01Context & responsibility
A small team was managing a growing service across time zones.
Secure Ground Transportation, or SGT, helps Amazon employees travelling on business to selected countries and cities with elevated risk. That risk can relate to disease outbreaks, severe weather, conflict, civil unrest or local safety conditions. The service covers the ground journey from the airport to the hotel, between the hotel and Amazon or business partner offices, and back to the airport.
Erika M, the Travel Risk Management (TRM) manager, approached our FinAuto UX team with a service that was becoming difficult to run through email. SGT had been agreed as a product within Trail, a suite intended to bring together employee travel and compliance services, including business travel and work or visit visas. I was assigned as the lead designer. Design and Engineering joined the work at the briefing stage.
TRM had 12 members working across different roles, time zones and levels of availability. As requests grew, the team was manually checking destinations, communicating decisions and coordinating with transport providers. Each request could generate another email conversation between TRM, the traveller and the provider.
The delay affected employees across levels, including VPs and senior leaders. Some needed to travel urgently, yet had to wait around 36 hours just to learn whether the request was approved or declined. Once they submitted the form, they could not change it themselves. Providers faced a different problem: cancelled requests could leave them with costs they had already committed to.
Traveller
Can I use SGT, and what happens next?
Understand coverage, receive a decision and manage an itinerary that may change.
TRM · 12 people
Which requests need our judgement?
Review the right cases and coordinate a growing service across time zones.
Transport provider
What has been arranged or changed?
Receive complete trip details and understand changes before committing resources.
My responsibility
I was the lead designer assigned from FinAuto. I led the UX work from discovery and journey mapping through wireframes, high-fidelity design, usability testing and handoff. Design and Engineering were involved from the initial brief, working closely with TRM as the product scope developed.
02Discovery & insight
The service was being mistaken for a cab booking tool.
Employees usually discovered SGT in the confirmation email after booking business travel through Concur. The link took them towards service information and a long intake form. Many assumed SGT was a cab service like Uber. They completed the intake even when their destination was outside the programme. The distinction between a risk-based travel service and everyday cab booking was not clear enough at the point of entry.
This mental-model mismatch was a major reason for declined requests. A traveller could spend time gathering information and completing the form, then wait for TRM to explain that the destination was not supported. The same misunderstanding created avoidable review work for a team with limited capacity.

Our customer journey exercise drew on five interview sessions. We traced the experience from discovering SGT to completing the trip. The findings included lengthy guidance, travel details scattered across different places, session expiry, long email threads and difficulty contacting drivers after arrival. We continued interviews and usability studies as the work moved through wireframes and high-fidelity designs.
Eligibility came too late.
The intake asked for effort before giving enough confidence about destination coverage. I moved country selection and guidance to the start of the request.
Submitting the form reduced the traveller’s control.
Changes required more coordination. I treated viewing, editing and cancelling as core tasks within the request lifecycle.
The handoff carried much of the friction.
Long email threads made arrangements difficult to follow. The design needed to give travellers clear next steps and providers a structured itinerary.
These findings changed how I approached the brief. Improving the form mattered, but the design also had to make the service understandable before submission and manageable afterwards.
03Customer journey
The journey continued well beyond the booking screen.
I mapped eight stages, starting with the Concur booking confirmation and ending with the final journey and payment arrangements. This kept the discovery work grounded in the traveller’s whole experience. The form was only one part of the journey. A person arriving in an unfamiliar location could face poor mobile connectivity, difficulty finding the driver or uncertainty about what transport had been arranged.
Arrange the trip.
Know what happens next.
Existing experience
Amazon employee travelling to a destination with elevated risk.
Goal: dependable ground transport throughout the trip.
