A real step-by-step guide to leaving Zendesk. What to export, what doesn't port, what the API limits look like, and how to avoid downtime.
TL;DR:
- Zendesk Suite runs $55 to $169 per agent per month, with AI add-ons billed separately on top. A 10-agent team on Suite Professional with AI pays roughly $1,800 to $2,500 per month before usage.
- The hard parts of leaving Zendesk are not exporting articles. They are macros, triggers, automations, custom ticket fields, and the routing rules that touch every conversation.
- The Zendesk Help Center API and Incremental Export API can pull what you need. The Incremental Tickets endpoint pages forward by time and is the right tool for multi-year archives.
- Run Zendesk and the new platform in parallel for two to four weeks on a single channel before full cutover.
- Skip the migration if you depend heavily on Zendesk Sunshine, multi-brand setups, or complex SLA agreements that the new platform cannot replicate.
Zendesk is a mature product. It is also expensive and slow to ship the AI features that newer platforms ship every quarter. By 2026, more teams are looking at the line item and asking whether the value still tracks the price.
This is the playbook for teams that have decided to migrate. It is not a sales pitch. Zendesk does things very few competitors match, and you should know what you are giving up.
Three patterns drive most migrations.
Zendesk Suite has four tiers per Zendesk's public pricing page: Team ($55), Growth ($89), Professional ($115), and Enterprise ($169), all per agent per month on annual billing. Suite Enterprise Plus is custom and typically lands above $200 per agent.
That is before AI. Zendesk's Advanced AI add-on is roughly $50 per agent per month on top of Suite. Workforce Engagement, Workforce Management, and Quality Assurance are each separate add-ons priced per agent. A 10-agent team on Suite Professional plus Advanced AI is around $1,650 per month before usage. Add WFM and QA and it climbs past $2,500.
The friction is not just the headline number. It is how many separate line items the bill has and how hard it is to predict next year's cost as the team grows.
Zendesk's AI Agents (formerly Ultimate, acquired in 2024) are billed per AI resolution, similar to Intercom Fin. The exact rate depends on contract size and is not published. Teams running both Zendesk Suite and AI Agents at scale report combined bills above $5,000 per month for moderate teams. The pattern is the same as Fin: the better the AI works, the more you pay.
This is a fair criticism, not a knock. Zendesk's strength was being a stable, dependable platform for 15 years. That strength is also why their AI features have shipped slower than newer entrants. Teams that want native model choice, modern RAG over their help center, or self-serve AI deployment without a sales cycle find Zendesk's AI roadmap frustrating.
This section saves migrations. Skip it and you will lose data you did not realize was load-bearing.
Easiest piece. Zendesk's Help Center API returns each article as HTML with metadata. Loop through paginated calls and dump to disk.
Three gotchas:
zdusercontent.com and zendesk.com subdomains). You need to download and re-host them, otherwise they break when the Zendesk subscription ends.Zendesk also provides a built-in Help Center export under Guide admin settings. The built-in export covers HTML content but does not include metadata like article labels, user segments, or vote counts. Use the API for full fidelity.
This is the painful one. Zendesk's macros (canned responses with actions), triggers (event-based rules), and automations (time-based rules) are stored in a proprietary format that does not port to any other platform.
Macros export as JSON via the Macros API, but the actions inside macros (set ticket status, assign to group, add tags) use Zendesk-specific concepts that need to be rebuilt in the new platform.
Triggers and automations export but cannot be replayed. Plan to rebuild from scratch.
Practical approach: pull the JSON for documentation, then sit down with the team and decide which macros are actually used (most teams find 60 to 80 percent of macros have not fired in the last 90 days). Rebuild the ones that matter.
Custom fields export via the Ticket Fields API with their type, options, and visibility settings. The new platform's custom field model will be different (different field types, different validation), so plan to map each one manually.
Critical detail: any field referenced by a macro, trigger, or automation needs to exist in the new platform before you migrate workflows. Pull custom fields before macros.
The Incremental Tickets endpoint pages forward by start_time and is the right tool for multi-year archives. Each request returns up to 1,000 tickets plus a cursor for the next page.
Per Zendesk's docs, the incremental endpoints have rate limits that vary by plan: typically 10 requests per minute on Professional and below, 200 requests per minute on Enterprise. A multi-year archive of 100,000 tickets at 1,000 tickets per request takes 100 paginated calls. Plan a few hours of API work, not seconds.
The browser-based export under Admin Center caps at 50,000 rows per CSV file and is sent via email, which can take an hour for large datasets. Use the API for serious volume.
Users and organizations export via the Users API and Organizations API. Both export cleanly. Custom user fields and organization fields export with the records.
Views (saved ticket queries) export as JSON but are tied to Zendesk's query language. Rebuild in the new platform.
SLAs are Zendesk Enterprise only and are configured per ticket priority and customer segment. If you have complex SLA agreements with enterprise customers, document each one and verify the new platform can replicate the policy structure. Most AI-first platforms have lighter SLA models.
Routing rules (round-robin, group routing, skills-based routing) need to be recreated by hand.
This sequence assumes you have picked the destination platform and have admin access to Zendesk.
Two weeks before any cutover, run a usage audit:
The output is a short list of what actually matters. Most teams discover 50 to 70 percent of their Zendesk configuration is dead weight from a previous CX leader and does not need to migrate.
The help center powers your AI agent. Export it first because the rest of the migration depends on AI being live in the new platform.
Pull articles, sections, and categories via the Help Center API. Download all embedded images and re-host on your CDN or inside the new platform. Validate that internal links between articles still work after re-hosting.
