Verified AI summary ◀▼
Auto assist is an AI-powered tool that helps agents handle tickets by suggesting replies and actions based on procedures, help center articles, or similar solved tickets. It reduces repetitive tasks, ensures consistency, and speeds up ticket resolution. Agents review and approve suggestions, which are tailored to the conversation's tone. Customers remain unaware of the AI's involvement during interactions.
Auto assist is an AI-powered assistant for your agents. Using large language model (LLM) technology, auto assist understands the contents of submitted tickets and makes suggestions to your agents on how to solve them.
With auto assist, agents spend less time on repetitive tickets, solve tickets in a more consistent way, and ultimately close more tickets.
Watch the demo video below to see auto assist in action, or read on for more information about auto assist. To get started, see Turning on and configuring auto assist.
Zendesk agent copilot featuring auto assist (4:43)
Auto assist makes suggestions to your agents based on specific procedures you define, generic information from its LLM training, your public help center articles, or similar solved tickets. These suggestions can be replies that the agent should send to the customer or actions that the agent should take.
If you've defined direct instructions for agents in your procedures, auto assist suggests the instructions to agents when they need them based on the flow of their conversation. Instructions are steps you've written that agents should perform manually. For example, you might write instructions for checking your internal stock management system if auto assist can't access it.
When using auto assist, agents don’t have to compose replies or execute actions themselves, except when following instructions you've defined. Instead, agents review and approve the replies or actions suggested by auto assist. Replies suggested by auto assist are automatically adapted to the conversation's tone and writing style so that agents can reduce the time spent on editing responses. When making suggestions, auto assist takes into account all public ticket comments in the conversation, including previously generated comments approved by the agent. Suggestions are made in response to end-user comments only, though, not agent comments.
Auto assist suggestions appear in tickets in the Agent Workspace interface in place of the composer. Any replies or actions suggested by auto assist require an agent’s approval before being sent to the customer or performed, and are sent or performed under the agent’s name. The customer isn’t aware of auto assist during their interactions with the human agent.
To learn more about the agent experience with auto assist, see Using auto assist to solve tickets.
Understanding how auto assist makes suggestions
Auto assist can suggest replies and actions based on the information provided in procedures and its generic LLM training information. It can also suggest replies based on your public help center articles, or similar solved tickets, when a relevant procedure or article isn't available. It can't suggest replies or actions based on other sources.
Auto assist can suggest replies on any topic based on the following information:
- Procedure content you've created, which is essential to achieving the best
performance. A procedure is a set of written steps that describe how a given
customer problem should be solved. You can think of defining procedures as
similar to explaining to a new agent how a specific problem should be solved
within your company.
Auto assist uses this procedure content, along with its built-in LLM training, to generate suggested replies or actions, which it then presents to agents as they work on tickets in the Agent Workspace.
Your procedure content can also include direct agent instructions that you've defined. These types of instructions are steps that agents should follow and perform manually.
- If a relevant procedure isn't found, then auto assist uses your public help center articles to generate suggested replies. Only articles with a brand and language matching the ticket brand and language are used for reply generation. For example, if the brand on a ticket is Obscura, then only articles with the Obscura brand are considered.
- If a relevant procedure or help center article isn’t available, then auto assist uses similar solved tickets to generate suggested replies.
Examples where auto assist can suggest replies include:
- Advising the customer on buying a product or service sold by your company
- Explaining a business policy to the customer
- Helping the customer troubleshoot a problem with a product
Examples of procedures where auto assist can suggest replies and actions include:
- Checking the status of a current order in Shopify, or canceling and refunding specific items or whole orders in Shopify
- Querying and modifying your own internal business systems or performing third-party actions via API
- Telling an agent to check your internal stock management system and adding a refunded item as 'back in stock'
- Instructing an agent on how to create and validate a new user account in your company's human resources system