Zendesk AI Agents in 2 minutes or less
Zendesk AI Agents in 2 minutes or less
Puzzled by Zendesk's AI or AI in general?
Let's break AI Agents down in a 2 minute read so you can get on with implementing some really useful features on your Zendesk support platform.
📒 Important Note
Zendesk's Advanced AI offering is not in the context of this article. It is only about AI Agents.
You've probably heard a lot about AI lately and you've possibly been bombarded by Zendesk's AI offerings since their Relate conference this year. If you struggle to grasp it all, don't fear; I'll try to give you only what you need to know here about the very basics.
It barely touches the surface of AI capabilities in general but might just help you increase your knowledge a nugget or two and get you on the way to something a little, well, intelligent.
Zendesk's AI Agent is an advanced, AI-driven virtual assistant designed to enhance customer support by handling complex interactions and automating a variety of tasks. In short, you can use it to deflect tickets thus reducing your human agent's workload. It will increase the scalability of your support operation if implemented well.
Zendesk AI Agent Autoreplies are really fundamental to ticket deflection
Autoreplies
This is probably the most fundamental of AI Agent features you can implement and honestly, every Zendesk instance should have it enabled. Autoreplies leverage your help centre articles as well as intents to formulate suggested answers to the customer's request.
In the context of email and web channels only, there are 2 ways to implement these:
Recommended Articles: Web Form Request Submission
Recommended Articles: Email Request Submission
Let’s dive into each of these a bit more.
Recommended Articles: Web Form Request Submission
This dialogue shows right after the end-user submits a request.
When enabled the above dialogue shows right after an end-user submits a request via a request form in your help centre. It suggests your help centre articles based on what the end-user has typed into the summary and description fields.
If you have any content in your help centre and use either web forms and/or email channels then you should definitely implement this. If you don't have a knowledge base of content in your help centre, you should start building one (then come back)!
If you want to give it a try, Zendesk has fairly clear instructions right here.
Recommended Articles: Email Request Submission
This dynamic content shows how we include recommended articles in the body of the email.
When configured via a trigger, this will send up to 3 suggested articles from your help centre based on the content of the email. In the above example, I am using dynamic content to implement autoreply in a multi-lingual environment.
If you have any content in your help centre and use either web forms and/or email channels then you should implement this.
You can find a fairly straightforward set of instructions on how to implement this here on Zendesk's Help Centre.
Beyond Auto-replies
Practically every other AI feature available in AI Agents is available only on the Messaging channel by adding a bot. A bot sits on top of your messaging channel.
Here's a run down in plain English.
Generative AI
You can immediately smarten up your Messaging by enabling Gen AI for the bot when you're setting it up.
Enable this to enable Generative AI
This means your bot is going to generate responses by leveraging your help centre content. If this is all you do, then you've already made progress.
Answers
This can be confusing so let's try to make it clear. Answers are the little pre-written conversation prompts you see in a chatbot.
Here's an example.
These are Answers
In this test, the bot could not find anything in the help centre so suggested two Answers which I have defined. Each of the two answers has a flow which I can define.
In the case of the Visit the Help Centre flow, it simply links the user to the Help Centre.
Intents
I think of intents as use cases or specific customer situations, for example, requesting a refund or account verification status.
To illustrate how they work, take a look at this example.
A sample Intent set to Generate a reply using AI leveraging the Help Centre content
Circled in blue is the intent called Account verification service. This intent is set to Generate a reply from our Help Centre. As we don't have any content remotely close to mentioning account verification services, the bot returns the default response when it can't find anything which is to give the two Answers:
Visit the Help Centre
Talk to an agent.
A sample Intent set to Talk to an agentAnswer instead
Now I've changed the intent's Answer to Talk to an agent instead of Generate a reply. In green, you can see the result is that the bot begins the Talk to an agent workflow.
When you create a bot for your Messaging channel you will need to build the flows for your business. Zendesk comes with about a dozen templates for common use cases.
This just touches the surface of Answers & Intents but should be enough to get a basic bot up & running in minutes.
Advanced AI
Beyond this, there is a host of more advanced topics to leverage from Zendesk's built-in AI Agents such as integrations and API calls but that's for another post.
For now, I'll leave you with a word of caution about pricing. Zendesk is introducing per-resolution pricing in August 2024. It means there will be a small charge per resolution considered to be due to AI Agents.
You should make yourself aware of this and know how to monitor your usage by heading into the Account > Usage > AI and bots usage screen.