AI Agents-Essential end-of-life

I agree that it can seem daunting. I’ve been using the current “Essentials” agent since 2023, back when it was originally considered the “Advanced” agent, and that was a lot to learn.

From what I understand, there will be a guided process to set it up. We are in the same boat here - using Suite Professional with no sandbox. Based on my experience, it helps to start with replies using knowledge content and add use cases over time. I’ve documented all of our current flows and written myself a “guidebook” of sorts to help with the transition since from what I’ve been told, there will be no way to convert the current AI agent into the new one. Documenting any instructions and flows you currently have set up will be key to getting the best results from the advanced agents. While it takes time, you’ll be glad to have made those preparations when you start working on the new agent.

Keep in mind that the new version will allow you to write prompt based flows. What I found helpful was documenting each step in my current flows and then you can feed that into ChatGPT or other preferred AI tool, asking it to create a prompt to be used for a Zendesk Advanced AI agent flow.

While I know it’s stressful, I believe it will ultimately be a good thing. Bug fixes end in August, with EOL actually scheduled for December, so you’ve got a bit of time.

Thanks, ​@KROB. We set up Essentials when it was launched for email/webform last fall. We don’t use messaging/chat, only autoreplies (and currently have no plans to use chat/messaging). We only have a couple of Instructions; I will make sure to save those off.
I am frankly baffled by the documentation for Advanced re email. Essentials doesn’t require any configuration, really. Just connect it to brand/data source and channels, and “poof” autoreplies start happening. Nothing of “flows” or “use cases” exists. I don’t really have any clue how to get started, especially since the documentation is almost entirely written assuming chat bots, not email/webform.

For the most part, focus on anything that would require collection of information before being forwarded to an agent, along with any external connections that require APIs. The content sources should be the same. If you don’t require APIs or collection of specific details, you shouldn’t have a ton of work to do during setup. One benefit is that customers can ask follow up questions, which is going to greatly improve the resolution rates. 

I understand your concern—this kind of transition feels risky, especially without a sandbox to test things safely. Sudden timeline changes and limited guidance can make it even more stressful.

From my experience working on tools like stipendiocalcolatore.it, even small changes in core functionality (like email workflows or calculations) can impact real users quickly if not handled carefully. That’s why having clear migration steps and proper documentation is so important before rolling out such updates.

Thank you, ​@Mick11. I look forward to the documentation on the new capabilities. I am hoping it includes not just technical implementation documentation, but also transition planning, especially for those of us with no sandbox to work in.

Regarding webform, I’m hoping that your team is aware of this issue that Instant Answers introduced that has caused us to disable ticket submission via webform entirely, and that you have a plan to fix this soon even though my ticket reporting the issue was closed as “working as intended”.

​@Mick O'Donnell I am seeing new documentation being posted, but so far I haven’t seen anything that relates to migrating from AI Agents-Essentials to the new model. I only see documentation that notes that AI Agents-Essential is now considered “legacy” and documentation for setting up AI Agents from scratch.

Where is the documentation on how to transition from AI Agents-Essential to the new model in a production instance with no sandbox, where we cannot risk breaking existing behaviors to learn new behaviors?
For example, when we transitioned from “autoreplies with articles” to AI Agents-Essential the entire configuration UI for “autoreplies with articles” disappeared from Admin Center. Fortunately, we didn’t need to revert behavior b/c AI Agents-Essentials was light-years better than “autoreplies with articles”.

But at this point we have a perfectly stable AI Agent-Essentials configuration which our end users and agents depend on. I can’t just take configure the new AI Agent and take a leap of faith that everything will be fine when I switch over.

HI David, 

We'll be provide detailed documentation to support migrations from Essential in the coming weeks. There's no rush on this. If you're happy with your current setup on Essential, stay with this for now.

Kind regards,

Mick

​@Mick O'Donnell Last week I got yet another about EOL for Essentials, advising “It is recommended that all customers currently using AI agents Essential or legacy AI agent functionality migrate to the new AI agent experience before August 31, 2026 to avoid disruption”

It also notes “Step-by-step migration documentation in the “What do I need to do?” section of this announcement to support each configuration type”. Well, not really.

There is documentation on how to create a new AI Agent that does the same thing our Essentials agent. This documentation doesn’t provide any guidance whatsoever in how to validate this functionality in a Professional instance without risking breaking my production environment’s behavior. I have built triggers, automations, custom statuses, and Agent SOP around how things work with Essentials. I think it’s highly likely that there are sufficient differences between Essentials and the new Agents that this will break, and I have no documentation or instructions to guide me.

Thie migration documentation still states

In the future, an automated, in-product migration tool will be offered that will help you recreate an essential AI agent without manual rebuild. You can either follow the steps in this guide to unlock the potential of the new platform earlier, or wait for the tool to be available. More details on this process will become available later in 2026.

WHEN will this tool be available, given that we’re being advised to do our migration by the end of next month? And what assurances can I get that it will drop-and-go with my triggers, automations, custom statuses, and Agent SOP - or that sufficient documentation will be provided that will let me update this automation with low risk of breaking our production environment.

I remain utterly perturbed about this migration. Nothing has given me any confidence that our support experience isn’t going to simply break.

Hey Dave,

I have been working on setting up AI advanced agent over the last couple of weeks on our end; we didn’t use essentials really (had a small stint where we dabbled) but now are moving towards an AI agent connected to our order system to help provide information on existing orders to our customers to remove the load from our staff. We are on enterprise so I have been doing this in Sandbox, and from what I can tell, there isn’t an easy way to just turn this on in Production. Once you enable the AI advanced bot to the email channels, it will start doing its thing. You can possibly stagger this by enabling the AI advanced bot, but not enabling any access to email channels until are you ready to test some tickets. 

For getting the AI to understand what it can or cannot handle, you will need to use the Use Case/Procedure set up. You’ll need to detail in the use case what the AI can handle, and make separate use case for what it cannot handle; the procedure instructions use a natural language system, and you need to fine tune ‘telling’ it how it should handle the email from start to finish. 

You can use a combination of configured actions and triggers in order to still have the Advanced AI agent set tags, but you likely will need to test the actual email tickets by eventually enabling the email channels in production. It might be a good approach (albeit, annoying) to go in between enabling/disabling the email channels for the AI bot so you can make tweaks without it handling too much in your production environment amongst the changes.

If you’d like a hand with anything, please don’t hesitate to reach out and I will gladly see what I can do to give you a hand with this. It has certainly been a learning process to get this to work as needed!