Welcome to Zuzana

What Zuzana is

Zuzana is a helpdesk with an AI agent working inside it. Your channels come into one inbox, your team works the tickets there, and the AI works the same tickets alongside them: reading the message, looking things up in your systems, and writing the reply.

What makes it yours is not a model, it is what you teach it. You describe how your business handles customer requests, connect the systems that hold your data, and decide how much the AI is allowed to send on its own. Everything below is a way of doing one of those three things.

The pieces you will actually touch

The product has a lot of screens, but almost all of your time goes on a handful of concepts.

• Knowledge sources 📚 The facts the AI answers from: your help articles, your product catalogue, your website. See Knowledge Sources and Articles.
• Policies 📋 Short, reusable rules for how a type of question is handled. How you word a delivery estimate, when you offer a goodwill discount. See Policies.
• Workflows 🔀 Step-by-step procedures for the involved cases: cancelling an order, handling a return, chasing a missing parcel. Steps, lookups and decisions rather than just wording. See Workflows.
• Tools 🔌 The connections that let the AI act instead of only talk: order lookups, refunds, shipping data, your own systems.
• Communication style 🎙 How every reply should sound, per market. See Communication Style.
• Markets 🌍 One team can run several storefronts, languages and brands, each with its own tone, links and hours. See Markets.

How a reply comes together

Every incoming message runs the same route. Knowing the order helps you work out where to make a change when the AI does something you did not expect.

1. It reads the request in the context of the whole conversation, not just the last message.
2. It searches your knowledge for relevant articles, products and pages.
3. It matches a policy or workflow, if one applies. A policy shapes a single answer; a workflow runs a procedure.
4. It calls tools when it needs the customer's actual order or account rather than a general answer.
5. It writes the reply in the customer's language and in your tone for that market.
6. It sends or holds, depending on the response method you have chosen.

You decide how much goes out on its own

Nothing forces you to hand over the keys on day one. The response method decides what happens with a finished reply: held as a draft for an agent to review and send, sent automatically while the ticket stays open, or sent and closed. You can set it differently for different situations and tighten or loosen it as your confidence grows. See Response Methods.

Note: Most teams start in draft mode on everything, watch the drafts for a week, and then automate the question types where the AI is consistently right. That is a good instinct. It is much easier to open up than to win back trust.

Your team in the helpdesk

People are organised around the work rather than the other way round.

• Users 👥 Invite colleagues by email and pick a role. See Inviting your team.
• Squads 🧩 Group agents so the right work reaches the right people. See Squads.
• Capabilities 🎚 Control which markets, channels and topics a person actually sees. See Capabilities.

Day to day, the helpdesk gives your agents Macros for wording they reuse, Tags for organising tickets, Snooze for bringing a case back later, and Side Conversations for when the answer has to come from a supplier or a carrier.

Where to start

Roughly in this order. Each step is worth doing properly before the next one, because the AI is only ever as good as what you have given it.

1. Get the channels in. Set up your sending domain, forward your support mailbox, and put the chat widget on your site. See How to setup your domain and the forwarding guides for Google Workspace and Microsoft 365.
2. Bring your history. Importing your existing tickets from Zendesk, Gorgias or Dixa gives you something real to test against from day one.
3. Feed the knowledge. Connect your product catalogue and website, and write articles for the questions you answer most.
4. Write your first policies. Start with the three questions that fill your inbox.
5. Automate one scenario end to end. Pick the most common one that involves a lookup, and build it as a workflow.
6. Invite your team and set up squads and capabilities so everyone sees their own work.

Note: You are not alone in this. If something in here does not match what you see on your screen, or you are unsure what to do next, ask us. We would rather set it up with you than have you guess.