Menu
Book a free call CV
ProPlanAI AI-UX Audit & Redesign
Boosting User Trust and Transparency
ProPlanAI is Hexagon’s AI-powered automation tool that reduces machine tool CAM programming time by up to 75%, using AI models trained on historic CAM data. The data upload and training process is critical—manufacturers need a seamless, trustworthy experience to upload their proprietary data securely and train AI models effectively, ensuring personalized and accurate AI-driven programming outcomes.

The challenge
Improve user concerns about data privacy, transparency during the training process, control over training parameters, and trust in AI recommendations based on uploaded data.
Company
Hexagon

Product
Hexagon's ESPRIT EDGE - ProPlan AI Feature
Deliverable
  • Baseline and follow-up user surveys
  • AI-UX audit report using customized heuristics
  • Redesigned data upload and AI training UI flows
  • Compliance alignment documentation (EU AI Act)
Timeline
6 weeks (Discovery to Deployment)
My Role
Lead UX Designer
Team
AI Product Manager, Legal/Compliance Officer, Design System Architect, Software Engineer
MY RESPONSABILITIES

I led the AI-UX audit and redesign focused on the data upload and training flow.

MY ROLE

Validated improvements through quantitative user feedback demonstrating increased trust and satisfaction.

Applied a detailed AI-UX heuristic framework to evaluate and identify pain points across transparency, control, and error handling.

Delivered targeted UI/UX recommendations, ensuring clarity, compliance with EU AI Act data governance, and minimizing user cognitive load.

Designed and executed baseline and follow-up user surveys on perceptions related to transparency, privacy, control, and trust during data upload and training.

Impacts
  • User trust in data handling and training transparency increased by 32% (survey).
  • Training-related errors decreased by 28%, with clearer feedback and validation reducing user mistakes.
  • 45% increase in user engagement with training controls. Improved confidence in managing proprietary data securely and effectively.
  • UI now fully compliant with the EU AI Act, enhancing readiness for global markets.
//
//
//
CHALLENGES
Regulatory compliance with EU AI Act for data governance, transparency, and human oversight.
//
Complexity and cognitive load in understanding AI training’s impact on CAM automation.
//
Insufficient control mechanisms for users to pause, cancel, or customize AI training operations.
Lack of transparent, real-time feedback on AI model training progress and status.
User concerns around privacy and security of proprietary CAM data during upload

ORG GOALS

  • Deliver a world-leading AI-powered CAM programming tool that streamlines manufacturing processes.
  • Ensure Hexagon’s AI system upholds strict data privacy, regulatory compliance, and user trust standards.
  • Accelerate product adoption through superior user experience and confidence in AI automation.

  • Understand and track AI training progress in simple, transparent ways.
  • Have granular control over training parameters and the ability to intervene if needed.
  • Minimize cognitive effort to interpret AI feedback and system status.
  • Access clear help and compliance information to make informed decisions.

USER GOALS

DESIGN GOALS

  • Create an intuitive UI for secure, transparent data upload.
  • Embed real-time feedback and error mitigation in the user flow.
  • Enable user intervention.
  • Align interface and communication with the EU AI Act for transparency and human oversight.
  • Reduce cognitive load and streamline complex workflows without sacrificing control.

Version before redesign.

Discover Baseline Survey

Conducted baseline user surveys with Lyssna to assess clarity about data usage, AI training progress, and privacy policies.

Key results from the survey:
  • Significant portions of users felt uncertain about what happens with their data after upload.
  • Many users expressed difficulty tracking AI training progress or understanding its impact.
  • There was notable concern about data privacy and compliance visibility.
  • Users desired more granular control over AI training parameters and the ability to intervene.
Visual data and specific metrics (e.g., trust score, clarity ratings) were collected to pinpoint frustration points and highlight areas for improvement.
Survey Insights

Discover AI-UX Audit

Performed heuristic evaluation of the data upload and training UX, focused on these principles:
  • Keep Users Informed: evaluated how well the current UI communicates training progress and data use.
  • Allow User Intervention: checked the presence/absence of pause, cancel, and parameter adjustment controls.
  • Minimize Anticipatory Errors: assessed data validation mechanisms to prevent upload issues.
  • Reduce Cognitive Load: reviewed clarity, feedback mechanisms, and instructional cues for users.
  • Provide Accessible Help and Compliance Info: audited help availability, clarity of privacy info, and compliance statements.
Combined survey insights with usability evaluations to form a detailed problem report guiding the prototype and redesign phases.
AI Design Heuristic Checklist

Audit Higlights

  • No overall progress indicator (e.g., “Processing 2 of 5 features”)
  • No visual progress bar for the entire prediction process
  • No clear mapping between prediction and source file/feature
  • Lacks explanation on how predictions are derived
  • No info on prediction confidence or methodology
  • No detailed feedback options for specific issues
  • No confirmation feedback was received or acted upon
  • All features are shown simultaneously without prioritization
  • Lacks explanatory context for predictions
  • Poor visual hierarchy; all features look equally important regardless of status
Missing

  • Providing clear, real-time feedback on system status and progress,
  • Educating users about AI predictions and data privacy,
  • Offering actionable and specific error messaging,
  • Enabling flexible workflows and user overrides,
  • Displaying uncertainty and confidence to help users gauge prediction reliability,
  • Issuing proactive warnings about system limitations,
  • Disclosing how predictions are made to avoid “black box” effects,
  • Communicating system accuracy and limitations honestly, and
  • Clarifying responsibility when errors occur for better accountability.

Define Problem Framing
& UX Strategy

Analyzed survey and audit findings to synthesize core user pain points and regulatory gaps.

Established design goals balancing trust, compliance, usability, and cognitive load reduction.
PROBLEM STAMENT
Defined problem statements focusing on the lack of real-time feedback, inadequate user control, and limited transparency on data privacy.

Develop Prototyping
& Iterative Design

Ideated and designed solutions introducing:
  • Real-time training feedback panels with progress and ETA
  • Prominently accessible data privacy information and permission management
  • Controls for pausing/canceling training and adjusting model parameters
  • Step-by-step error handling with guided recovery instructions
  • Prototyped designs and iterated based on user and stakeholder feedback to optimize usability and ensure compliance with EU AI Act data and transparency requirements.
If a building becomes architecture, then it is art

Delivery Post-Redesign Survey & Impact Measurement

Following the implementation of the redesign, a second round of user surveys was conducted to measure changes in user perceptions and usability. The results demonstrated strong positive shifts:
  • User trust in data privacy and security increased by 35%, with 85% of respondents expressing confidence in how their proprietary data is handled compared to 50% before the redesign.
  • Clarity of AI training progress improved substantially; 78% of users felt well-informed about training status versus 42% in the baseline survey.
  • Control over training parameters was rated effective by 82% of respondents, reflecting the introduction of pause, cancel, and customization features.
  • Error handling satisfaction rose by 49%, with users appreciating clearer, actionable error messages.
  • Overall user satisfaction improved by 31%, highlighting a meaningful reduction in cognitive load and enhanced confidence using ProPlanAI.
//
//
KEY TAKEWAYS

Transparent communication of AI training status is vital for user trust in automated systems.

Iterative user feedback and data-driven auditing create meaningful UX improvements in AI systems.

//

Clear, accessible data privacy and compliance information are non-negotiable in AI data workflows.

//

Robust error prevention and recovery, coupled with user control, empower users and reduce frustration.

Made on
Tilda