Zendesk AI Migration Readiness Guide: Preparing Your Support Data for AI

AI capabilities are one of the main reasons organizations migrate to Zendesk. But whether those features deliver value — or only create new problems — hinges on the quality of the data behind them.

How well Zendesk AI features perform once you're live is determined by ticket history and knowledge base content, along with the metadata connecting them. This guide covers what to check before you migrate, when to migrate, and how to get Zendesk AI working the way it's supposed to.

What Is Zendesk AI? 2026 Update

When asking "what is Zendesk AI?", you'll get a different answer today than you would have before the 2026 rollout, when each feature worked inside its own corner of support. Now Zendesk AI is one connected platform with Zendesk AI Agents, Copilot, Intelligent Triage, and AI-powered Knowledge running together across the entire support workflow, from the first customer message to ticket resolution.

AI Agents

Zendesk AI Agents are autonomous AI assistants designed to resolve customer requests without predefined conversation flows. You may still be researching Zendesk AI bot capabilities, but Zendesk now positions AI Agents as the successor to traditional support bots.

Instead of following scripted paths like traditional Zendesk AI chatbots, they analyze each request, determine the next best action, and resolve the issue once they have enough information and context to do so.

Zendesk Copilot

While AI Agents work directly with customers, Zendesk AI Copilot works alongside your support team. It summarizes conversations, suggests replies, recommends relevant knowledge base articles, and surfaces the next best action based on the current ticket.

Zendesk Knowledge

Zendesk Knowledge brings help center articles, policies, web pages, and other approved content into one connected knowledge system. AI Agents, Copilot, self-service, and support teams all reference the same information, keeping answers consistent across customer interactions.

Intelligent Triage

Every support request has to land in the right queue. Intelligent Triage automates that process by identifying customer intent, detecting sentiment and language, and using those signals to classify and prioritize incoming tickets.
Combined with tags, categories, custom fields, and priority levels, this cuts down on manual routing and helps support teams respond consistently as ticket volume grows.

Why AI Readiness Starts Before You Migrate

Every Zendesk migration moves data. The question is whether that data gives Zendesk AI what it needs to work effectively.

Your support data helps Zendesk AI

Your support data helps Zendesk AI understand how your support organization works. That includes:

• Historical tickets, which show how similar customer issues were handled and resolved.
• Knowledge base articles, which AI Agents and Copilot use to answer questions and recommend solutions.
• Tags, categories, custom fields, and routing rules, which help AI classify requests and recognize recurring patterns.
• Ticket metadata — priorities, statuses, and attributes attached to each ticket — which help AI summarize conversations and recommend the next action.

Clean, well-structured data gives Zendesk AI a stronger foundation.

The hidden risk of migrating with “dirty” data

Not every migration improves data quality. If duplicate records, inconsistent ticket fields, outdated knowledge base articles, or unused custom fields are moved into Zendesk unchanged, those same issues will affect how AI performs.

For example:

• Inconsistent categories and tags make it harder for Intelligent Triage to classify and route requests consistently.
• Outdated or duplicate knowledge base articles can lead AI Agents and Copilot to retrieve irrelevant information.
• Incomplete ticket metadata limits the information available for AI to summarize conversations or recommend the next action.
• Unused or inconsistent custom fields risk making reporting, routing, and AI-assisted workflows inaccurate.

That's the groundwork a full pre-migration audit builds on, which is exactly what we’ll cover in the next sections.

What Zendesk legacy AI deprecation means for your timeline

Pre-Migration AI Readiness Audit

A migration project is also a data quality project. Reviewing your knowledge base, ticket history, and metadata before moving to Zendesk means starting clean, not cleaning up after AI features are live.

Step 1: Audit your ticket taxonomy and tagging schema

Review 100–200 recent tickets for inconsistent tags, renamed categories, or abandoned custom fields. These are common signs your taxonomy needs cleanup.

By doing so, you’ll have a cleaner foundation for the Zendesk AI features you'll turn on after migration.

Step 2: Review your knowledge base for AI readability

Knowledge Copilot evaluates your knowledge base against three health metrics: Coverage, Freshness, and AI Readability. AI Readability measures how well your articles are structured for AI processing. Media-heavy pages, long unstructured articles, or content with excessive links are harder for AI Agents and generative search to use as content sources.

Step 3: Map metadata fields that power intelligent triage

List the metadata your support team relies on, including custom fields, priority levels, statuses, tags, and ticket forms. Then check how consistently those fields are populated across your existing tickets.

