• How can I use Power Virtual Agents in a language that’s not part of the 23 supported languages?
• Can I create a single Power Virtual Agents chatbot that responds to different languages instead of creating a separate bot for each language?

While support for single-bot multi-language is work in progress, there are also scenarios where a real-time translation layer can benefit chatbot makers and end-users, so that a single bot can be customized once, in one language, and automated translations are provided for all the messages between the user and the bot, allowing to deploy to more markets quickly, while reducing the implementation effort.

What is the Translator Bot sample solution for Power Virtual Agents?

The Translation Bot solution is a code sample that our customers can re-use and adapt to their needs. It showcases how to add a real-time translation layer by creating a relay bot on the Azure Bot Service as well as a connection to the Azure Cognitive Services Translator. It communicates with Power Virtual Agents using the Direct Line API.

The solution lets chatbot makers configure language recognition at every turn, allowing to translate various languages that customers input into the same Power Virtual Agents chatbot. Additionally, chatbot makers can customize dictionaries to offer specific translations for user queries – for example for specific industry terms.

The Translation Bot sample solution also supports adaptive cards, Power Automate cloud flows, the Dynamics 365 Omnichannel for Customer Service chat widgets and the Dynamics 365 Omnichannel Voice channel.

In the below example, the Power Virtual Agents chatbot recognizes the initial language used by the user – in this case, Spanish – and responds in the user language. When the user starts using English, the chatbot answers in English.

Frequently Asked Questions

How does the Translator Bot work, and what components do I need to set this up?

Here’s an overview of how the solution works. You will need an Azure Service Bot, Azure Cognitive Services Translator, and a Power Virtual Agents chatbot:

1. The user sends a message in their desired language.
2. The middleware inside the Azure Bot Service intercepts the message, translates it, and sends it to the Power Virtual Agents bot.
3. The Power Virtual Agents bot triggers a topic based on the message received from the user.
4. The Power Virtual Agents bot’s response is sent back to the user.
5. The middleware intercepts and translates the message according to the user’s language.
6. The user receives the translated message from the Power Virtual Agents bot.

You can choose to detect the language at every turn or to set it at the beginning of the conversation.

What are the languages supported by the Translator Bot?

This solution supports all languages available through Azure Cognitive Services Translator: Language support – Translator – Azure Cognitive Services.

What is the cost of using the Translator Bot solution?

In addition to Power Virtual Agents pricing, the additional required Azure components would be billed on an Azure subscription. While most Azure services have a free tier to get started, billing is based on usage and on the selected tier. See pricing for the Translator Service and  App Service pricing

How can I deploy the Translator Bot solution in my tenant?

Setting up the Translation Bot should be straightforward once the required Azure resources are provisioned. The step-by-step for this is detailed here: PowerVirtualAgentsSamples/MultilingualBotSample at master · microsoft/PowerVirtualAgentsSamples · GitHub

How is the Translation Bot supported by Microsoft?

Although the underlying features and components used to build the Translation Bot are fully supported, the code itself represents a sample implementation of these features.

Our customers and community can use and customize these features to meet their requirements. Any issue with the Translation Bot sample solution should be raised on the Power Virtual Agents Samples GitHub repository and not through Microsoft Support.

The post Make your Power Virtual Agents chatbots multi-lingual with the Translator Bot sample solution appeared first on Microsoft Copilot Blog.

1. Empower subject matter experts to drive engagement in your business using Power Virtual Agents
2. Build a bot that integrates with your backend systems with Power Virtual Agent
3. Best practices for building an intelligent bot with Power Virtual Agents

	Empower subject matter experts to drive engagement in your business using Power Virtual Agents Cleber Mori; Emma Archer; Omar Aftab; Pawan Taparia Learn more about how customers are improving engagement and productivity and reducing costs through personalized conversational intelligent virtual agents. In this session, get an overview of all new…
	Build a bot that integrates with your backend systems with Power Virtual Agent Marina Kolomiets; Pawan Taparia; Salem Bacha Learn how to integrate Microsoft Power Virtual Agents with your backend systems to enable bots to act on behalf of your users. As part of this session, we cover invoking Power Automate flows and call…
	Best practices for building an intelligent bot with Power Virtual Agents Murali Kumanduri Learn more about best practices that our customers follow to build successful virtual agents. This session covers what it takes to build a virtual agent that provides the best business outcomes, and how to improve your virtual agent and drive engagement…

Visit us at our Community Forum to post questions and discuss Power Virtual Agents.

