Unifying a fragmented customer service platform to empower global support agents with modern, efficient tooling that eliminated legacy system bottlenecks.
Client
Autodesk
My Role
Lead UX Consultant
Domain
Enterprise SaaS · Customer Service
Year
2023
Tools
Figma · FigJam · Dovetail · Miro
Overview
Autodesk's global customer support organization employs thousands of agents who handle technical support, licensing, billing, and enterprise account management. These agents are the frontline of Autodesk's customer experience — and they were struggling.
The root cause wasn't agent skill or motivation. It was the tooling. A collection of aging, siloed systems had accumulated over years of acquisitions and incremental product additions. Agents routinely toggled between multiple applications to resolve a single customer issue, manually copying information between systems and losing precious minutes — and customer trust — at every handoff.
As Lead UX Consultant, I was embedded with the Autodesk team to lead research, synthesis, and design for a unified agent experience platform.
The Challenge
The existing agent desktop was not a unified platform — it was a loosely stitched collection of browser tabs, internal wikis, and custom tools that agents had learned to navigate through tribal knowledge and personal workarounds.
Fragmented tooling — Agents averaged 7+ different applications per customer interaction. The cognitive overhead of context-switching was measurable in both handle time and error rate.
Broken knowledge access — Support articles, product documentation, and escalation procedures lived in separate systems with no unified search. Agents often resolved issues from memory because finding official guidance took too long.
Invisible customer context — Agents frequently began interactions without visibility into the customer's product history, previous cases, or account health — forcing customers to repeat information they'd already shared.
No performance feedback loop — Agents received CSAT scores but no actionable signals during interactions. High performers had developed personal heuristics that were never codified or shared.
Research
I designed a research program anchored in observational methods — the only reliable way to understand what agents actually do versus what they report doing. I collaborated with the Autodesk team to align on research goals, recruit participants across tiers and geographies, and coordinate in-person and remote sessions.
In-person observation of agents across 4 global locations (US, India, Ireland, Philippines) — watching real customer interactions unfold in real time.
24 in-depth interviews with agents segmented by tier (T1/T2/T3), tenure (new vs. veteran), and specialization (technical, billing, enterprise).
Mapped the full agent journey from case intake to resolution across all channels — phone, chat, email — to identify systemic failure points and duplication.
Facilitated synthesis workshops with the Autodesk design and product team, clustering 400+ data points into themes, patterns, and opportunity areas.
Analyzed 6 months of operational metrics — average handle time, first-contact resolution, CSAT by tier — to quantify the cost of the current experience.
Built 4 agent personas (New Hire Nav, Veteran Specialist, Escalation Lead, Enterprise Account Manager) to anchor design decisions throughout the project.
Critical Finding
Veteran agents had developed elaborate personal systems — physical notes, browser bookmarks, copy-paste scripts — to compensate for tooling gaps. This "shadow infrastructure" was invisible to management and impossible to onboard new agents to.
Opportunity
If the platform could codify the best practices of top-performing agents into the interface itself — smart defaults, contextual guidance, unified customer history — it could compress the experience gap between new and veteran agents significantly.
Design Process
Co-creation was central to this project. I facilitated weekly design workshops with Autodesk product managers, engineers, and — critically — agents themselves. Involving agents in ideation, not just validation, produced designs that reflected how people actually thought about their work.
Collaborative sessions with Autodesk to align on findings, prioritize problems, and establish design principles from evidence.
Sketching sprints and low-fidelity wireframes reviewed with agents to validate conceptual direction before investing in high-fidelity work.
Three rounds of interactive Figma prototypes, each validated in moderated testing sessions with agents from different tiers and locations.
Detailed design specifications, component documentation, and design rationale presented to engineering and QA for phased implementation.
Core Solutions
Customer 360 Panel — A single-pane view of every relevant customer data point: product licenses, case history, account health, recent interactions, and account tier — loaded automatically when a case opens, with zero manual lookup required.
Unified Knowledge Base — Integrated search across all support documentation, internal wikis, and escalation playbooks — surfacing the most relevant articles contextually based on case category and customer profile.
Guided Resolution Flows — For the 20 highest-volume issue types, step-by-step resolution flows that encoded veteran agent expertise — reducing reliance on tribal knowledge and dramatically improving new agent performance.
Real-Time Sentiment Indicators — Subtle conversation health signals (escalation risk, customer frustration cues) surfaced within the case view to help agents adjust their approach before a situation deteriorated.
Impact
↓ 38%
Reduction in average handle time within the first quarter of the unified platform rollout
↑ 22%
Improvement in first-contact resolution rate — customers getting answers without escalation or callback
↓ 60%
Reduction in new agent onboarding time to proficiency — from 8 weeks to approximately 3 weeks
Key Learnings
Shadow infrastructure is a signal, not a problem. When employees create workarounds, they're telling you exactly what the official system fails to provide. Shadow sessions revealed a design brief that no stakeholder meeting ever could have produced.
Codifying expertise is a design challenge. The gap between new and veteran agents was primarily a knowledge access problem, not a skill problem. Embedding expert heuristics directly into the interface democratized excellence in a way that training alone never would.
Global research changes everything. The pain points in Autodesk's India support center were meaningfully different from those in the US. Multi-location research wasn't a nice-to-have — it was the only way to design something that worked globally rather than just in the home market.