Designing your conversational messaging workflow

• Messaging deployment guide: Introduction
• Part 1: About conversational support with messaging
• Part 2: Designing your conversational messaging workflow
• Part 3: Planning your staffing and operational requirements
• Part 4: Rolling out conversational messaging support

Articles in the series

At its core, messaging is a way for your customers to have direct conversations with your agents. This is known as conversational support.

As your support needs grow, you may want to automate some of this communication, such as offering a simple greeting before handing the customer over to an agent or deflecting inquiries using knowledge base articles. Later, design a more complex bot that automates more of the conversation, asks the customer questions to help clearly define their support issue, routes conversations to the right agent, collects useful data from customers, and presents options to help them self-solve their problems. For many accounts, the ideal configuration is somewhere in between no automations and very complex automations.

This article contains the following topics:

• Identifying your conversational support goals
• Choosing your messaging channels
• Considering offline scenarios
• Choosing a conversation style
• Considering automating conversations with bots
• Next steps

Identifying your conversational support goals

Customers want their issues resolved quickly, through a single conversation, across a range of channels, and without having to repeat themselves. Conversational support with messaging provides this and more.

There are many reasons you might add conversational support, but it's important to understand the specific reasons it's right for your organization. Your goals and intentions should inform the decisions you make about setting up and maintaining your conversational support workflows in the future.

To help articulate your goals, consider the following questions:

Understanding your customers

• What channels are your customers most likely to use?
• What are your customers' communication preferences?
• What devices do your customers use when seeking support?
• When do you receive the heaviest volumes of messages? Time of day? Day of the week? Specific times of month or year?
• How likely are your customers to have simultaneous separate conversations with your agents?

Understanding your business

• How long does it take for your agents to engage with a ticket? What about solving a ticket? On average? What about longest and shortest times?
• Are your customers satisfied with the queue wait times and resolution times? Are you hoping to improve customer satisfaction?
• How many agents are typically involved in solving a ticket?
• Do you need extensive reporting capabilities?
• How much of your agents' work is repetitive? How much of the repetition could be eliminated with bots and automation?
  • If you're planning to use bots and automation to deflect tickets away from agents, what are your specific goals around deflection? Completely deflect the tickets with self-service options? Deflect some of the tickets, but expect a majority to still reach an agent? Provide customers with more information to aid the conversation when an agent is available?
• What channels can you realistically support with your current organization?
  • What channels do you currently use?
  • Are there any channels you can remove because they are under-utilized or too expensive for the value they provide?
  • Are there specific channels you're looking to add based on agent or customer feedback?
  • If you're considering social messaging channels, which specific social channels would work well for you?
  • Are there specific types of inquiries that need more immediate, conversation-based support?
  • Are you able to adjust staffing and schedules to support more or different channels?
• How do you want to handle customer requests that come in during non-business hours?
• How important is efficiency? How efficient are your agents currently? Are you looking to increase efficiency?
• How do you route work to agents? Do you need a push-model for work assignment, or does a pull-model work well in your organization?

Choosing your messaging channels

After identifying your goals for providing conversational support with messaging, you should have a clear idea of the messaging channels you want to implement. These are likely a compromise between what's most affordable for your company to provide and what your customers will benefit from the most. You can choose from the following:

• Website channels
• Mobile channels for Android, iOS, or Unity SDKs
• Social channels

Considering offline scenarios

Messaging offers a persistent and often asynchronous style of communication. It's important to plan for situations where either an agent or customer isn't available to respond immediately to a message. Consider the following common scenarios you should take into account when designing your messaging workflows.

Agent is offline or customer responds after business hours

There are a few ways you can handle this scenario:

• If you have agents online 24/7, you can implement a policy that agents should reassign tickets before their shifts end to prevent communication bottlenecks from occuring while they're offline. You could create a view of unassigned messaging tickets to aid in this.
• If you don't have 24/7 customer support and an end user responds to an unsolved ticket while your agents are offline, the agent will see the message in the notification list when they next sign in to Zendesk.
• For longer periods of agent unavailability, consider using the Out of Office app from the Zendesk Marketplace.

Customer is no longer responsive

A conversation is considered idle if no message has been received in 10 minutes. Likewise, users are considered idle after 10 minutes of inactivity. If an end user stops responding mid-conversation, you have two options:

• The agent can ask whether additional help is still needed. Then, after the end user has been idle for 10 minutes, the agent can communicate to the end user that they are updating the ticket status to Pending or Solved.
• You can leverage automation to change the ticket status to Pending, Solved, or Closed after the 10-minute idleness threshold is reached.

In both cases, you should keep in mind that changing the ticket status to Solved could trigger your bots to display a customer satisfaction (CSAT) query. It's considered a best practice to exclude auto-solved tickets from the CSAT trigger, if you have one.

Customer responds after the ticket was solved or closed

When a customer responds to a solved ticket, the conversation goes back to the original agent with a status of Open. However, if a customer responds to a closed ticket, they start over with your messaging workflow.

Choosing a conversation style

Depending on your organizational needs, you can build a messaging workflow that's as simple or complex as you like. We recommend adopting this golden rule: don't over-engineer your messaging workflows.

It can be quite difficult to undo bots and automations after you've implemented them. Therefore, it's important to be pragmatic. Design your workflows for simplicity first. Then, as your knowledge of messaging and conversational support grows, you can iterate and improve your messaging workflows with more complex automations.