Scroll across to follow all eight stages. Landscape gives the map more room on a phone.
| Journey stage | 01Discover SGT | 02Understand | 03Request | 04Coordinate | 05Meet the driver | 06First journey | 07During the stay | 08Complete |
|---|---|---|---|---|---|---|---|---|
| Touchpoint | Concur confirmation email | SGT wiki and guidance | Long intake form | TRM and provider email | Airport and phone contact | Airport to hotel or office | Hotel, offices and other stops | Final journey and payment |
| Traveller action | Book business travel and notice the SGT service link. | Read about the programme before requesting transport. | Collect trip details and submit a transport request. | Wait for a decision and agree transport arrangements. | Locate and contact the driver after landing. | Begin the first ground journey at the destination. | Use transport for the remaining business itinerary. | Finish the service and resolve payment arrangements. |
| Traveller concern | Understand whether SGT is relevant to this trip. | Find a clear answer about coverage and the next step. | Know the destination is supported before completing the intake. | Know who will respond and when the trip will be confirmed. | Reach the right person in an unfamiliar place. | Understand the agreed route and initial drop-off. | Know what is included throughout the stay. | Understand who handles payment and any outstanding charges. |
| Pain point | Low awareness. The service could be mistaken for a general cab booking option. | Lengthy guidance made important information difficult to find. | Details were scattered, sessions could expire and submitted requests could not be edited. | About 36 hours for the initial decision, followed by long email conversations. | Roaming could fail and locating the driver was difficult. | Uncertainty about the arrangements could continue after meeting the driver. | The extent of the service during the trip was not always clear. | Payment expectations could remain unclear until the service ended. |
| Emotional journeyMore at ease ↑ More concerned ↓ | ||||||||
| Design opportunity | Explain the purpose of SGT and make the service easier to discover.Awareness + mental models | Explain the service plainly and answer the most important questions early.Information architecture | Use progressive disclosure, drafts and a manageable request lifecycle.Cognitive load + user control | Clarify approval routing, request status and the provider’s next step.Visibility of system status | Capture an alternative contact channel and share provider information in advance.Service continuity | Communicate the pickup, route and drop-off details before arrival.Expectation setting | Organise service needs by city and confirm transport for the required duration.Shared understanding | Communicate arrangements to the traveller and provider in advance.Service closure |
Journey themes from five user interview sessions, across eight stages of the service. The emotional curve is a qualitative interpretation of those themes, not a measured sentiment score. Opportunities include the booking experience and the wider service.
The map helped separate the booking work from wider service opportunities. Progressive disclosure and request management addressed problems in the product. Driver contact, service expectations and payment arrangements also depended on the information exchanged with providers.
Explore the journey with the people who understood the service.
After the interviews and customer journey mapping, I hand-sketched an initial flow and wireframes to explore how someone would create an SGT request. I used the sketches in brainstorming sessions with Erika from TRM. The purpose was to make our questions visible while the journey was still easy to change.
We worked through what travellers needed to understand about SGT, what happened when a destination was not recommended, how to handle several cities or colleagues, and which travel details the provider needed. I also explored where a late-booking warning belonged and what confirmation should communicate. Erika clarified the scope, answered service questions and supplied the details we needed to move the flow forward.
Make the open questions visible before refining the screens.
- 01
Explain SGT
What is SGT? Safe trip Coverage Help Continue What does the traveller need to understand before requesting it?
- 02
Choose a destination
Where to? Destination country Select country Check destination guidance What happens when SGT is not recommended?
- 03
Build the itinerary
Other cities? Yes No City Add Repeat for each stop Can one request include several cities and colleagues?
- 04
Capture travel needs
Travel details Arrival Departure Hotel / pickup Service preference Contacts + co-travellers Which contact, pickup and service details does the provider need?
- 05
Review and submit
Review request Destination City details Traveller Late booking warning? Submit When do we show a late-booking warning, and what follows submission?
The early exploration separated service explanation, travel needs and service preferences into individual steps. As the scope developed, I consolidated the intake into the five-stage journey shown below. Destination guidance stayed close to country and city selection, while each city held its own travel details. This reduced repetition without losing the questions the service needed answered.
I worked through the decision paths before refining the high-fidelity screens. The flow needed to account for unavailable destinations, additional city stops, saved drafts, requests requiring TRM review and changes close to arrival. This gave the team a shared view of the interaction logic and the operational boundaries behind each screen.
DETAILED USER FLOW
Work through the decisions before refining the interface.
Request creation, approval routing and the changes that follow.
Open each flow to follow its branches. On a phone, use landscape or scroll across the diagram.
