IBM: Watsonx for Support
Designing an AI-native platform to eliminate manual case work for IT support engineers and clients
To honor my NDA, I’ve kept this case study free of confidential info. If you would like to know more, I'm happy to discuss more details! Just shoot me a message.
COMPANY & ROLE
IBM, Design Lead
TEAM
IBM CIO (Chief Information Officer Organization)
FOCUS
0 to 1, Design strategy, AI Product design, Service design
TIMELINE
Fall 2023 - Spring 2025
17,000+
Hours Saved Annually, driven by automated responses for low-complexity client cases
31 min
Saved Per Case, achieved through automated log analysis and anomaly detection
56,000+
Hours Saved Monthly, via automated workflow triggers across piloted business units
01. The Core Challenge & System Constraints
IBM Support handles millions of complex enterprise cases annually. Engineers were losing hours to fragmented tools, manually reading through thousands of log lines, and drafting repetitive responses.
How might we integrate AI into existing IT support workflows without breaking engineer & customer trust or causing context-switching friction?
Note on NDA: Recreated designs use public design systems (IBM Carbon / Salesforce Lightning) and public metrics.
02. My Role and Process
Design Lead: Owned discovery through prototyping end to end, from the first workshop to the assistant that shipped.
Discovery: Co-led two structured workshops with support, product, and engineering stakeholders; 1.) mapping where AI could realistically internally & externally help 2.) turning that into a shipped direction and roadmap.
Prototyping: Built and tested the assistant directly with platform teams to validate technical feasibility before handoff to engineering.
03. Architecture Trade-offs: Choosing the Right Approach
Before designing screens, we evaluated three structural ways AI could live within the workflow:
Option A: Automated Contextual Overlays
Required real-time user-action tracking beyond our initial technical capabilities.
Option B: Browser Extension
Failed to maintain cross-session context between client and backend support tools.
Option C: Embedded Assistant
Provided the flexibility needed to maintain state across platforms while integrating directly into existing UI paradigms.
04. Key Solutions & UX Highlights
Smart Case Summaries & Confidence Scoring
Log reviews often slow support engineers down. This concept integrates AI-powered analysis to identify trends, surface relevant past cases, and suggest next steps within the same screen.
Built-In Log Analysis
Log reviews often slow support engineers down. This concept integrates AI-powered analysis to identify trends, surface relevant past cases, and suggest next steps within the same screen (saving 31 minutes per case).
Assistive Response Drafting
From client responses to internal wrap-ups, I explored how support engineer could use context-aware prompt triggers that help engineers generate client updates and internal notes, moving them from "blank page" to "editor/reviewer."
Dual Experience
I prototyped two watsonx assistants: an internal one that lets engineers order parts, schedule support, and escalate cases without leaving the platform, and a client one that answers questions, surfaces updates, and walks clients through resolution. Note that they are on different platforms, so different design systems: Salesforce Lightning for engineers, IBM Carbon for clients.
Designing for Implementation: Turning Vision into Scalable Systems
With foundational concepts and interface prototypes in place, the next step was translating design intent into a roadmap that could work across IBM’s layered platforms and global teams. Translating high-level AI vision into shipped software required balancing two distinct design systems (Salesforce Lightning for internal engineers, IBM Carbon for clients) powered by a shared watsonx back-end.
How we grounded design in implementation:
- Unified Interaction Guidelines: Defined standardized UI patterns for AI confidence, prompt triggers, and fallback states across both Carbon and Lightning environments.
- Feasibility-First Architecture: Partnered directly with platform engineers to define data requirements and conversational UX models before writing production code.
- Strategic Phasing: Prioritized high-impact engineer capabilities for the first pilot run (case summaries and response drafting) to build internal momentum and before expanding to deep log analysis and client-facing tools.
05. Impact
The experience patterns and assistant architecture I led drove a systemic shift across IBM’s Cognitive Support Platform (powered by watsonx.ai and watsonx Orchestrate).
By moving engineers away from manual log hunting, our integrated log analysis slashed resolution time by 31 minutes per case. At scale, offloading low-complexity cases to AI-generated drafts saved 17,000+ hours annually for clients, while automated workflow triggers eliminated over 56,000 hours of repetitive manual tasks every month.
What started as a fragmented, reactive support ecosystem is now an intelligent, proactive platform that continues to scale across IBM business units.
07. Testimonial
Some very kind words I recieved from a Lead Service Designer I worked closely with, especially during the design workshops.
Thank you, Kate!