The post Microsoft Business Applications Summit 2020 – Power Virtual Agents sessions quick reference appeared first on Microsoft Copilot Blog.

A Crisis FAQ bot can help you address some common questions of employees, for example, in the face of crisis. You can sign up for the Power Virtual Agents trial version to create a bot (trial is free for 30 days and currently extendable after 30 days) and use the pointers outlined in this blog to create and customize topics as needed for your audience. This blog also walks you through an example with step by step guidance that you can expand for other scenarios. The key goal of this blog is to help you get a Crisis FAQ bot deployed quickly.

Identify the topics to design

Taking some time to plan your topic coverage for your Crisis FAQ bot will help you make most efficient use of your authoring time and give your bot the best chance to successfully answer the questions from your users.

When first thinking about what topics your Crisis FAQ bot will need, it can be helpful to think about all the interaction points customers/employees/students have with your organization. It’s essential to consider this from the point of view of the people who will be using the bot.

Below is an example of some key suggested topic groups for a bot built to handle a crisis response. This example is based on similar crisis efforts like the CDC site FAQ.

1. Emergency contact information for your organization
2. Advisories specific to your organization
3. General public health information (for example, based on CDC site FAQ)
4. Preventive measures and tips (based on CDC site FAQ)
5. Exposure assessment (based on CDC site FAQ)
6. Frequently Asked Questions (based on CDC site FAQ and on actual questions that have been asked of your organization from all sources, including your bot in production)

These topics can be created and configured in the Power Virtual Agents authoring canvas as shown below.

You can also automatically generate a set of topics from the CDC site FAQ or another online content source, using the Suggested topics feature in Power Virtual Agents. This will generate a list of topics and trigger phrases using the content from the FAQ website. You can then review and add the applicable topics from this auto-generated list to your bot.

Getting the user to the information they need

Write down the questions the user might ask the bot that should start a topic. These are your topic “triggers.”

Then, write down the questions the bot should ask to narrow down the scope of the problem. (Asking as few questions as possible in the bot conversation keeps your user engaged and keeps your bot authoring efforts efficient.) What are the possible responses to your questions?

Use these questions to create a logical conversation tree for the steps needed to determine which answer your user needs (in other words, so that you know exactly what the issue is).

In this content design example, a user saying “hi” or asking a vague question (“I need help”) will see the Greeting topic, which redirects to a topic that asks which general topic the user is interested in. Each topic group will ask clarifying questions to branch out to individual topics, and the user will be guided to the answer they are looking for. If the user starts by asking a specific question, it will trigger the specific topic directly, based on the trigger phrase associated with it.

Some guidelines to keep in mind as you build out the trigger phrases and topics:

1. You can think of each topic as having a unique verb, noun, or verb + noun combination (e.g.: travel, meetings, working remotely). If you find that it is hard to differentiate the triggers for two topics, consider combining those two topics and guiding your users with questions within the combined topic.
2. Within the trigger phrases for a topic, use a variety of words that your users will likely use to mean the same thing. (eg, “travel” “trip” “visit a customer” “go to an event”)
3. For words that are common across multiple topics (such as your company name)  there is no value in adding them to trigger phrases because it does not help the AI to differentiate the topics.

Let’s take an example use case and walk it through.

Example topic: travel policy

User wants to know company specific information regarding travel policy during crisis.

Triggers

• Need info on business travel policy
• What is the company guidance on travel to APAC?
• Can I still go to a customer event?
• I just got back from a trip
• Recently visited an affected country

Conversation tree

This conversation tree takes the user through a guided path from the high level topic groups to specific advice for employees regarding business travel. Your topics can branch based on variations of policy or different situations. For example, the topic below branches based on the user’s response to a question about their travel status.

Other things to keep in mind

You can use Power Automate to take action during the virtual agent conversation. Consider what actions can be built for your topic. You can use Power Automate to look up and provide information, take action on behalf of the user, and keep the user in the virtual agent experience instead of sending them to other sites. For example, in the context of crisis response, you can build actions for “Reporting your work status,” so the employee can update their work status through the bot. This can be done through a combination of Power Virtual Agents Authentication and a Power Automate action.

You can efficiently gather information about the user by asking them to log in to their account from the bot using the Authentication node and collecting the user context (such as office location) as variables to use within the topic. You can also use entities and slot filling to skip questions you already have the answer to.