01Create the request Coverage, city stops, traveller details, drafts and review
02Approval and provider handoff Automatic approval, TRM review and service communication
03Change or cancel a request Edit boundaries, city removal, confirmations and cancellation
04Request & approval
Establish the destination and approval path before asking for more detail.
I structured the request into five stages: destination country, destination cities, traveller details, attachments and notes, then review and submit. Country selection came first because it affected whether the traveller could proceed through the booking tool.
We worked with TRM and Engineering to classify destinations and identify the criteria for automatic approval or TRM review. The design made that distinction part of the guidance and signifiers shown during the request. Where review was required, the request went to TRM over email for an approval or rejection decision. A fuller SGT admin console for managing those decisions was part of the roadmap beyond the traveller MVP.
Engineering created the Request Lifecycle Management APIs, the underlying services supporting how requests were created and managed. In parallel, Design worked with TRM and Engineering on destination classification and approval behaviour. My design responsibility was to translate those rules into guidance, signifiers and clear request states, so the traveller could understand whether a request qualified for instant approval or required a TRM decision.
Keep the booking journey connected to the people delivering it.
Discover SGT through the Concur confirmation email.
Select the country and review city guidance.
Itinerary, traveller details, notes and review.
Destination unavailable in the tool: SGT guidance and a route to TRM remain available.
Approval routing · Agreed destination and request criteria
Eligible requests proceed directly.
The request meets the criteria defined with TRM, Design and Engineering.
A person makes the decision.
TRM receives an email notification and makes an approval or rejection decision. A dedicated administration console was planned for the next phase.
Approved: continue to the provider handoff.
Declined: the request does not proceed to booking.
Approval and provider confirmation remain separate steps.
The traveller receives the relevant next steps and provider contact details.
A structured email includes the traveller in the conversation.
The provider follows up. TRM remains the escalation route.
Country coverage, city recommendations and approval status needed distinct meanings. A destination being available in the tool did not, by itself, explain every condition of the service. City-level recommendations and contextual guidance helped explain the service. Approval routing determined whether the request could proceed directly or needed a TRM decision. I kept these meanings separate: a recommendation is guidance about the destination, while approval is a decision about the request.
Create a transport request
Select a country, add a city and fill its travel details. Follow the request through to submission.
Build the itinerary around the city stops.
A business trip could include several cities, each with its own pickup, accommodation, service and departure details. I grouped those details by city so the traveller could build the itinerary in parts and review the whole request before submitting it.
The original journey map showed that trip details were scattered and sessions could expire while people gathered information. The step indicator provided orientation, and Save and exit let a traveller pause without treating the whole intake as one uninterrupted task. Notes and attachments carried additional context, including colleagues travelling together. This kept the main sequence consistent while allowing the request to contain the information a provider would need.
05Changes & cancellation
Give travellers control when plans change.
The existing form left travellers dependent on follow-up messages after submission. I brought viewing, editing and cancellation into the booking details, keeping the itinerary and provider contact information available alongside the request.
That control had operational limits. The design allowed self-service edits when arrival was at least 72 hours away. Closer to arrival, it directed the traveller to contact the provider. Changing the country required a new request. Removing a city warned that its travel details would be lost.
I made the final edit action explicit: submit changes to the provider. This distinguished reviewing or changing information on screen from committing an update that another party would act on.
Manage a changing itinerary
Open Trip actions to edit or cancel. Change the arrival scenario to explore the 72-hour boundary.
Explain the cancellation consequence before the traveller commits.
Providers had raised the cost of cancellations in the original brief. The cancellation flow therefore paired the action with a warning that cancelling within 72 hours of arrival could incur a fee. It also collected a reason, such as a change in plans or pricing.
The warning appeared at more than one decision point: review before submission, the traveller’s notification and the cancellation confirmation. This gave people an earlier opportunity to understand the consequence. The message left any applicable fee and follow-up with the provider.
06Service communication
Design the information each person needs at the handoff.
Email remained part of service delivery. I treated the notifications as part of the user experience, with a clear purpose for each recipient.
The traveller’s message acknowledged the request, summarised the destinations, supplied provider contact details and explained what would happen next. It set an expectation that the provider would make contact within two business days. This was a service expectation, separate from the approval decision.
