top of page


Aiva - Amazon IT Service 

Aiva is a web based AI-powered IT support tool that helps internal Amazon employees troubleshoot and self-serve IT issues

Aiva Laptop.png

My Role

I was a User Experience (UX) Design Intern who explored the scope, developed, and flushed out the Aiva project. 



Google Forms

Amazon Design Systems

I conducted Competitive and Comparative Analysis, hosted a Design Jam, created Customer Journey Maps and User Flows, conducted User Research and Analysis, Wire framed and prototyped designs, and developed a brand identity for Aiva


Kendall Arata, Amazon UX IT Services Team: Michael Berg (Manager), Selene Lee (Mentor), Yao Wang (UX Designer), Jessica Bao (UX Researcher), Sue Chen (UX Designer), Rahul Potgham (Product Manager), Andrea Calhoun (Product Manager), Ryan Clark (SWE)


Amazon UX Design Internship

12 weeks, June 2022-September 2022


What is Aiva?

Aiva (Amazon IT Virtual Assistant) is an AI-driven tool designed to provide self-service IT support to Amazon employees. It utilizes machine learning and intent-detection modeling to identify and address specific IT issues, offering solutions or directing users to relevant resources, ultimately reducing the need for live support assistance.

Currently, Aiva serves as a reactive support method. She is encountered once a customer experiences an IT problem and searches for a solution on or initiates a contact with IT support. Recent research indicates that more prominent placement of Aiva might incentivize use. Previous research also revealed that customers who encountered Aiva felt like she offered good solutions, but because a direct customer-Aiva interaction was not readily apparent, she was not optimized for efficient IT problem solving and troubleshooting. Therefore, there is a desire to develop and test concepts for appearance, placement, and form that Aiva takes on in the customer IT problem process, including exploring and envisioning Aiva as not only a reactive, but proactive support figure in the customer IT support journey.

Defining the Project

One of the first challenges I encountered at the beginning of this project was narrowing down the broad scope of my project. There wasn't much direction other than the fact that I was supposed to focus on Amazon's IT Service chatbot, Aiva. This was both a blessing and a curse, because I had so much room for exploration, but it was almost too much. So, I began reading up on every single chatbot/AI self-service resource available.


To start, I not only read up on what makes a chatbot helpful, but also started looking at how competitor tech companies served their customers when they had problems. I began conducting a comparative and competitive analysis on how companies offered their customers support. Apple, Kayak, T-mobile, Bank of America were among some of the companies I looked at, and I noted differences in user flow on how they each implemented AI/Customer Support in their Customer Service Journey. Because the overarching goal for each chatbot was different for each company (i.e. Apple's is for mostly to solve hardware and software issues, T-mobile is more for phone specific issues), I had to factor these into account when comparing them to Amazon's internal-facing self-service tool. 


Some of the key insights I noted were:

  • Most chatbots served as an intermediary between the customer's own self-service (i.e. looking through FAQs and searching) and a live support representative


  • Companies like Kayak, T-mobile, and Bank of America utilized chatbots as more humanized FAQs to minimize the amount of high frequency requests sent to live support


  • The accuracy and efficiency of bots like T-mobile and Kayak were low if an issue that was rare or unique occurred. Most times ended up being directed to a queue for a live service representative


  • Erica, Bank of America's AI bot, was extremely intelligent and efficient, breaking down expenditures and seemingly overwhelming numbers into digestible, visually supplemented diagrams and charts for the user to see.


  • Apple did not utilize a form of AI, instead, they had the user filter down his/her issue based on category of issue and provided helpful articles for troubleshooting. If user still could not find the answer, they were directed to a live support representative who was provided the context of the issue for smarter diagnoses


Based on these differences, I also created a user flow mapping out the customers thoughts and actions when encountering an IT problem in the current Amazon self-service model and focused on how the user's mental models affected how the they interacted with Aiva in chatbot and search form. From this user flow, I extracted some key insights that helped me get down to the root cause of why Aiva was not noticeable/optimized in her effectiveness.