Consider the following conversation styles:

• Short and simple

  If a streamlined, live chat-like approach works best for your agents and customers, you can implement a workflow with little or no automation. With minimal configuration, you can offer an automated greeting and create a simple form to collect essential information from the customer to aid the agent who helps them.

• Intricate and comprehensive

  Alternatively, you can use conversation bots to address complex customer service scenarios and provide equally complex messaging solutions. Conversation bots can include preconfigured options for your customer to choose from, pull information from external sources into the conversation using API calls, offer choices based on customer location or time of day, and more.

• Meet in the middle

  If you're like most Zendesk admins, you probably want to design a messaging workflow that falls somewhere between simple and intricate. You can use bots to automate as much or as little of your messaging workflow as you want. Just remember the golden rule: don't over-engineer your messaging workflows.

Considering automating conversations with bots

While choosing your conversation style, you might have realized you want to incorporate some bots and automations. Before you start planning it all out, it's important to understand the three basic components of a messaging conversation:

• Greetings are the customer's first interaction with your customer support organization, either from an agent or an automated bot.
• Self-service options can be used as a next step to deflect simple tickets, leaving your agents more bandwidth to focus on more complex requests.
• Conversation workflow includes an agent's or bot's back-and-forth with the customer, how and when a bot passes the conversation to a live agent, and the path tickets take to being solved.

Each of these components can be automated, so think back to the conversation style you want to achieve and your goals for your messaging workflow when deciding how to configure them. Then consider the configuration options for each of these components.

Greetings

Greetings set the tone for the messaging conversation. When you configure a messaging widget for web or mobile, the standard messaging response is activated. This includes a basic conversation greeting that customers see when they launch the widget, followed by a message letting them know they're being connected to an agent.

The standard greeting can be modified to suit your needs. Similarly, you can configure automatic responsesto new customer requests for social messages, too.

Consider the following questions when designing your customer's first interaction:

• What do you want your first impression to be? Casual and friendly, or more formal and professional?
• What information do you want to immediately convey to your customers? You can use an automated greeting to set customer expectations for services you'll provide, agent availability, and possible wait times.
• How personalized do you want your first message to the customer to be?

Self-service

Deflecting tickets can benefit both your customers and agents by saving time and expediting the resolution of customer issues. There are a number of ways you can implement automated self-service options within your messaging workflow. Usually these will come at the beginning of the conversation, but they could be useful at other points as well.

• Suggest help center articles. The standard autoreply botevaluates new customer requests submitted through email or web forms for key concepts, identifies relevant content in your help center, and automatically replies to the customer request with links to the articles that are most likely to have the information the customer needs to solve their problem.
• Provide predefined answers to common questions. Identifying common questions and automating answers to them can provide customers with immediate answers.
• Pull data from external sources into conversations. Using API calls, you can automatically incorporate third-party data in response to customer requests. This is particularly helpful when customers need information that's updated frequently or unique to them.
• Ask whether a question is resolved. After providing one or more of the self-service options listed above, you can also prompt the customer to tell you whether the self-serve options resolved their issue. This way, they can close tickets before it gets to an agent.

Conversation workflow

Regardless of the complexity of your messaging workflow and bots, there will always be some customer support requests that need to be transferred to a live agent. When configuring messaging with any level of automation, you need to consider how and when the transfer to an agent occurs and establish rules for how the conversation is managed after the transfer.

Design your bot and automation

If you choose to incorporate a conversation bot into your messaging workflow, take time to map out the flow of the bot before you start implementing it. It will be easier to implement a single version all at once instead of potentially creating multiple versions if you use a phased approach.

We recommend using process mapping to create this plan, but you can use any tool you find helpful. The important thing is that you have a visual representation of every action the customer can take when interacting with your conversation bot, where those actions take them, and–most importantly–label these actions with the bot step type, feature, or messaging functionality that makes them possible.

After you create this plan, preserve it in a way that you and your team can refer back to.

Consider the handoff from bot to agent

The handoff to an agent can be a simple message letting the customer know they're being transferred and then adding the conversation to an agent's queue. If necessary, you can incorporate any of the following into the handoff interaction:

• Forms: Gather additional information the agent might need to solve the customer's issue.
• Estimated wait times: Set customer expectations appropriately.
• Notification options: Give the customer options for how they want to be contacted when an agent is available to assist them.

Decide how conversations are routed to agents and resolved

You also need to plan how the conversation is managed after the bots hand it off to an agent. Consider the following:

• How do agents receive messaging conversations? Are they assigned to groups and agents (push model) or do agents need to pick them up themselves from views (pull model)? You can choose from the following options:
  • Omnichannel routing provides Zendesk's most sophisticated routing logic and the ability to route tickets from multiple channels based on agent availability and capacity. On Professional plans and above, tickets can also be routed based on skills and priority. Omnichannel routing works out of the box for messaging tickets.
  • Notification routing broadcasts notifications to all agents for all relevant conversations. Agents can click Serve Request to start working on the message.
  • Assigned routing evenly distributes messaging conversations among online agents. Only one agent is notified for each incoming message at any given time.
• How are you going to notify customers when an agent enters the conversation?
• How are you going to manage the tickets generated by messaging conversations?

Next steps

Continue walking through the Messaging deployment guide. The next article in the series is Planning your staffing and operational requirements for messaging, which will help you consider your staffing needs for conversational messaging.