Volvo Group • Gen AI • 2024
Volvo Group’s fleet management platform
Timeline
3 months
Role
Product Designer (Me!)
Team
Business Development
Product Manager
Engineering
TL;DR
Volvo Group serves 250,000 connected vehicles across 100,000 fleets. I shaped the vision for integrating Generative AI into Volvo Group’s fleet management solution as part of the MS-HCI program’s industry capstone at UC Santa Cruz, showcasing Volvo Group’s digital innovation.
Partnering with business development, product management, and engineering, I ensured alignment with business goals, user needs, and technical constraints. Through this collaboration, we delivered AI-powered tools in six months, improving discoverability and actionability of operational insights and reducing cognitive load for fleet managers.
↓ 90%
Time on task
↑ ~7500
MAU (projected)
↓ 35%
Support calls (projected)
Solution
Here are some of the key flows
Intelligent Alerts
This feature reduces alert fatigue by only highlighting critical fault codes while also supporting subsequent triage.
Advanced Analysis
This feature delivers advanced telematics insights without the need for analytical expertise or additional time.
Specific Data Inquiry
This feature grounds every response in verified asset data, eliminating hallucinations and mix-ups. (Issue discovered during testing)
Overview
Volvo Group is a global manufacturer of trucks, buses, construction equipment, and marine and industrial engines.
Volvo Group is a global manufacturer of trucks, buses, construction equipment, and marine/industrial engines. Founded in 1927, headquartered in Gothenburg, Sweden. The company employs over 100,000 people across 180 markets worldwide.
NASDAQ:VOLV B
Problem
Volvo Connect struggles to convert onboarded customers into active users.
Along with the purchase of a Volvo Truck, customers are provided with two years of complimentary access to Volvo Connect, Volvo Group’s fleet management solution, which they’re then onboarded to at the point of sale. Of the 118k fleet managers onboarded thus far, only about 9% use the platform regularly.
User Research
User interviews uncovered difficulty deriving operational insights from available fleet telematics data.
My first instinct was to understand why users weren’t taking advantage of it, since secondary research had shown telematics to be a significant competitive advantage. Some disengagement was to be expected, but the volume suggested a deeper problem in the product.
Through user interviews, we found that users recognise the value of fleet telematics data, but felt that the time and effort required to derive actionable operational insights using the platform’s current methods far outweigh the benefits it offers. The most recurring contributing factors were as follows.
Data w/o operational insight
Volvo Connect provides a wealth of telematics data but stops short of insight, leaving little of it readily actionable.
Alert fatigue
Faults arrive by the hundreds with limited triage, making it challenging to distinguish critical faults from routine ones.
Fragmented ecosystem
The trucking industry is historically siloed, forcing fleet managers to navigate multiple applications before building a complete picture.
Partial fleet visibility
Volvo Connect is single-OEM, so users running mixed fleets turn to brand agnostic third-party tools for a full fleet view.
User workarounds are driving up support costs and driving down engagement.
While most users disengage from the platform entirely, the small minority who stay active export the data to work with it elsewhere — taking their activity outside the product.
Larger fleets skip interfacing with the platform altogether, requesting dedicated Volvo representatives to support their telematics analysis directly — an expensive workaround that’s challenging to scale.
How Might We
Help users derive operational insights with minimal effort?
User goal
Drive platform adoption?
Business goal
Ideation
User research and business strategy converged on the same idea, which was a conversational assistant.
I facilitated a workshop with business development, product management, and engineering, sketching against both HMWs. We generated a wide set of ideas and discussed each against user value, business value, and feasibility.
Several ideas satisfied the brief, but it was 2024, and Volvo Group had a vested interest in showcasing their digital innovation — Generative AI was the obvious vehicle for that, and a conversational assistant was what both could agree on.
Design Process
I started by taking a capability inventory.
Before designing anything, I sat down with engineering to understand what the product could reach — what data it could access, what APIs it could call, and what actions it was permitted to take.
We were limited to what already existed within the product, since the other service systems were heavily siloed. However, the backend was built atop Amazon Quick, which had agentic capabilities baked in.
AI-Powered Business Intelligence
To meet tight deadlines, I prioritized rapid iteration — building directly in high-fidelity by leveraging Volvo's experience system.
Reusing established components and patterns let me design in high-fidelity from the first iteration, putting a version of the final product in front of stakeholders and users within weeks. I also contributed back to the design system, extending the library for future work.
Design Decisions
Finding areas of improvement in the current experience through iteration and testing.
I walked through the current experience while comparing it against all the insights we learnt during user research. When a need felt unmet, I wrote down what was missing. Those notes became the focus of future iterations, some of the results are shown below.
BEFORE
No way to export data or verify logic.
AFTER
Export csv data for each query and review SQL code to verify logic.
BEFORE
No way to verify information provided.
AFTER
See sources that Fleet Assist relied on to arrive at the answer provided.
Outcome
We project that the number of fleet managers leveraging telematics data will go up in the months ahead — especially among small fleets.
↓ 90%
Time on task
↑ ~7,500
MAU (Projected)
↓ 35%
Support calls (Projected)
Reflection
My key takeaways and learnings!
Start recruiting participants early!
Especially in niche domain, recruiting representative users can be challenging so start early and reach out to company contacts.
Iteration! Iteration! Testing!
While data and research showed us exactly what to fix, we wouldn't have arrived at the right solutions without feedback from users.
Curious to know more?
This is just a small part of the design process — for the full story, get in touch