The provider needed considerably more detail: traveller contact information, each city’s arrival and departure arrangements, hotel, service preferences, notes and attachments. The template grouped the itinerary by city and included the traveller in the email conversation. TRM remained the route for help or escalation.
Connect the traveller and the provider
Switch recipient and screen size to compare the information each person needs.
We designed desktop and mobile versions of the provider email. A clear reading order mattered when someone was checking the itinerary away from their desk. The contact fields also captured whether the traveller used WhatsApp, responding to the difficulties around international calling and airport coordination identified in discovery.
07Iteration & delivery
Keep the journey coherent as the delivery scope changes.
The service vision stayed clear, but the scope changed as engineering capacity, resource availability and compliance requirements became clearer. The work went through more than five collaborative brainstorming sessions and over ten scope and design iterations. Interviews and usability studies continued from discovery through wireframes and high-fidelity screens.
I had to keep the main tasks connected while those boundaries changed. The traveller still needed to understand the service, submit a useful request, know what happened next and manage a change. A scope decision affecting one of those tasks also had consequences for TRM or the provider.
Automation + review
Make the decision route visible.
Automatic approval served requests that met the agreed criteria. Requests needing judgement followed a TRM review route. A fuller admin console was planned to support that operational work.
Structured data + flexibility
Keep room for group travel.
The main journey used consistent traveller fields. Additional colleague details could be supplied through notes or attachments, leaving a more structured group flow as a future improvement.
Self-service + operations
Show where direct contact is needed.
The edit cutoff protected an operational boundary. Provider details gave the traveller a next step when self-service editing was no longer available.
More than five usability studies informed the MVP, alongside the interviews and journey mapping in discovery. The release began with a controlled pilot in one country before wider rollout. The work covered both the first request and return visits: unavailable destinations, saved drafts, city removal, restricted edits and cancellation. Maintaining those states across iterations was important because the service had to remain understandable when plans changed.
08Results & reflections
Less waiting for travellers. Less coordination for TRM.
The MVP connected destination guidance, a structured intake and request management within Trail. Travellers gained a clearer way to understand the service and manage changes. TRM could focus its review on requests that required judgement, while providers received a consistent itinerary and clearer change notifications.
Less time from request to confirmation
The project reported a reduction from 36 to 18 hours in elapsed service time.
Low-risk requests approved faster
More than 80% were approved within 15 to 30 minutes.
Increase in unique travellers
More employees used SGT to arrange their ground transport.
Fewer booking-related enquiries
Fewer questions about booking confusion, pricing and eligibility.
These were shared outcomes across Design, Engineering and TRM. My contribution was to connect the research findings to destination guidance, progressive disclosure, self-service request management and communication that supported the people arranging each trip.
What the MVP taught us
The MVP addressed the booking journey, but the customer journey map showed that access became even more important after the traveller left their desk. At an airport, a local SIM might not work, carrying or opening an Amazon laptop could be impractical, and access to internal tools could be difficult. A desktop booking experience alone could not support every moment of the trip.
TRM as the product team, Design and Engineering identified the next releases together. The roadmap addressed mobile access, provider operations and the administrative work around approvals. These plans extended the service beyond the scope of the traveller MVP.
A dedicated mobile app
The team planned mobile access to trip details, provider contact and request status, so travellers would not need to open an Amazon laptop at the airport. The design direction accounted for unreliable SIM access and the practical limits of using internal desktop tools while travelling.
Provider onboarding and lifecycle management
A fuller provider experience was planned for onboarding transport partners and managing their ongoing participation in the service. This would give the team a more structured foundation as SGT expanded across locations.
A dedicated administration console
The planned console would bring requests requiring manual approval or rejection into a shared workspace for TRM. It would support the review work behind the traveller experience and reduce reliance on separate email conversations.
Concur integration and in-product communication
We also identified retrieving an existing itinerary through a Concur booking ID and keeping provider conversations within the request. These opportunities addressed repeated entry and the fragmented context of email or WhatsApp exchanges.
For me, the main lesson was that service design depends on understanding both the traveller’s context and the operational work behind it. The delivery constraints changed what could fit into the MVP. The journey map helped the team retain a shared view of the experience while deciding what to launch first and what belonged in later releases.