Empowering Evaluation with AI Summaries
Summary
InsightEval is a HIPAA-compliant, AI-powered platform designed to help forensic psychologists manage overwhelming case records and produce accurate evaluations faster. By combining AI-driven document summarization, chronological timelines, and an integrated report drafting workspace, the tool streamlines a process that traditionally takes days into hours, without compromising trust or clinical judgment.
Responsibility
End-to-End Process (User research, defining problems, workflow mapping, prototyping, usability testing)
Collaborators
Engineers
Clinical Advisors
Product Designer (that's me!)
Duration
5 months
Challenge
Psychologists are often tasked with reviewing tens of thousands of pages of unstructured medical and legal records per case. Outdated workflows, fragmented tools, and no centralized workspace mean that preparing a single court-ready evaluation can take 30+ hours, resulting in inefficiency, increased stress, and a higher risk of missed details.
Impact
26% faster report preparation (~3 hours saved per case)
Positive adoption feedback: “I finally feel like the tool works with me, not around me.”
Prototype is now used to demonstrate the product vision and future roadmap
User Problem
Users are overwhelmed by heavy caseloads and manual workflows, resulting in delayed evaluations and lower-quality results.
Research
Gathering insights
User quotes
“Sometimes I’m reviewing 5,000+ pages for one case. I’m constantly afraid I’ll miss something important.”
“I spend hours just reorganizing records into chronological order before I even begin writing.”
“I have templates, but they’re in Word, and I still need to reformat everything each time and copy and paste from one to another.”
“I’m open to AI tools, but it can’t replace my judgment. I need to see where the information comes from.”
Workflow mapping
Breaking down the entire process from start to finish helps to clarify each step and make the whole process easier to understand.
Case Intake
Receive referral + uploads bulk PDFs
Organize folders manually
Unstructured data
Hard to find what’s relevant
Document Review
Read 1,000–50,000+ pages
Highlight by hand
Use PDF viewers / printouts
Manual scanning
Can’t find things later
Overwhelmed 😩
Note-Taking & Synthesis
Take notes in MS Word or paper
Trie to build timeline & categories manually
High cognitive load
Duplicates effort
Mentally drained 😓
Report Writing
Open template in Word
Rewrite from scratch
Re-check original docs
Repetitive
Easy to miss citations
Frustrated 😠
Final Review & Submission
Proofreads
Email or upload the final report
Revise if needed
High pressure
Re-checking sources
Stressed
Key insights
1.
High cognitive load
Managing timelines, case complexity, and legal standards while reviewing huge volumes of records slows insight extraction and increases errors
2.
Fragile trust in AI
Users require transparency, source traceability, and full control over report content
3.
Inefficient and fragmented tools & methods
Switching between PDFs, notes, and Word documents disrupts focus and workflow. Locating specific evidence is slow and often inaccurate.
Define
Prioritization
With limited time and resources, we focused the MVP on must-solve priorities that balanced security, usability, and trust.
1.
HIPAA compliance
Non-negotiable. All outputs must remain secure, private, and ethically defensible.
2.
User trust & legal defensibility
Every AI output is editable, transparent, and traceable back to the source.
3.
Evaluator workflow variation
Psychologists work differently. Started with flexible templates and customizable components.
4.
Unstructured input data
Prioritized assistive tools (tagging, highlighting, summarizing) instead of risky full automation.
Goal
Streamline documentation and insight extraction to produce accurate, structured reports more efficiently
outcomes
Final Design
“Surface what matters”
User pain
Information overload from massive, unstructured records makes it slow and error-prone to find relevant details and organize them chronologically.
Design goal
An AI-assisted workspace that extracts key insights from large datasets, organizes them chronologically and by category, and allows evaluators to highlight and verify everything in context.
AI-generated case summaries
Summaries are only based on uploaded medical and legal records.
Psychologists can scan highlights instead of full records.
Reduce time spent scanning thousands of pages.
Source traceability
Every AI-suggested insight is expandable and linked to the original document for verification.
Source traceability
Every AI-suggested insight is expandable and linked to the original document for verification.
Timeline Viewer
Chronological grouping of events and medical history reduces manual time spent organizing timelines, which was identified in interviews as a top pain point.
Smart categorization with police reports, medical records, incidents, and testing)
Timeline Viewer
Chronological timeline of events and medical history
Timeline grouping based on uploaded medical and legal records.
Chronological grouping of events: Reduces manual time spent organizing timelines, identified in interviews as a top pain point.
Editable report drafting workspace
“Write with confidence”
User pain
Manual report writing is repetitive, unstructured, and time-consuming. Scattered notes and formatting tasks increase the risk of inconsistency and error.
Design goal
An AI-assisted workspace that extracts key insights from large datasets, organizes them chronologically and by category, and allows evaluators to highlight and verify everything in context.
Side-by-side report-building interface
End-to-end workspace improves from workflow fragmentation
Minimized switching between tools and stages, reducing cognitive overload
Editable section-based templates (Executive Summary, Clinical History, Mental Status, Competency Assessment, Recommendations)
Right-panel AI assistant for tone, clarity, and legal relevance suggestions
Supports internal reasoning and documentation of decisions—reduces rewriting from memory.
Users can quickly accept, edit, or reject AI drafts, mirroring their voice and logic, giving evaluators full control while saving time.
One-click export to court-ready format
Impact
The results
26%
faster report preparation
In early testing up, to 3 hours saved per case
Shaping product vision
Prototype now used in demos to showcase product vision and inform the product roadmap.
Positive user sentiment:
"Having everything in one place (records, notes, and the draft) has completely changed the way I work."
– Psychologist, early pilot user
Reflection
What I Learned
In high-stakes workflows, trust is as important as speed. Designing for clarity, control, and verification builds lasting user confidence.
Clear communication and visual cues are essential to explain AI’s reasoning and drive adoption.
Focus on core value delivery before expanding features
What’s Next
Design more engaging ways to showcase source references to strengthen trust and usability.
Run A/B tests on AI-powered writing suggestions to optimize clarity, tone, and adoption.