  • Users groups who were directed to Aiva belonged to one of two groups: they either had a high productivity impact issue–defined as a problem requiring immediate assistance affecting an Amazonian's daily productivity, or a low productivity impact issue–defined as an issue which could be resolved later but not affecting an Amazonian's daily productivity dramatically


  • The low productivity impact issues could accumulate over time and turn into high productivity impact issues if not addressed within a reasonable amount of time


  • Users already had a negative association of IT and entered the process with a negative and/or frustrated mindset


  • Aiva was not noticeable on the IT page directly unless a user clicked on the "Get Support" button in the upper right hand corner of the page

The Problem

Keeping my comparative research in mind and from the user flow I constructed of an Amazonian with IT issues, I identified four main problems with Aiva's current representation and placement in the User IT Journey.

  • Aiva is not discoverable but perceived as a good option for resolving issues - Customers are unaware that Aiva exists and cannot recognize her as a definitive figure. When users that have IT questions, they are not recognized as a pathway to a solution because Aiva is not recognized as a tool to support them; she is not memorable nor exposed enough to the customer and is thus an underutilized resource. 

  • Current Aiva placements along customer journey are not optimized to predict and resolve customer IT problems given her capabilities - We do not know the best placement, presentation, and visual identity for Aiva during the customer’s user journey to facilitate solving the user’s IT problems. Where are the best places for Aiva to appear to help the user troubleshoot and solve their IT problems in the quickest and most satisfying way possible? Previous research has shown that solutions provided by Aiva are useful but often overlooked (e.g. “Aiva says” feature) or inefficient because Aiva’s answers are seen as duplicated from self-search.


  • The current Aiva’s tone does not resonate with user’s mental models when they encounter an IT issue - Previous research on Amazon IT support revealed that customers describe IT support as “punitive, policed, and matter of fact” Generally, customers perceive IT support as cold and unapproachable, lacking personality.


  • Current Aiva is solely utilized for reactive purposes and does not provide predictive and preventative support - Aiva is only prompted to help customers solve IT problems after they have occurred (e.g. troubleshooting hardware issues, laptop crash, etc), thereby she is only encountered toward the end of the customer IT problem process when customers are already in a negative mindset. Her ML capabilities can be leveraged to serve the customer proactively, thus reducing the number of IT issues Amazonians encounter in the first place and preventing future issues from occurring.

Because most of these problems originated from lack of user research on Aiva's current placement and personality, I decided to focus on understanding and crafting the experience the user has before encountering and solving an IT problem and during the IT problem-solving process to optimize the overall Amazon self-service pipeline as a whole. The solutions that I would create to address this main issue would be designed to consider best customer value without being constrained by Aiva’s current and future technical limitations.

User Journey Map

To start, I wanted to identify placements for Aiva along the user journey where she could facilitate customer self-service. So, I constructed a user journey map. I decided to segregate it into both a high and low productivity impact problem scenario that the customer could encounter to better visualize the gaps in self-service. From there, I imagined various ways that Aiva could appear and help along the user journey. Some of the key ways I imagined Aiva were:

  • Tamagotchi Aiva - Aiva could adopt a personable form and act as a overall IT health indicator, leveraging the concept of gamification to incentivize the user to keep their computer "healthy"

  • A figure pop-up that scheduled 1:1 meetings to dedicate time toward educating an Amazonian on how to keep their computer and corporate devices up to date


  • A virtual pocket assistant that could be accessible on personal devices 24/7 in case of an emergency

For the high productivity impact issue, I focused on the issue of password expiration–every 1 year, an Amazonian's password on their secure account is reset and the user is required to change it to something else. If the user does not change her password in time, she is locked out of her computer and must either access First Aid on her phone by going to or call a live service representative.

For the low productivity impact issue, I focused on the issue of hardware replacement, and I imagined that an Amazonian wants to request a new laptop because her model keeps crashing. However, she does not know where to go to request a new laptop replacement, nor does she know there is a general eligibility requirement that either allows her to request a new laptop or not.

