Introduction
Currently, people everywhere use multiple communication channels such as email, web chat, telephone, and platforms such as Slack, Microsoft Teams, and Facebook. Organizations need to be accessible to their customers over these multiple channels; moreover, they need to respond consistently in the same manner, regardless of which channel that a customer decides to use.
Users and customers have become increasingly demanding, requiring constant daily response, responses that are tailored to their behaviors and characteristics, and answers to more complex requests.
Conversational AI, also known as agents, respond to human questions and requests in an intelligent way. Organizations use agents to provide the first line of response to customers, handling common inquiries across multiple channels. Agents instigate a conversational dialog with the human, whether it's over the web or another channel such as SMS text service.
Issues around building agents
The following statistics depict the complex challenges that organizations currently experience when dealing with both employees and customers:
66 percent of customers opt for using self-service first before contacting a support representative, because contacting support representatives can be time-consuming, and the customer waits in a queue when all they want is an immediate answer.
90 percent of customers expect consistency across channels. Whether they're on the phone, website, or retail store, customers want the same treatment from all channels.
59 percent of channels are managed in silos; call centers have different incentives than retail stores, the website is managed separately, and so on.
Agents can help address these challenges, but agents can also be difficult to create and expensive to maintain. Traditionally, creating an agent requires skill and effort; you need developers to code the agent, subject matter experts (SMEs) to define the conversational flow, AI specialists to handle the language processing, and the agent needs to connect to your operating systems to access relevant data.
Microsoft Copilot Studio Agents
Microsoft Copilot Studio is designed to create agents whether the user is external (your customers) or internal (your employees). SMEs can build conversational flows by using a graphical interface, with no code, in a browser.
Developers and administrators can assist SMEs in building conversational flows and integrating agents with line-of-business applications by using Microsoft Power Automate cloud flows. Microsoft Copilot Studio includes AI language processing to create natural, conversational AI.
Microsoft Copilot Studio makes it possible for anyone to build an agent, which greatly reduces the time and cost that is involved in deploying an agent.
Use cases for agents
You can add conversational AI into an application to increase the user experience. Common uses for external agents include:
Answer questions or inquiries
Process return and exchange requests
Resolve a complaint or problem
Make a purchase, check inventory, make a recommendation, or track shipping
Introduce a product and use as a help desk
Make a reservation or reserve a ticket to an event
Manage email subscription preferences
Pay a bill or process a claim
Find a qualified person (for example, professional services)
Provide triage for healthcare and crisis communication
Request customer feedback and create quizzes, surveys, and competitions
A popular use for agents is in customer-support scenarios. For example, an agent can assist you with troubleshooting, warranty, and repairs.
Most people's experience with agents is in customer-support scenarios; however, more powerful use cases exist when using agents inside of an organization:
HR support
Employee onboarding
Employee ideas
Change management
IT support
Sales support
Case management
Issue management
Inspections
Internal process information
Reasons to consider building an internal agent include:
Consistent internal communication
Convenience, speed, and accessibility
All day, everyday availability
Its ability to act on behalf of the user for routine tasks
Its ability to replace intranet or internal email for frequently accessed information
Solution architect's role in determining the need for agents
Solution architects should consider using agents as part of the solution when:
Replacing web forms, including lead generation forms and inquiry forms
Helping customers navigate and finding a person or information instead of using a website search
Redirecting inquiries to the right person in the organization, such as customer service agents or professional services staff
Agents aren't only for websites. You can also deploy Microsoft Copilot Studio to the following entities:
Facebook
Slack
Twilio
Email
Mobile apps
If the solution architect decides to use agents, they'll need to ensure that the project team follows the principles and best practices for building agents, as discussed in the remainder of this module.
Responsible AI
Microsoft has defined six principles for the responsible use of AI. Because agents are AI-based services, solution architects need to keep these in mind when designing agents. The Transparency principle is one that particularly applies to agents.
The Transparency principle is concerned with ensuring that the human understands that they're interacting with an agent. When an agent conversation is started, the agent should clearly state that it's a agent. The agent should state its purpose and limitations, such as listing the scope of what it can answer or do. An agent should enable the user to escalate or transfer to a human.
Agents work well when they're limited solely to their purpose and don't try to be too generic.