Certified User Experience Design leader
Designing an Agentic AI for Asset Health
How can AI transform reactive maintenance into proactive, When the users are not ready?
Context & Goal
While Asset owners and Reliability Engineers are Maintaining asset they face data fragmentation, unclear priorities, and underused capabilities, leading to Reactive & Risk Based Maintenance.
Problem Statement
In Present APM tool, Asset Managers struggled with:
-
Fragmented data views across multiple systems, slowing decision-making.
-
Unclear priorities, with no quick way to identify assets needing urgent attention.
-
Manual analysis required to extract actionable insights from complex datasets.
-
Low trust in AI, limiting delegation due to opaque outputs and setup complexity.
They needed a system that delivered contextual insights, prioritized urgency, and built trust for gradual AI delegation.
My Role
Craft
Collaboration
-
Led cross-functional alignment
-
UX Designer
-
Data Science
-
Content Design
-
Product Management
Leadership
-
Facilitated design thinking workshop
-
Mentored UXR and Content design
-
MVP Alignment
Research and Ideation
Discovery Methods
Objective
-
To define value proposition
-
AI Knowledge of users
-
Assess cognitive load reduction in daily workflows.
-
Validate if the agent improved prioritization efficiency.
-
Test personalization for role-specific needs.
Findings
-
Before: 14/17 users felt “lost” post-login, overwhelmed by static dashboards.
-
After: 12/17 reported the agent as a “time saver,” with 80% success in NLQ interactions.
-
Users valued proactive suggestions but wanted clearer explainability for complex alerts.
Impact
Quantitative Wins:
-
Reduced task prioritization time by ~25% in pilot testing (based on workflow audits).
-
Increased user adoption from 20% to 65% within 3 months of rollout.
-
Contributed to a 10% reduction in unplanned downtime in pilot sites.
Key Learning
- User Control is Paramount
Users need to feel in control. Overly directive AI leads to hesitation.
- Explainability Builds Trust
Users delegate when they understand the "why" behind AI suggestions.
- Smart Personalization (Not Bloat)
Universal personalization is unscalable.
Prototype
Note: Original IBM designs are under NDA and available upon request during portfolio reviews.
Anonymized Figma prototypes are linked above.