient. Analysts wanted to understand how each conclusion was reached before acting on the recommendation. An on-demand explanation experience provided AI conclusions, supporting evidence, reasoning, and policy context without overwhelming the primary interface.

The AI Analysis panel, reused unchanged from the Status Labels and Colors pattern, sits on the left with five rows (Access control review, Third-party vendor, Data retention, Access logging, Incident response) marked DETECTED in blue or UNDETECTED in orange, each with a small circular ? icon. A dotted leader line runs from the first row's ? icon (Access control review) directly to the AI Conclusion and Explanation modal on the right, shown at full size with a soft drop shadow that reads as an overlay rather than a separate screen. The modal identifies itself with an Evaluation: Access Control Review subheader, then shows AI Confidence, an AI Conclusion (Status: Detected, Suggested Action: Escalate to Tier-2 Manual Review) set apart by a left accent bar, and quieter supporting sections for Input Data, Reasoning Path, and Policy Context. Together the two elements communicate a single idea: selecting an evaluation's ? icon opens an on-demand explanation for that specific finding.
The AI Conclusion and Suggested Action lead the modal as the primary output, with supporting evidence given a quieter visual treatment below.

Impact

Qualitative Outcomes

Outcomes

Reflections