Supporting case study
From thirteen systems to one service task
United Airlines
At IBM iX, I helped consolidate thirteen customer-care systems into one task-centered dashboard. During the pilot, average throughput rose from two to seven complaint resolutions per hour.
- Role
- UX Designer
Facts
- Client
- United Airlines
- Employer
- IBM iX
- Dates
- February to December 2018
- Research
- Interviews and ethnographic observation in Chicago and Houston
- Scope
- Thirteen legacy systems used in customer care work
- Sequence
- Customer care dashboard, followed by a rebooking assignment
Watch the complaint come together
Reconstruct the case before redesigning it
Customer care representatives handled one complaint by piecing together information from thirteen legacy systems. I interviewed and observed representatives in Chicago and Houston, then helped turn those sessions into a current-state assessment. Customer history, itinerary details, complaint status, compensation, notes, and correspondence were split across the work. Agents had to rebuild the case before they could resolve it.
I documented dashboard component behavior in Flinto and worked with United and IBM partners to translate the research into a new information architecture. An agent needed to understand the customer, the trip, the complaint, and the available action without reconstructing that picture in separate tools.
Contribution
My work covered field research, the current-state assessment, information architecture, interaction design, and component behavior.
From scattered context to a working case
See the working state
Make the complaint the organizing unit
I organized the workspace around the complaint an agent needed to resolve rather than the legacy systems that stored each piece of information.
Make the actions mean what they say
The April 2, 2018 testing notes show why consolidating information was only part of the job. A participant could see the complaint and the relevant customer details, but still had trouble understanding the scope of an action.
When responding to a complaint, the participant was unsure whether Edit response also changed the compensation amount. When asked to add a note, they looked to editing and could not identify the nested note action. In the history views, they confused the current case with related cases and expected related cases to share an itinerary.
These observations put a sharper question behind my information architecture work: can an agent tell what an action changes, and which case it applies to? Keeping context visible matters, but the labels and placement of the controls have to explain that distinction too.
Scope
The testing helped the team distinguish an action's meaning from its location. The later pilot evaluated the consolidated workflow.
Organize the screen around the task
Keep context visible and let supporting detail wait
Putting every field on one page would have preserved the same problem in a denser form. I kept the customer and itinerary visible, then gave the active case, its status, and the next action priority. Compensation, coding, notes, attachments, and correspondence opened as supporting detail instead of competing for the first glance.
The placement rule was simple. Put information where the agent needs it in the task, not where a legacy system happens to store it. That rule shaped the cards, their expanded states, and the reading order across the workspace.
Watch the supporting panels change
Preserve the itinerary reading order
Place information at its point of use
Customer and itinerary context stayed visible. Supporting controls opened where the complaint required them, giving product, design, and engineering partners one rule for reviewing each screen.
Measure completed work
Count resolved complaints
The pilot measured completed complaint resolutions per hour, not preference for a mockup or a forecast about future use. Average throughput moved from two to seven complaint resolutions per hour with the consolidated dashboard in the pilot.
Keep the result at the workflow level
I report the pilot result for the consolidated workflow rather than assigning it to an individual card or interaction.
Carry the rule into rebooking
Change the controls, not the organizing rule
After the dashboard pilot, United gave IBM a follow-on rebooking assignment. The workflow asked a different question. Which itinerary should the agent change, and what option could replace it?
I helped the IBM team keep passenger and trip context together and design controls for comparing flights and adjusting the itinerary. The task-centered rule carried into a related service workflow without forcing the complaint dashboard into a different job.
Follow the rebooking flow
Reuse the information model, not the dashboard
The rebooking work kept context and actions together while using controls suited to flight comparison and itinerary changes.
Results
- During the pilot, average complaint throughput rose from two to seven complaint resolutions per hour.
- United then gave IBM a follow-on assignment to apply a similar approach to rebooking.
Evidence limit
Surviving sources do not establish long-term production performance or feature-level causation.
Related work
A research-led direction for ANA's global website
In 2019, I assessed earlier agency work, interviewed ANA stakeholders, and helped facilitate a two-day design-thinking workshop in Tokyo. I turned the findings into UX 2020, design principles and a roadmap for western business travelers.
- UX research
- Design strategy
- Workshop facilitation
- Concept design
A coffee order starts with the right site
I worked with BP product owners in the UK and Australia on coffee pre-ordering for BPme. I refined the requirements through research, then created wireframes and a clickable prototype that carried the order through to collection.
- Interaction design
- Mobile commerce
- Prototyping
- Service design
Designing and building AI workflows at AstraZeneca
At AstraZeneca, I designed compliance tools, forecasting interfaces, and workflows for a shared AI workbench. I wrote requirements, built frontend prototypes with AI assistance, and developed components and guidance for engineers building their own screens.
- AI product design
- Design systems
- Frontend prototyping