Add New Signals to Your Skill That Alexa can Consider for Name-Free Requests

Today, we are announcing the name-free interaction (NFI) Toolkit, which makes it easier for customers to find and open skills without having to remember and say the skill’s name. Alexa uses keywords, skill descriptions, and categories to surface relevant skills when customers say things like “play an adventure game.” We’re continuously improving name-free interaction for customers so they can find a skill without using the skill’s name.

The NFI toolkit offers two new capabilities. First, you can configure up to five suggested launch phrases customer might use to launch your skill. Second, you can identify specific intents from your interaction model that you believe are good candidates to handle name-free requests. These two capabilities allow you to provide Alexa with additional signals to consider when matching a name-free request to a helpful skill. The NFI toolkit is available in preview for English language skills in the US, UK, and IN and includes new authoring tools within the Alexa Skills Kit (ASK) Command Line Interface (CLI) and Skill Management APIs (SMAPI). Skills using the NFI toolkit will work with existing publishing tools and processes and will need to adhere to additional validations and policies that ensure customers get the best responses and highest-quality experiences.

Configuring your skill for name-free interactions with the NFI toolkit can create new pathways for customers to discover your skill. Customers will also benefit from being able to ask Alexa questions naturally to find a relevant, helpful skill to fulfill their request.

Sign up for the preview here. 

Today, customers have to explicitly ask Alexa for a skill by name. The invocation name must be used with a phrase like: [ask, open, tell, etc.] <your invocation name> to [do something]. For instance, for an imaginary trip planning skill called “Leo’s Trip Planner,” a customer must request to open the skill with a phrase like, ”Alexa, ask Leo’s trip planner to plan my trip to Grand Central Station.“ As a developer, you have to strategically choose an invocation name and adhere to the best practices published here to ensure a successful response. Until today, you had no way to update your skill’s launch phrase based on how customers used your skill. 

Now, with the NFI toolkit (preview), you can use a suite of build-time tools, runtime APIs, and analytics dashboards to configure your skill to handle name-free interactions. You can suggest up to five skill launch phrases and indicate the intents in your skill that Alexa can consider when routing name-free requests. For instance, for the imaginary trip planning skill, “Leo’s Trip Planner”, one of these phrases can be, “Alexa, help me plan my trip.” Notice that no invocation name was used. In that same example, you can also enable a “Get Train Schedule” intent for NFI, and so when a customer asks, “Alexa, when is the next train to Grand Central Station?” Alexa can consider your skill to answer.

The NFI toolkit lets you provide build-time signals, such as name-free intents and utterances, and runtime signals (canFulfillIntentRequest) to teach Alexa what utterances your skill can handle. If accepted into the preview, your skill can be considered for name-free interactions when it meets the eligibility criteria (see FAQs #3 and #4). After your skill is onboarded, Alexa can consider your skill for name-free interactions. When Alexa receives a name-free request from a customer, Alexa evaluates a variety of signals in real time to provide a relevant skill or experience to fulfill the customer’s request. In addition, Alexa occasionally asks clarifying questions and constantly improves the routing decisions based on direct and indirect customer feedback to improve routing over time

An important component of name-free interaction is targeting the customer requests that are the right match for your skill’s capabilities. You can configure the intents and utterances you want Alexa to consider for name-free interaction via build-time and runtime signals. After being accepted into the NFI toolkit preview, there are three steps to enable your skill to handle name-free requests.

First, you define the NFI container in your interaction model. Inside the container, you define launch phrases, or utterances, that customers might use to launch your skill. You also identify which intents you want Alexa to consider for name-free interactions. The launch phrases and intents you identify should be the ones that customers would request naturally if they were familiar with your skill’s functionality, and you should feel confident your skill can handle the request to the customer’s satisfaction. Patterns of customer dissatisfaction, like customers repeatedly interrupting your skill by saying “Alexa, stop,” may affect your NFI eligibility over time.

Returning to Leo’s Trip Planner skill, intents that you might identify for NFI include those that help customers start trip planning, like reading out a bus schedule, or reading out a train schedule. You should not identify intents for NFI that would be unnatural for customers to land on, like intents to get help or to retrieve a past trip. Launch phrases would include common phrases a customer may use when trying to find a skill like yours, like “help me plan a trip” or “plan my commute”.

Second, CanFulfillIntentRequest (CFIR) is a highly recommended interface that helps to ensure that Alexa only routes customer requests that are a good match for your skill given the context of the customer at runtime, or when they use your skill. It can reduce false positives derived from your NFI container. For example, in Leo’s Trip Planning Skill, you identified the "get train schedule" intent for NFI. However, let’s assume Leo’s Trip Planner Skill focuses on schedules for a geographical area, like Southern California. CFIR helps you signal to Alexa situations in which your skill wouldn't be a good match for getting a train schedule - for instance if the customer asks for train schedules to Baton Rouge.

To complete the process, you can trigger a build and will receive new notifications or error messages confirming the status of your skill and the NFI models. After a successful build, you can test your skill in isolation with name-free requests, and the simulator will attempt to match the name-free request to the relevant intent as defined by the interaction model. Once testing and iteration is complete, you can submit your skill for certification. Your updated skill will be ready for usage right away, but it can take up to 30 days for Alexa to begin considering your skill in the name-free scenarios you indicated in your NFI container.

The new invocation dashboard (preview) in the analytics section in the Alexa Developer Console shows you what paths customers use to invoke your skill, including direct invocations, one-shot invocations, name-free interactions, quick links, and more. This dashboard is included as part of the NFI toolkit preview. All data on the dashboard can be filtered to show data for new and returning customers. You can view frequent utterance data for the name-free invocation channel to learn the most common name-free requests sent to your skill and make adjustments as needed to your skill’s NFI container. 

Developers in the preview will have access to the NFI toolkit in the Alexa Developer Console and the CLI. With the NFI toolkit, you can enable natural, flexible ways for customers to discover and use your skills. To apply for the preview and review frequently asked questions about the NFI toolkit preview, please visit our website.