User Flow

After mapping it out, I noticed that Aiva's touchpoints for each scenario were located along different points of the user journey. For high productivity impact issues, her touchpoints were more proactive, located before or just when the user becomes aware of the problem. For low productivity impact issues, her touchpoints were more reactive, located when the user is identifying the problem or when they are implementing the solution. I thought this was interesting, so I iterated on this user journey map to further delve deeper into the different proactive and reactive ways Aiva could help a user before, during and after encountering an IT problem.

Based on these new user flows, I was able to envision Aiva as a tool that could be used as both proactively AND reactively to better educate users, identify problems, and solve user issues in the overall Amazonian daily experience. 

Design Goals

After revising the user flows, I was able to pinpoint goals that my designs and wireframes should address, including:

  • Driving customer adoption of Aiva as the main IT support figure and go-to IT resource for solving IT problems by building and maintaining customer trust. Developing and implementing a consistent personality and recognizable brand identity. This includes exploring different touchpoints along the customer IT problem process where Aiva can offer assistance. Imbuing Aiva with a memorable, fun personality and implementing gamification methods to encourage product familiarity and promote a positive perception of Amazon IT support. 

  • Uncovering Aiva's opportunities on different touch points of users journey to prevent and diagnose IT issues

    • ​Employing proactive support methods to anticipate hardware/software issues and prevent high-frustration, high severity IT problems (e.g. computer crashing, losing access to ACME) from occurring and encourage proactivity on user’s part by creating gamification methods to encourage engagement. 
    • Enhancing the customer experience of Aiva reactive support methods 


  • Exploring different variations of the Aiva experience for the user to optimize her capabilities. Determining the best appearance, placement, and form that Aiva appears as well as the features she could add into her current capabilities to make the the user IT support process as quick, easy, and satisfying as possible. 

Design Workshop (Inclusivity & Accessibility)

Since this was envisioning a novel idea, I had to get some quick insight into what type of form Aiva could take that users would be most receptive to and combat users' already negative connotation of IT into a positive one to mock up some design concepts. To do this, I hosted a Design Jam with my team to develop Aiva into a more humanized bot who was friendly, educational, and had a personality that most users were receptive to.

Design Jam Workshop - 3 hours with team members and cross-functional teams

The most popular form that Aiva took was the dog/pet, because it not only tied in well with Amazon's corporate pet policy, but it also had an overall positive and friendly connotation and allowed the user to take care of it analogous to taking care of their computer.

Design Concepts and Wireframing

Iteration #1

The first iteration of designs focused on Aiva as a gamified experience to incentivize the user to take care of her and their computer. She is integrated in dog/pet form into an external web app that pops up on the users computer when they have updates or when IT issues are detected. Users can take action and resolve detected issues in a timely manner to keep Aiva and their respective Amazon devices "healthy" to prevent the frequency of high productivity impact issues.

Iteration #2

My second iteration focused on refining the designs to web app standards, as well as incorporating another proactive feature with Aiva called "Tip of the Day." This feature is a daily tip relating to solving common IT problems that is given by Aiva and rewards the user for engaging with the Tip of the Day, again utilizing the strategy of gamification.

I settled on three features:

  • Tamagotchi Aiva

  • Aiva Tip of the Day

  • Aiva Outlook Integration


After deciding these three features, I created final wireframes and prototypes to demonstrate the functionality and features of my app

Clickable Prototype

Final Presentation

Final Presentation Deck used to pitch design concepts to senior leaders and managers


After pitching the design concepts to senior leaders, I earned a return offer with the opportunity to continue working on Aiva, focusing more on usability testing and delving deeper into the user research. My pitch effectively aligned design goals with Amazon's strategic business goals, and it influenced a 3 year plan to develop the web application I designed with eventual global rollout and company-wide adoption.

What's Next?

  • Explore Aiva as an all encompassing support assistant in-app experience

  • Implement user testing to further validate Aiva's touchpoints along user journey and determine the most efficient features that solve both high and low productivity impact issues


  • Conduct more user research to define user conception of productivity

  • Determine classification between high and low productivity impact issues

  • Investigate impact of variations of Aiva’s personality


  • Explore cultural perceptions of Aiva from users of different cultures in order to promote inclusivity and diversity

What's Next?
bottom of page