What exactly is Zendesk Agent Copilot?
Published
Read time
6 minutes
Article summary
Zendesk Agent Copilot is a proactive AI assistant built into the service agent's Agent Workspace. Where AI agents handle routine queries end to end, Agent Copilot guides the human agent through more complex cases with suggested replies, writing tools and back-end system integrations. It pays to roll it out gradually rather than forcing it on the whole team at once.
Which one to choose — an AI agent or Agent Copilot?
Most of a support agent's day is spent on two kinds of situations: repetitive routine questions, or complex problems where the agent has to dig up information across different systems and follow multi-step instructions. AI helps resolve tickets in both cases, but in different ways:
Zendesk AI agents: These handle the customer interaction from start to finish without a human (so-called _one-touch_ tickets). They're a great fit for high-volume routines such as order tracking, password resets or common FAQs.
Zendesk Agent Copilot: This is designed for situations that call for human judgement, empathy or complex problem-solving (so-called _multi-touch_ tickets). Agent Copilot doesn't talk to the customer directly — it works in the background inside the agent's workspace: it pulls the information together, suggests a reply and guides the agent step by step through the right process.
Rule of thumb: If a ticket only needs a single API lookup or a straightforward answer, let a Zendesk AI agent handle it. If it needs judgement, several messages back and forth, or manual clicking around back-end systems, turn it into a Procedure for Agent Copilot.
Thanks to this division of labour, agents can focus on the actual customer interaction while AI handles the manual investigation and clicking in the background. But what exactly does Zendesk Agent Copilot include, and how does it make an agent's day easier?
What Zendesk Agent Copilot includes and how it works
Zendesk Agent Copilot works directly in the _Agent Workspace_, supporting agents in real time. Its capabilities fall into three main areas:
Intelligent triage
Before a single agent even sees a new ticket, AI analyses its content in the background.
Intent: Automatically identifies what the contact is about.
Entities: Extracts the key details from the message — such as order numbers, email addresses or products — and can attach them straight to the ticket's data fields.
Sentiment and language: Detects the customer's mood (e.g. angry) and the language they're using.
What's the benefit of intelligent triage? Manual tagging becomes a thing of the past. Tickets can be routed automatically to the right experts or prioritised by language and sentiment. If it's a simple one-touch query, an autonomous AI agent can handle it directly, while harder cases reach the agent pre-chewed.
Auto Assist
Once a ticket lands with an agent, Auto Assist steps in with real-time guidance and automation:
Procedures: Agent Copilot suggests step-by-step procedures based on instructions written in plain language. It understands how the conversation is progressing and guides the agent on what to ask or do next. If the customer already gave their order number in the first message, Agent Copilot recognises it and won't ask the agent to request it again.
Actions: Auto Assist can run system actions directly from within the ticket, with the agent's approval (_Action Flows_). For example, it can issue a refund in an e-commerce or billing system (such as Shopify or Stripe) or fetch customer data from an external CRM with a single click.
Suggested Replies: Generates ready-made drafts based on Help Center articles, macros and previously proven answers.
This makes an AI-ready knowledge base a critical foundation — if the Help Center content is out of date, Agent Copilot's suggestions can be wrong too. Read here how to build an AI-ready knowledge base.
Ticket Summaries: Turns long conversation threads into a quick summary so the agent can get up to speed in seconds without reading the whole history. Especially handy when a ticket is handed from one agent to another.
Enhance writing: Helps the agent adjust the tone of a reply, expand short notes into a full email, or simplify text into clear instructions on the fly.
Similar Tickets: Finds and surfaces comparable past resolutions so the agent can reuse answers that already worked.
Merge Suggestions: Detects when the same customer has opened multiple tickets across different channels and suggests merging them to avoid duplicate work.
What's the benefit of Auto Assist? _Agents no longer have to jump between systems in the middle of writing a chat or email. Procedures standardise service quality and are especially helpful in getting new hires up to speed faster. Integrations and suggested replies cut out needless typing and clicking, which noticeably shortens handling times._
Analytics
Measurement and continuous optimisation decide whether you actually get value out of Zendesk Agent Copilot. Zendesk provides a ready-made Agent Productivity Dashboard for this, measuring three key areas:
Procedure performance (Auto Assist): The dashboard shows how many tickets each procedure has been applied to and what its acceptance rate is.
Use of AI suggestions: The system measures how agents respond to Agent Copilot's suggestions. This tracks, for example, the use of similar past resolutions, merge suggestions and macros.
Use of generative writing tools (AI agent tools): The dashboard records data on how often and how agents use AI to polish replies — for example expanding text, changing tone or summarising long threads.
What's the benefit of analytics? Improving efficiency is based on data, not guesswork. The dashboard tells you directly which procedures work and which need tuning, and how much time the AI tools save your team. That gives you direct visibility into the return on your investment.
How to roll out Zendesk Agent Copilot
At Zennius we recommend approaching a Zendesk Agent Copilot rollout in three steps:
Build the foundation (Intelligent Triage)
All automation needs good data. Turn on Intelligent Triage first and analyse your team's top 5 contact reasons. If they can be solved in one go (one-touch), route them to an AI agent. If they need judgement (multi-touch), prepare them for Agent Copilot.
Start piloting (Auto Assist)
Don't switch Auto Assist on for the whole team or every ticket at once. Pick a test group and create a trigger that adds an agent_copilot_enabled tag only for certain limited situations (or add the tag manually to individual test tickets at first). Start by writing the 1–2 most common and clearest procedures based on your Intelligent Triage data. Once the test group notices they're saving time and the procedures have been fine-tuned, you can gradually expand to new topics and the whole team.
Monitor and improve
The real game begins after rollout. Track the analytics — such as how often agents actually accept Agent Copilot's suggestions — and create new procedures based on the accumulated data and the AI's own recommendations. Little by little, ticket resolution speeds up and more time is freed up for the harder contacts.
Need a hand with the rollout?
If rolling out Zendesk Copilot is on your agenda, or you'd like a sparring partner to take your support automation to the next level, book a short meeting from our calendar.
Zennius is an AI-expert-certified partner. With experience from over 100 Zendesk AI rollouts, we know these features below the surface and make sure you get the best possible return on your investment.
