Designing an AI Assistant for high-stress work environments.
Role
Product Design & Research Lead
Tools/Skills
Figma, Miro, Qualitative Research, Workshop Facilitation
Timeline
January 2026, 4 Weeks
Impact
Stakeholder Alignment, Optimized Product Roadmap
Overview
Aligning Ford's Service Advisor AI Strategy
Ford was building a generative AI assistant for dealership service advisors, but leadership feedback was clear: the prototype felt too much like "smart search" rather than an assistant. I was brought in to provide a design perspective to facilitate cross-functional alignment, define what constitutes a meaningful AI "lift," and translate field research into actionable product features that respect domain boundaries.
CHALLENGE
Beyond "Smart Search"
The initial AI prototype was functioning as little more than "smart search", it could answer questions, but leadership needed the team to push beyond basic lookup functionality and deliver genuine strategic value to service advisors.
Service advisors operate in high-pressure, noisy environments where speed is critical. The challenge was reconciling disparate data sources like vehicle telematics, historical service records, and real-time customer sentiment, into a single, trustworthy interface.
Research
Strategic Facilitation First
Rather than jumping to design, I facilitated a workshop to define AI-human collaboration boundaries, and develop use-cases for prototype testing with end users. As part of this work, I developed an AI Intervention Framework that bucketed the levels of autonomy we would allow the AI to have.
1
OBSERVE
Passive Monitoring
• AI watches patterns
• Surfaces insights
• No actions taken
• Human maintains full control
Example: Fleet trend analysis
2
SUGGEST
Active Recommendation
• AI recommends actions
• Provides reasoning
• Advisor decides
• Explainability required
Example: Warranty check alerts
3
AUTOMATE
Conditional Action
• AI takes action
• Within guardrails
• Human can override
• Transparency maintained
Example: Daily morning report
With these boundaries defined, I could prioritize addressing leadership feedback of moving past the 'smart-search' limitations, defining what constitutes a "Lift" for our AI assistant.
Defining a "Lift"
As a baseline, I added a requirement that in order to qualify as a genuine lift, the AI had to tap into at least two separate systems. Otherwise, the service advisor could simply complete the task themselves.
Example: Generating a repair order requires accessing parts inventory + pricing + tax calculation. In reality, most lifts integrated 3-5+ systems, but the "two-system minimum" became our bar for value.
I developed the Trigger-Lift-Output template to align the workshop group and give everyone a consistent framing for their AI examples, that the group developed from their in-person visits to dealerships.
Trigger → Lift → Output Template Examples
Trigger
Similar vehicles serviced
Lift
Identify common patterns
Output
Fleet insight dashboard
Level 1: Observe
Fleet Learning Feature
Trigger
Customer describes symptom
Lift
Suggest follow-up questions
Output
Structured symptom report
Level 2: Suggest
Symptom Capture Feature
Trigger
Warranty expiring within 7 days
Lift
Auto-flag on morning report
Output
Priority alert to advisor
Level 3: Automate
Daily Report Feature
Respecting Domain Boundaries
User research revealed a hard line: advisors wanted AI to handle administrative work, not vehicle diagnostics. They rejected AI-led technical analysis, reinforcing that the value is in workflow automation, not encroaching on the technician's expertise.
AI Should Support
Administrative tasks: Report generation, data entry, scheduling
Customer-facing lifts: Warranty checks, pricing lookups, history retrieval
Pattern detection: Fleet insights, maintenance trends, upsell opportunities
Workflow optimization: Appointment prioritization, time management
AI Should NOT Override
Vehicle diagnostics: Issue analysis belongs to technicians, not AI
Final pricing decisions: Advisors must maintain authority with customers
Customer relationship: Human judgment required for sensitive conversations
Technical expertise: Respect professional domain boundaries
IMPACT
Key Strategic Outcomes
This project redefined our approach to an AI Assistant. It proved that the role of AI in a professional environment wasn't just about efficiency, it's about Domain Respect.
By establishing where the AI's knowledge ends and the Human's expertise begins, we built a tool that advisors actually want to use. The success came from deeply understanding the cultural dynamics of the service bay and respecting the professional identity of service advisors.
Want to dig into the research, the stakeholder dynamics, or the strategic rationale?
I'd love to walk you through something more detailed. Reach out at hello@alexander-christian.com or connect on LinkedIn.