Koch Industries Atonix Digital Product Design AI / ML Enterprise

Asset Performance
Management

Helping industrial operations surface critical issues, identify root causes, and act with confidence without requiring deep expertise in every system they manage.

Role
Product Design, End-to-end
Scope
Detection, Diagnosis, Action
Team
Cross-functional
DesignOI Asset Performance Management

Industrial operations are growing more complex. The expertise needed to run them is not keeping pace.

Mid-sized industrial facilities operate some of the most complex physical systems in existence. As infrastructure ages and experienced workers retire, the institutional knowledge needed to interpret that complexity is disappearing.

Atonix Digital serves large-scale industrial operators, Koch Industries, JPMorgan Chase, and others, who manage complex assets across multiple systems. The work was to design a decision-support platform that surfaces what matters, explains why it matters, and recommends what to do next. The goal: faster, more confident decisions without requiring deep expertise across every system a team oversees.

The design challenge was not to expose more data. Industrial facilities already have more data than any team can meaningfully process. The challenge was to transform continuous sensor data and machine learning outputs into clear, actionable guidance, at the moment operators need it.

100s
Of sensors generating continuous data across a single facility
4
Core views: Evaluate, Analyze, Monitor, Resolve
E2E
Ownership from system model through interaction design

Too many alerts.
Too little clarity.
No consistent path forward.

Operations teams are responsible for monitoring environments of extraordinary complexity with limited time, limited staff, and increasingly limited institutional knowledge.

The problem space

The conditions that create this problem compound each other. Hundreds of sensors generate continuous streams of data. Teams have limited capacity to investigate every signal. Critical knowledge is held by experienced operators who are retiring faster than organizations can replace them. When something goes wrong, operators are left with alerts they cannot fully interpret, and no reliable system for determining what to do next.

"Critical issues were often identified too late, or not fully understood until the window to act had already closed."
Operations context, mid-sized industrial facility
The core insight
The goal is not to expose more data. It is to surface what matters, provide context, and guide users toward the right action. Turn data into guidance.

Design a system that works the way an expert operator thinks.

An expert operator does not read every sensor. They know which signals matter, how to interpret combinations of signals in context, and what to do when something looks wrong. The opportunity was to encode that expertise into the product itself.

Rather than presenting operators with more information to process, the system continuously monitors in the background, surfaces only what requires attention, provides clear context around why something was flagged, and recommends a course of action. The design model follows the same decision sequence an experienced operator uses: awareness, understanding, diagnosis, action.

Alerts ideation

Early ideation around how to surface and prioritize alerts without overwhelming operators already managing high cognitive loads.

The interaction model
Instead of navigating disconnected tools, users follow a single natural flow: an issue is surfaced, the system provides context, contributing factors are highlighted, and a course of action is recommended. The system guides users from awareness to action.

Four views. One continuous flow from signal to resolution.

Each view serves a distinct moment in the operator's decision process. Together they create an unbroken path from detecting a problem to resolving it.

Evaluate view
01 · Evaluate
Rapid orientation
High-level system health across the site. Meaningful signals surfaced, areas of concern highlighted, and the most relevant starting points for investigation made immediately clear. Operators know where to focus before they take a single action.
Analyze view
02 · Analyze
Deep investigation
Detailed insights into system behavior and root causes. The underlying factors contributing to an issue are surfaced in context, supporting informed decision-making without requiring operators to hold complex system knowledge in their heads.
Monitor view
03 · Monitor
Continuous oversight
Real-time updates on system performance and emerging issues. Operators maintain situational awareness across the full facility and can respond promptly to changes without actively watching every sensor feed.
Resolve view
04 · Resolve
Effective action
Actionable insights and recommended steps to address identified problems. Operators implement solutions with confidence, with clear guidance on what to do and assurance that issues are being resolved rather than masked.

The hardest problems were not visual. They were structural.

Every significant challenge in this work came back to the same root tension: operators need guidance, but they also need to trust the system giving it. Those two requirements pull in opposite directions if the design is not deliberate about how it resolves them.

