Scopio Labs

What happens when AI isn't sure?

Product Designer B2B SaaS Hematology AI
Default view
Split view

~15–30 min manual diagnosis / case

Cell by cell ↓ Validated in bulk
AI uncertainty ↓ Decisions with context
Microscope ↓ Trust Scopio
Background

Scopio's AI-powered scanner captures and classifies thousands of blood cells per slide. Pathologists review the AI's work before signing off.

Problem

The review workflow forced cell-by-cell validation: slow, tedious, and misaligned with how experts actually think. AI confidence scores had no clear mapping to action.

Approach

Mapped expert decision patterns, identified where AI ambiguity stalled the flow, and redesigned the interface around bulk validation with contextual confidence.

Result

A redesigned Cells Review enabling bulk validation, contextual AI confidence, and split-view comparison.

Overview

The product: digitalizing hematology

Scopio transforms the traditional microscope-based lab into a fully digital workflow.

1

Hardware: Scans physical slides at high resolution (100X).

2

AI Software: Automatically detects, classifies, and groups thousands of cells.

The Shift: Experts move from manually searching for cells under a microscope to validating AI suggestions on a screen.

Product overview
Clinical Laboratory Scientist persona

The users

  • Clinical Laboratory Scientists (CLSs) working under constant pressure to reduce Turnaround Time (TAT)
  • Increasing test volumes with limited staffing
  • Demand safety and accuracy above all, willing to trust the system but need to feel in control

The Product Challenge

Solving the "Unclassified" friction

Finalizing a BMA report requires manually classifying dozens of cells below the AI's safety threshold. Currently a slow, single-cell process.

The Solution

Safety-first bulk classification

Smart validation interface grouping uncertain cells into tentative suggestions. Experts process dozens of cells in seconds while maintaining full control.

Challenge and Solution overview

Design challenges

01

Designing for low-confidence AI

Surface low-confidence AI suggestions without creating diagnostic bias. The system needed to inform without leading.

02

Integrating bulk actions

Add high-speed validation tools without disrupting established single-cell workflows. New affordances had to feel native.

Logic mapping

Logic Mapping diagram

The solution

Designed screens

Research-led screens refined through wireframes and aligned with development constraints.

AI suggestions: reveal on hover

  • Hover-reveal pattern solves visual noise and diagnostic bias
  • Suggestion appears only on user intent (hover)
  • Enables immediate bulk classification without anchoring judgment

Tailoring the view

  • Two modes: continuous stream (speed) or row-based breakdown (structure)
  • Clean hover-reveal pattern maintained in both modes
  • Introduced based on user feedback
Continuous stream view
Row-based breakdown view

Bulk actions & editing

  • "Select All" triggers the native Floating Toolbar
  • Refine selection and perform advanced bulk actions
  • Stays within the existing workflow, with no context switch

Success metrics & behavioral analysis

Time-to-report reduction

Metric: Avg. time to clear the "Unclassified" bucket, Bulk vs. single-cell flow

Goal: ~20% target reduction in manual classification time per BMA report

Layout usage & adoption

Metric: 68% adoption rate for the "Split View" layout

Goal: Validate the necessity of customizable workspaces for different user personas

Selection-to-action conversion

Metric: 80% of "Select All" clicks leading to a successful Toolbar action

Goal: Verify the "Global Selection" pattern is intuitive and leads to task completion

Next case studies

Scopio
Safe by default
fear → trust
Replaced open CLI access with a governed, safe-by-default interface.
The Vets
Built for adoption
98% team adoption
Veterinary documentation, rebuilt around how users actually work.

Thanks for reading.

If you have any questions, feedback or would like to discuss my work further, feel free to get in touch.