For multi-brand setups, export each brand's help center separately and keep them separate in the new platform.
Zendesk organizes tickets by tags, custom fields, and routing rules. AI-first platforms organize by intents and topics. The mapping is not one-to-one.
Pull the top 50 ticket tags by volume from the last 90 days. Group them into 10 to 20 intent categories. Each category becomes a topic for the AI agent to handle (refunds, account access, billing questions, feature requests, bug reports).
This step is where most migrations go wrong. Skipping the mapping means the new AI agent's reporting will not align with your historical Zendesk reporting, and the CX team will lose confidence.
Pick the lowest-stakes channel first. For most teams, that is the website chat widget. Email and phone come later.
Deploy the new AI agent on the website only. Keep Zendesk handling email and any other channels. Run for one to two weeks while you tune the AI agent's responses, escalation rules, and tone.
Do not cut over all channels on day one. Channel-by-channel migration is the difference between a quiet rollout and a CX team revolt.
Measure on the new platform:
Compare to the same metrics on Zendesk for the same time window. If the new platform is within 10 percent of Zendesk's metrics, you are ready to expand. If it is meaningfully worse, find out why before expanding.
This is also when you discover edge cases your audit missed: the customer who always emails about a specific integration, the refund flow that requires manual approval, the VIP routing rule no one documented.
Migrate remaining channels (email, social, SMS). Keep the Zendesk account active and paid for 30 days after cutover. Do not cancel.
The 30-day standby covers:
After 30 days of clean operation on the new platform, cancel Zendesk.
| Zendesk feature | How to replicate in an AI-first platform |
|---|---|
| Help Center | Crawl the URL or import HTML; serve as RAG source |
| Macros | Saved replies and AI quick-actions; rebuild action steps |
| Triggers | Event-based webhooks or workflow rules |
| Automations | Scheduled rules or time-based workflows |
| Views | Custom inbox filters or saved searches |
| Custom ticket fields | Custom conversation properties |
| Tags | Tags (most platforms support them directly) |
| SLAs | Available on enterprise tiers, with simpler policy structure |
| Round-robin routing | Auto-assignment by group or skill |
| Multi-brand | Multiple workspaces or multiple deployments |
| Sunshine custom objects | External database plus webhook lookup |
| Talk (voice) | Pair with a voice CCaaS (Aircall, Dialpad); not native in most AI-first tools |
| Guide knowledge base | Native knowledge base with AI search |
| Side Conversations | Internal notes and team chat |
| Triggers calling external systems | Webhooks and tool calling |
Three patterns kill migrations. Avoid these.
The AI agent is only as good as the content it reads. If your Zendesk help center has 800 articles, 300 of which are outdated, the new AI agent will give wrong answers from day one.
Spend two weeks auditing content before cutover. Archive anything that has not been viewed in 12 months. Update articles that reference deprecated features. Consolidate duplicates.
Keep Zendesk paid and live for 30 days minimum after cutover. The temptation to cancel the day you launch the new platform is real because the bill is large. Resist it. The cost of one month of Zendesk is much smaller than the cost of a botched migration with no rollback.
If you have signed SLA agreements with enterprise customers in contracts, verify the new platform can meet them before you cut over. Most AI-first platforms have lighter SLA models than Zendesk Enterprise. Some do not support time-to-first-response SLAs measured outside business hours. Read the fine print.
Honest disclaimers. Stay on Zendesk if any of these apply:
If three or more apply, the migration cost likely exceeds the savings. Stay and optimize Zendesk instead.
For a 5 to 20 agent team with one brand and a clean help center, plan four to six weeks from kickoff to full cutover. That includes two weeks of audit and export, two weeks of parallel running, and a two-week tuning window. Larger teams with complex automations, multi-brand setups, or Sunshine integrations should plan three to six months.
Mostly yes, with caveats. Articles, users, organizations, tickets, custom fields, and macros export cleanly via documented APIs. Triggers, automations, views, SLAs, and routing rules export as JSON but the actions inside them are Zendesk-specific and need to be rebuilt. Sunshine custom objects export but the data model is not portable to most other platforms.
For a team under 10 agents that does not need complex ticketing, a combination of a lightweight helpdesk and a standalone AI agent often comes in under $500 per month total. Help Scout starts at $25 per user per month with simpler workflows than Zendesk. Pair with Chatsy at $140 to $475 per month for AI and the combined bill is roughly one-fifth of a Zendesk Suite Professional plus Advanced AI setup at the same headcount.
Probably better, depending on the platform. Zendesk's AI Agents (acquired from Ultimate in 2024) are competent but were built on an older RAG architecture. Newer platforms ship model upgrades faster and let you pick between GPT-5.2, Claude Opus 4.6, and Gemini 3 Pro. Honest answer: do a two-week parallel test on a single channel before committing.
Zendesk is not a bad product. For some teams it is still the right answer. If you have complex multi-brand workflows, deep Sunshine integrations, or enterprise SLA contracts, stay.
If you are leaving because the bill scaled faster than the value, the migration is doable. Start with the audit. Export the help center. Run parallel on one channel. Cut over channel by channel. Keep Zendesk in standby for 30 days.
Want a place to land that does not require a sales call to evaluate? Start with Chatsy. Free plan to test, transparent pricing on the pricing page, self-serve setup, and a mailbox plus AI agent plus knowledge base in one product. You can be live this afternoon.
Intercom's Fin charges $0.99 per AI resolution on top of seats. Chatsy is flat. Honest comparison with real cost math at 500, 1,500, and 5,000 resolutions.