01
Signal vs noise
Reduce alert fatigue through prioritization, not suppression
The existing state was alert overload. Every signal surfaced with equal urgency, which meant no signal could be trusted as urgent. The design response was not to show fewer alerts, but to establish a clear prioritization model so operators could immediately understand which signals required attention now, which could wait, and which were informational. Grouping related alerts reduced fragmentation without hiding information.
02
Scaling expertise
Embed institutional knowledge into the system itself
Experienced operators carry decades of pattern recognition that cannot be documented in a manual. The design approach was to surface that knowledge at the moment of need: contextual explanations of why a combination of signals is concerning, historical comparisons that give operators a reference point, and recommended actions drawn from how expert operators have responded to similar conditions in the past.
03
Clarity in complexity
Simplify multi-variable environments without losing fidelity
Industrial systems involve dozens of interdependent variables. Simplifying that complexity for an operator means making the relationships legible, not hiding them. The design used progressive disclosure to surface the right level of detail at each stage of investigation, giving operators a clear view without overwhelming them before they have established context.
04
Trust and transparency
Make the reasoning visible, not just the recommendation
Operators who do not understand why a system flagged an issue will not act on its recommendation. Every surfaced issue includes the contributing factors, the severity rationale, and the basis for the recommended action. This is not just good UX practice in industrial settings, acting on an unexplained recommendation can create safety risk. Transparency is a safety feature.
05
Accessibility
High-density interfaces built for expert, time-pressured users
Operations environments are not typical software contexts. Operators may be scanning dashboards from a distance, under time pressure, or managing several systems simultaneously. Every interface decision was evaluated against those conditions: contrast ratios at distance, scan paths for time-constrained users, and interaction patterns that do not require fine motor precision under stress.

What the work changed.

The measure of a decision-support system is not whether it surfaces more information. It is whether operators make better decisions, faster, with more confidence.

User impact
Faster identification of critical issues before they escalate
Reduced reliance on individual expertise for routine diagnosis
Increased operator confidence in decision-making under pressure
Consistent decision quality across varying experience levels
Product impact
Established a scalable decision-support model applicable across facility types
Enabled consistent workflows across diverse operational environments
Created a reusable design system for high-density industrial interfaces
Reduced design and engineering overhead for future feature expansion
Business impact
Increased accessibility for mid-sized facilities without large expert teams
Helped mitigate operational risk from workforce and knowledge gaps
Positioned the product as a long-term operational intelligence platform
Reduced time-to-value for new facility onboarding
As systems grow more complex and expertise becomes harder to retain, the role of design shifts. From presenting information to preserving knowledge and guiding action.
Key takeaway

What industrial design taught me about trust.

Consumer software earns trust by being useful. Industrial software earns trust by being right. The stakes of an operator acting on a bad recommendation are fundamentally different from the stakes of a consumer ignoring a notification. That difference shaped every design decision in this work.

What worked well

Designing the four-view flow as a single continuous decision process rather than four separate tools gave operators a coherent mental model from day one.
Making reasoning visible alongside recommendations built operator trust faster than any other design decision in the system.
Progressive disclosure at the Analyze stage let expert users go deep without overwhelming operators who needed only the top-level diagnosis.
Embedding accessibility standards from the start produced a better interface for all users, not just those with specific access needs.

What's next

Predictive maintenance integration, surfacing degradation patterns before they become issues rather than after the first alert fires.
Collaborative resolution workflows so multiple operators can coordinate on a single issue without duplicating effort or creating conflicting actions.
Mobile-first monitoring for operators who are physically moving through a facility and need situational awareness without returning to a fixed terminal.
Feedback loops that let operators confirm or correct system recommendations, improving the underlying ML models over time through use.
Let's talk about your
next big idea.

I'm currently open to new design leadership roles. Whether you're building something ambitious or improving what you already have, I'd love to connect.

Get in touch