Intelligent Triage classifies incoming requests using customer intent, sentiment, and language, and routing rules depend on clean, consistent metadata. Preserving those fields during migration keeps your existing workflows running as expected once Zendesk is live.

Step 4: Identify content gaps before migration

Compare your knowledge base with your most common support requests. If customers regularly contact your team about a topic that isn't covered in your documentation, it's worth closing that gap before migrating.

Help Desk Migration supports this process with flexible data mapping and knowledge base migration that preserves supported article structure, formatting, attachments, language versions, and other content relationships. Internal article links can also be updated during migration, making it easier to validate your knowledge base before the final transfer.

What Data Should You Migrate for Zendesk AI?

What data do I need for Zendesk AI to work? There is no single answer to this question. That’s because different Zendesk AI features rely on different types of data. Below, we break down which records are typically worth preserving and which considerations to keep in mind.

Essential data to preserve

Here's what typically needs to survive the move, and why it matters for Zendesk AI:

Knowledge base considerations before migration

• Migrating a knowledge base is a technical move as much as a content one. Before you migrate, check whether:
• Internal links, images, and attachments carry over without breaking
• Articles pull from external sources such as Confluence or Google Drive that require separate handling
• Structural markup, including headings, lists, and tables, is preserved
• Translations remain linked to the correct article if you publish in multiple languages

Then review the content itself. Remove outdated or duplicate articles, split broad topics into focused pages, and replace image-only instructions with searchable text. Clear, well-structured content gives Zendesk AI customer service features more reliable information to work with before they're enabled.

Why field mapping matters for Zendesk AI features

Field mapping determines where your existing ticket fields, tags, categories, priorities, and other metadata end up after migration. Rather than dumping everything into generic fields, each source field is matched to its Zendesk equivalent, keeping information organized and existing workflows intact.

This is particularly important if your current help desk relies on custom fields, ticket tags, or platform-specific properties. These fields support routing rules, automations, and reporting, and provide additional context used across Zendesk AI agents features and agent workflows. They also help maintain a consistent Zendesk AI integration with the information your support team already uses.

Help Desk Migration maps standard fields automatically and lets you configure custom mappings whenever source and destination fields don't match. A Demo Migration lets you validate field mappings, preview how your data will appear in Zendesk, and adjust mappings before the Full Migration.

A two-step migration strategy

If you're adopting Zendesk AI as part of your migration, Help Desk Migration, with its AI-first approach, can split the process into two steps, validating your setup before transferring the rest of your data.

Post-Migration AI Activation

Migrating your support data into Zendesk gets the technical work done. What's left is introducing Zendesk advanced AI capabilities one at a time, confirming each one behaves as expected before expanding the rollout.

Phase 1: Validating migrated support data

Before enabling AI, review the information your support team relies on every day. Ticket history, knowledge base articles, custom fields, tags, organizations, users, and attachments should all be present and mapped correctly. Confirm that automations, triggers, views, and routing rules behave the way they did in your previous help desk. Use this Zendesk setup checklist as a reference.

Phase 2: Turning on AI Agents

To configure Zendesk AI Agents, connect them to the knowledge sources they should draw from, then assign each a channel. Every agent is built for either messaging or email, so a team running both channels ends up building two agents, not one.

Once an agent is live, watch its resolution rate alongside how often conversations escalate to a human. Review escalated conversations to determine whether they point to gaps in your knowledge base, routing configuration, or supported workflows before expanding AI to additional channels.

Phase 3: Enabling Copilot

Zendesk AI Copilot works from your migrated ticket history and knowledge base once both are available in Agent Workspace. Don’t rely on dashboards and metrics alone. Have several support agents use it inside their normal queue, and compare their handling time against your baseline. Review their feedback before expanding adoption.

Phase 4: Turning on Intelligent Triage

Start with a limited set of categories or queues before rolling it out across your whole support operation.

Frequent manual corrections usually signal that categories, routing rules, or metadata need refining before expanding Intelligent Triage further.

Phase 5: Expanding to Voice AI agents

Voice AI agents run on top of Zendesk Voice (Talk) and are currently available through Zendesk's Early Access Program (EAP). Confirm that your account is eligible before making them part of your rollout plan.

If it is, connect Voice AI to your phone channels and test routing, language selection, and escalation scenarios before directing production traffic through it.

Testing before you go live

Before making Zendesk AI and automation features available to customers, test conversation flows thoroughly across the scenarios each channel is most likely to handle.
The specifics vary by channel, so it’s best to check them one at a time:

Pre-activation review by channel:

Zendesk AI Features That Depend on Data Quality

Different Zendesk AI features rely on different parts of your support data. The impact of poor data quality also varies. Some features lose context, others produce less consistent results, while some depend primarily on configuration rather than migrated historical records.

