Amazon Alexa

Skill Builder


The new #Alexa Skill Builder user interface is amazing! I had no trouble understanding how to build my first interaction model #FirstSkill

- Sanath Kumar Ramesh



Team & My Role.

  • 1 UX Designer (me)

  • 4 Product Managers

  • 20 Developers

I led the design of the new Alexa Skill Builder since the inception of the project in October 2016.  I helped uncover insights and translate concepts into features that address customer behaviors and motivations. Created frameworks and prototypes to share the vision, design principles and content strategy. This helped to evangelize ideas, gain alignment and drive decision making. I partnered with my product manager to define the product. I evangelized customer goals and balanced business goals. I prioritized and negotiated features for launch and beyond. Designed across, collaborating with partner product teams and tech partners to translate product features while also designing down on defining the application framework. I executed journeys, wireframes, prototypes and all other design artifacts. I presented work to gain buy‐in from executives and stakeholders and many other teams in Alexa throughout the project lifecycle.



Introduction to the Alexa Skills Kit.

The Alexa Skills Kit (ASK) is a collection of self-service APIs, tools, documentation, and code samples that makes it fast and easy for you to add skills to Alexa. By using ASK, you can leverage Amazon’s knowledge and pioneering work in the field of voice design.

The Alexa Skill Builder was a project to develop a new interaction model designer in the developer portal which includes visual designer for intents, Multi-turn dialogs and search/browse NLU 2.0 built-in library.

The Alexa/Lambda testing story is pretty tasty. You can just copy the JSON input from the Alexa test simulator and create tests inside the Lambda editor. Much more efficient for testing text output.
— Sam Strong

Assessing the old console.

In favor of speed to market, the ASK team ramped up development and launched the SDK with minimal design consideration. This proved to be the right move to get developers bought into ASK, but created risks for scalability and rapid iterations based on community feedback. 



Putting the user at the center of the narrative.

I synthesized the developer insights into a set of overarching design principles to guide all ideation and provide a framework for assessing concepts. The key principle for maximizing usability is to employ iterative design, which progressively refines the design through evaluation from the early stages of design. The evaluation steps enable the designers and developers to incorporate user and client feedback until the system reaches an acceptable level of usability. By adhering to the following user centered design principles, we were able to associate tactics throughout our ideation and iteration.


Understanding our developers.

Collaborating with the marketing team, our team created provisional personas of potential ASK users based on online research and developer survey results.


Identifying and prioritizing pain points.

Through affinity map exercises with skill developers, results from our developer surveys, qualitative feedback from Alexa champions and ASK developer community forums, we were able to group pain points of the existing skill builder and prioritize through a cross-team design vision workshop.



The voice of the customer.

The following friction points and developer insights provided the foundation for our design and iteration.

Screen Shot 2018-06-12 at 4.32.59 PM.png


With the complexity of interaction model creation being limited to a form box requiring JSON code input, users had a difficult time understanding what an interaction model (the skill’s conversational model) consisted of. Conversation and voice design is heavily baked into code today which requires skill builders to concurrently design and code to build their skill's voice user interface. Providing a UI.


Interaction model is not only esoteric to build, but is only as good as the contextual logic that accompanies it to deliver a great conversational experience. Skill builders are unable to realize quick results and are resorting to higher level tools (templates) which are highly inflexible. The structure to what makes a skill is not clear. Terminology needs to be clear and consistent. We need to make sure to avoid common terms that can confuse the developer.


This high friction entry point lead to high bounce rates. Skill builders come to the developer portal intending to create a skill but are impeded by the current interaction model text boxes. We hypothesize that a more intuitive interface will encourage them to explore further. Fear of not knowing how to code discouraged users from creating an interaction model. Novice skill developers with limited or no coding experience found it very difficult to get started with the tool. Skill builders are looking for simpler tools to build and publish skills.




I came up with several explorations for each of the pain points and made some sketch and Lo-Fi wireframes of my design solutions. I did some preliminary validation with the internal developers for high level concepts on the Lo-Fi UI wireframes and used the guerrilla feedback to refine my wireframes and narrow down my solutions for the interactive prototype.


Interactive prototypes.

I jumped into Axure to create an interactive prototype of the proposed solutions to get into usability labs for testing. The testing phase is crucial as it helps us determine which aspects are worthwhile and which parts need to be revised or discarded.. Creating the prototype allowed our design team to not only evaluate, but also test the product and ensure optimal developer experience.



Assessing and de-risking.

Our team conducted a study to evaluate Alexa skill interaction model creation with representative customers to test whether the Skill Builder meets specific usability criteria. Usability testing was conducted by having skill creators complete a series of structured tasks specifically designed to uncover issues and inform product iteration.

The goals of the study were:

  1. Evaluate if creators can complete key tasks within an optimal success rate;

  2. Identify if there are key missing patterns in the interaction flow;

  3. Determine if creators exhibit anxiety, confusion or frustration during their journey; and

  4. Test, evaluate, and validate the high value, risky assumptions and hypotheses for IxM creation.


Screens upon screens.



Navigation & Wayfinding.


Interstitials & Editors.


Design system

I created sets of specification docs to communicate requirements to both engineering and aid QA in writing test cases. These deliverables consisted of the CX Spec—requirements, customer journeys and the design system (informing typography, attributes, size rules, button sizing and behaviors, hierarchical content organization, iconography, measures, spacing and styles for all patterns). I also authored complementary documentation to communicate animation and timing keyframes for micro‐interactions.




We launched the public beta of the new Alexa Skill Builder, April of 2017, available to all Alexa developers in the developer portal. The following adoption KPIs are for the span of 3 months. 6.9k of the 24.5k skills created since launch are opted into the Madden beta (28%), and 1,171 include dialog models. 61 of the 171 skills created since launch that are currently in certification are opted into Madden (36%), and 10 include dialog models. 194 of the 987 live skills created since launch are opted into Madden (20%), and 22 include dialog models.


What the community is saying.


What the press is saying.



Learnings + Takeaways.


The Alexa Skills Kit is a great product with a grand mission to get developers and content creators to start thinking voice first. To achieve this, it faces one of the biggest challenges — changing a person’s expectation of what an intuitive voice platform can do and what are the best practices for conversational design. The skills kit is packed with a lot of powerful tools to make intuitive voice experiences, such as the self service APIs and CLIs, in depth tech documentation along with voice best design practices guidelines. But changing a user’s perception is painful, so it’s even more important to make it as easy and friction-less as possible, without sacrificing functionality, for a user to build conversational applications. Which is why it is so crucial to include and make sure you do rigorous reviews wth partner product, tech  and testing teams. Always work backwards from the customer voice, championing and advocating for the user. This may all sound like a no-brainer, but it often takes more than one may imagine. It requires a bit of skill development to carry out successful validation sessions with all the perspectives being in one room. I am a firm believer and practitioner of the idea that UX not only needs to be agile, UXers can also lead the rest of the team in a way that puts team energy and skills to optimal use.