AI Experience Design · B2B SaaS · Industrial Support
Not a ticket problem.
A context problem.
A B2B SaaS startup wanted to add AI to their support ticket platform for industrial machinery manufacturers. They thought they needed a chatbot. What they actually needed was a completely different approach — and a system designed for both sides of the ticket at once.
Client identity and specific product details are withheld under NDA. Metrics shown are projected based on industry benchmarks.
Most support AI is designed to deflect tickets. That doesn't work in industrial B2B — operators arrive at the portal with a broken machine, not a simple question. The real bottleneck wasn't ticket volume. It was what happened inside each ticket: incomplete descriptions, back-and-forth rounds, technicians starting from zero every time. The AI isn't here to avoid the ticket. It's here to make every ticket worth opening — from the first second, for both the operator who opens it and the technician who receives it.
System architecture
Two agent systems · one ticket in the middle
Operator
plant technician
Client system
Intake side
what the operator experiences
Enriched ticket
Technician
manufacturer's support agent
Technical system
Resolution side
what the technician experiences
Operator
plant technician
Client system
Intake side
what the operator experiences
Enriched ticket
Technician
manufacturer's support agent
Technical system
Resolution side
what the technician experiences
Shared context layer
The operator sees one interface · the technician sees another · the agents work in parallel. Context passes as a structured object — not free text.
Key results
−50%
Time to resolution
Tickets arrive with 75–88% of context pre-filled. Technicians act immediately instead of asking.
+35%
Technician capacity
Same team handles more cases — without adding headcount.
−60%
Portal abandonment
Operators under pressure stay in the system because the assistant guides them, not filters them.
These are projected metrics based on industry benchmarks and the design specifications delivered during the engagement.
What makes this different
Designed for both sides simultaneously
Most AI support tools design for the user or the agent. This system designs for both — and for the handoff between them. That's the discipline I call HAXD: Human-Agent Experience Design.
The assistant doesn't replace the ticket
It builds it. Every conversation the operator has with the assistant automatically structures the ticket — context, urgency, error codes, previous attempts. No forms. No repetition.
The technician never starts from zero
By the time a ticket reaches the technician, the AI has already crossed it with machine history, similar resolved cases, and a suggested diagnosis with an explicit confidence level.
Honesty as a design principle
Every AI suggestion comes with a confidence score and its source. The system knows when to step back and escalate — and it says so clearly. That's what builds trust in high-stakes industrial environments.
Explore the case study
The client agent
How I designed the experience for operators arriving under pressure — and the three-agent system that guides them without feeling like a filter.
Read more →The technical agent
How I designed the experience for technicians receiving tickets — diagnosis suggestions, confidence levels, escalation flows, and schedule management.
Read more →System design
The full architecture — context design, shared object, agent IDs, guardrails, evals, and monitoring. How the two systems connect and improve over time.
Read more →About this work
This case study documents a real engagement with a B2B SaaS startup building support infrastructure for industrial machinery manufacturers. Client identity and specific product details are withheld under NDA. The design decisions, prototypes, agent architecture, and strategic framework reflect the actual work developed during the engagement.
If you're building something similar — or want to talk through your AI product decisions — I'd like to hear about it.