Intelligent Triage: Routing Becomes Less Consistent

Intelligent Triage classifies requests using tags, categories, custom fields, and routing metadata. Duplicate tags, overlapping categories, or inconsistent field values make routing rules harder to maintain and reduce classification consistency.

AI Ticket Summaries: Missing Context Carries Forward

AI Ticket Summaries reflect the information available in each conversation. Without comment threads, internal notes, and ticket metadata, agents lose the context behind each customer interaction.

Knowledge Copilot: Suggestions reflect your knowledge base

Knowledge Copilot retrieves articles from your knowledge base. When multiple articles answer the same question differently, or documentation no longer matches the product, agents spend more time reviewing suggested content before using it.

AI Agents: Documentation gaps increase escalations

AI Agents handle customer questions using your knowledge base and configured workflows. When documentation doesn't cover a request or contains conflicting information, more conversations are transferred to support representatives instead of being resolved automatically.

Voice AI and writing tools: Configuration has a greater impact

Voice AI depends primarily on phone configuration, routing rules, and connected channels. Writing Tools work from the current conversation, so they rely less on migrated historical data than features built around your knowledge base or ticket history.

Zendesk AI Pricing and Tier Breakdown in 2026

Zendesk AI pricing splits into three separate decisions: which Suite tier you choose, which add-ons you layer on top, and how many Automated Resolutions your AI Agents use beyond what's included. Migrating onto Zendesk without mapping out all three usually means budgeting for the subscription alone, then getting a surprise bill once usage-based costs kick in.

What's included in Zendesk Suite when no add-on is required

Zendesk Suite includes built-in AI capabilities, so teams can start using core AI features without purchasing a separate AI add-on.

A Zendesk AI-ready migration preserves your knowledge base, ticket history, users, custom fields, tags, and routing configuration so your support team can continue using AI capabilities alongside existing workflows after the migration.

What the Copilot add-on covers

For organizations that need more advanced AI-assisted workflows, Zendesk offers the Copilot add-on for Suite and Support Professional plans and above. It extends Zendesk Suite’s built-in AI capabilities with additional AI assistance for support teams.

Zendesk AI Copilot currently includes these features:

If you're researching Zendesk Advanced AI add-on pricing, note that older documentation may still reference “AI Agents – Advanced.” Zendesk is now retiring the Essential and Advanced AI Agents packaging in favor of a unified AI Agents offering, while Copilot remains a separate add-on for agent-assist features.

Outcome-based pricing: automated resolutions explained

Zendesk AI pricing combines your subscription plan with usage-based pricing for AI Agents. An Automated Resolution is a customer request that an AI Agent resolves from start to finish without handing the conversation over to a human agent.

Each plan includes a monthly allowance before additional Automated Resolutions are billed.

Four factors determine the total AI cost: your subscription tier, the number of agent seats, any optional add-ons such as Copilot, and the number of Automated Resolutions that exceed your monthly allowance.

For example, a team with 100 Suite Professional agents has 1,000 Automated Resolutions included in its monthly plan. If AI resolves 1,500 customer requests, the additional 500 Automated Resolutions are billed separately at the rate specified in its Zendesk contract.

ROI framing: Migration cost vs. AI deflection savings

A Zendesk migration is a one-time investment, while Zendesk AI agents deliver ongoing savings by reducing the number of assisted support interactions over time.

Illustrative example:

• Gartner reported a median assisted contact cost of $13.50. For this scenario, imagine your support team handles 20,000 assisted contacts per month.
• If Zendesk AI agents automate a conservative 20% of those interactions, around 4,000 contacts move out of the assisted queue.
• At Gartner's benchmark, that’s roughly $54,000 in assisted service costs avoided each month.
• For a typical mid-market project using Help Desk Migration, direct migration costs can range from $25K to $50K, depending on data volume, customizations, and migration requirements.

Comparing those figures against your Zendesk AI pricing and implementation costs gives you a practical way to estimate potential ROI and break-even point.
Looking at migration costs alongside expected AI savings gives you a clearer picture of the long-term financial impact.

FAQs About Zendesk AI Migration

Written by

Taras Velychko

Product Manager, Help Desk Migration

Taras is leading product strategy at Help Desk Migration, specializing in UX/UI, business intelligence, and vendor partnerships. His data-driven approach has improved conversion rates by 50% while keeping the product simple and customer-focused.