What is new in Zoho Desk 2026?
Zoho Desk’s 2026 release covers more than sixty changes. These are the ones that change how a service team works.
| Feature | What it does |
|---|---|
| Blended agents | AI agents that work with human staff, built from the Agent store with roles, permissions and department access; hand over to a person with context; can be paused for approval |
| Automation Studio | One workflow built from Action, If, Fork and Delay blocks, for example to handle refunds end to end |
| Data Privacy Control | Detects and anonymises personal data before AI processing, then restores values in Desk; rules per department; playground for testing |
| Masking and redaction | Masking within ticket threads with permission-based temporary viewing; redaction removes data permanently while keeping ticket history |
| Azure-hosted OpenAI models | Your own Azure OpenAI models for summaries, insights, writing and reply assistance |
| Deluge script generation | Generates Deluge custom functions from a prompt, for review and testing before saving |
| Multiple sandboxes | Up to three sandboxes so teams test and deploy independently |
| Channels | WhatsApp voice calling, Arattai Business as a messaging channel, payments inside guided conversations |
| llms.txt for the help centre | A Markdown file of help-centre content that AI tools can read |
Why does it matter for service teams?
For service heads at distributors and D2C brands. A large share of tickets ask for order status, stock or a document. An agent that answers those and passes refunds or credits to a person cuts queue time without giving up control.
For compliance and data protection officers. Anonymising personal data before it reaches a model, and redacting it permanently where needed, supports the reasonable security safeguards that section 8 of the DPDP Act, 2023 expects of a Data Fiduciary.
For IT. Three sandboxes mean the service team, developers and a partner can test changes in parallel without touching live tickets.
How do you roll out AI agents in Zoho Desk?
- Pick three intents. For example: “where is my order”, “send my invoice copy”, “change my delivery address”. Measure their share of tickets for two weeks.
- Switch on Data Privacy Control first. Define masking rules per department and test them in the playground with real ticket samples.
- Build one Automation Studio flow. Route the three intents, with an If block for high-value customers and a Delay block for follow-ups.
- Pilot a blended agent in a sandbox. Give it read access to order and invoice data only. Any refund, credit note or address change goes to a person.
- Go live on one channel, such as WhatsApp, and review transcripts weekly for a month.
- Publish llms.txt for your help centre once articles are current, so AI tools read approved answers.
What are the risks and limits?
- Confirm edition availability and any model-usage charges for your account before rollout.
- An agent that can update customer accounts can make wrong updates. Keep write actions behind human approval until error rates are known.
- Generated Deluge scripts need code review and testing before they go live.
- Masking rules are only as good as their patterns. Test with Indian identifiers such as PAN, Aadhaar, GSTIN and phone numbers.
Related reading: Connecting Claude and ChatGPT to Zoho Books and Zoho CRM: what changed in November 2026, and the controls finance needs and Zoho One in November 2026: five updates to switch on, and what each one needs.
How can Straits Partners help?
We design and build Zoho Desk for Indian service teams: channels, SLAs, Automation Studio flows, AI agents with approval steps, personal-data rules and integration with Zoho Books, Inventory or ERP for order and invoice data. See Zoho + AI and consulting and implementation.
Want AI to answer routine tickets while your team keeps control? Send us an enquiry with your top ticket types, or use the requirement finder.
Frequently asked questions
What is a blended agent in Zoho Desk?
An AI agent that works with human staff. It is built from the Agent store with assigned roles, permissions and department access. It can hand a conversation to a person with its history and context, and staff can pause it to review or approve an action.
How does Data Privacy Control protect customer data?
It detects personal data, anonymises it before AI processing, and restores the values inside Desk. Masking rules are set per department and can be tested in a playground first.
Can Desk use our own Azure OpenAI deployment?
Yes. Azure-hosted OpenAI models can power ticket summaries, insights, writing assistance and reply assistance.
Which Desk plan do we need?
Edition availability for these features is not yet consistently published. Check your plan inside Desk, or ask us, before planning a rollout.