Finally, the easiest way to deploy the Crisis FAQ bot is using the out of box options for publishing the bot and deploying to your website (“Custom website”) using the code snippet generated.

After deploying the Crisis FAQ Bot, you can track what your users are asking about and how the topics are performing, using the in-built Analytics capabilities in Power Virtual Agents. This will help you adapt your topics to address what your users are asking your bot.

DISCLAIMER
The bot topics and related guidance in this blogpost are samples and may be used with Microsoft Power Virtual Agents for dissemination of reference information only. The samples are not designed or intended to be substitutes for professional medical advice, diagnosis, treatment, or judgment and should not be used as such. The samples should not be used for emergencies and they do not support emergency communications. Microsoft does not warrant that the samples, or any materials provided in connection with the samples, will be sufficient for any medical purposes or meet the health or medical requirements of any person. The samples are not intended or made available for use as medical devices, clinical support, diagnostic tools, or other technology intended to be used in the diagnosis, cure, mitigation, treatment, or prevention of disease or other conditions, and no license or right is granted by Microsoft to use the samples for such purposes. You bear the sole responsibility for any use of the samples, including incorporation into any product or service intended for medical or clinical use, and for providing end users with appropriate warnings about using your crisis response bot.

The post Building a Crisis FAQ bot using Power Virtual Agents appeared first on Microsoft Copilot Blog.

Align the bot with your organization’s goals

The first step in onboarding is to align on business goals; in other words, why your organization needs a chat bot.
It is very important that everyone on the project be clear about what you are and aren’t trying to achieve through the bot, and why.
Some of the reasons for the “why” we have heard include:

• Handling repetitive information requests when a skilled human agent is not required
• Reducing the time it takes to get answers to end users
• Lowering support call volume, while also handling high call volume
• Improving support KPIs
• Using a bot as part of the brand strategy

Identify the right requirements

Next you need to start thinking about the initial and overall experience you want your customers to have with your company.
Brand, personality and the rest will come from the initial experience, as customer experience is connected to brand identity very closely. Bots should ideally be an extension of the company’s brand​, so it is key to understand the bot’s role in the company’s overall branding, what personality the bot should have, and the tone of the conversations throughout the experience.

Apart from the initial requirements analysis, the bot content also needs ongoing improvements to keep the bot relevant.
Factor this maintenance effort in as you identify your requirements.

Start small and scale up

Starting with pilot bots serving a limited number of topics and controlled exposure is an effective approach to ensure that your bot development plan is addressing the right areas you want to address.

For the pilot bot, you can pick a limited set of topics with a mix of simple (direct answers) and advanced topics (such as those that require access to back-end systems) to learn how customers are using the bot and how to improve the initial experience.

The reach of the bot largely depends on the entry points and channels it is exposed in, so pick the key channels that make sense for reaching your target customers.

Make sure your placement of the bot on your site contributes to a high likelihood of success for the topics you’ve chosen to cover, either by funneling only customers with those issues to your bot or by clearly stating which topics to use the bot for help with.

The traffic to the pilot bot should be gated in the initial stages and then ramped up gradually as you see stabilization in performance KPIs. Balance risk with the need to get enough traffic to start to get valid data.

Track analytics for ongoing improvements

After launching the bot, the performance of the bot must be closely monitored to ensure that it is meeting your KPIs and success metrics. Monitoring also allows you to identify areas for improvement.

Analytics tell you a story, for example, a high abandon rate may show you that your initial dialog is too complex or too long to follow for your customer. Do you really need to embed your full business process inside the bot? Or can you reference a web page to make it easier on your customer to fulfill the process? Remember that the quality and quantity of the content the bot has plays a major role in its success

Some of the common bot KPIs that you can track include engagement rate, resolution rate, escalation rate, customer satisfaction (CSAT) and abandon rate. You can measure these in the Power Virtual Agents analytics dashboard.

I hope you’ve found some of the insights and recommendations in this blog to be helpful when planning your bot system. Stay tuned for more blogs about best practices using Power Virtual Agents.

We’d love to hear about your experience with best practices for bot building!

Please visit our community forum at https://aka.ms/PowerVirtualAgentsForum and share your feedback.

If you have any idea requests, please submit them at https://aka.ms/PowerVirtualAgentsIdeas.

The post Best practices when planning to use chat bots appeared first on Microsoft Copilot Blog.