Zendesk now offers predictive AI agents as a standalone add-on module, covering request resolution without human intervention, intent detection, task automation, agent assistance, and quality evaluation.
What does this change in practice? Until now, predictive AI was often bundled with other components, making it difficult for decision-makers to clearly assess ROI. By separating it out, Zendesk makes gradual deployment easier. You can test predictive AI within a limited scope (a specific ticket category or channel) without adopting an entire ecosystem of features you are not yet planning to use.
At SEIF, we recommend never activating an AI module without first doing the reverse exercise: identify what your teams should stop handling manually, and under what conditions. This organizational question is what determines success—not the technical activation of the module. This is exactly the approach reflected in the Quitoque case study, where the company built ticket-handling autonomy around just four carefully selected use cases, rather than attempting large-scale automation overnight.
Google Drive, Amazon S3, and Box can now be connected as external content sources and leveraged by Help Center search, generative search, AI agents, and the Agent Workspace.
What does this change in practice? Most customer service teams we work with have documentation scattered across Zendesk, a shared Drive, and sometimes an internal wiki. The result is an AI agent with incomplete knowledge because it only has access to part of the company’s actual information. These connectors remove the document migration step that often held back the start of an AI project.
Another new feature: PDF imports. AI agents can now use knowledge contained in PDF files to answer customer queries.
A key point to watch: connecting a source does not mean the content is clean. Before enabling a connector, a quick content audit (outdated documents, duplicates, conflicting information) can prevent your AI from scaling inaccurate answers across your customer base.
Zendesk Contact Center (Talk) is now integrated directly into the Agent Workspace, with a unified status, simplified authentication, native settings, and broader language support.
What does this change in practice? For teams that have been switching between multiple interfaces (Zendesk + a third-party telephony solution), this is a significant simplification of the agent experience. Fewer tool switches mechanically means lower handling time and fewer status errors (an agent marked as “available” in one channel and “busy” in another).
This unification is consistent with what we see at SEIF among teams with well-integrated telephony in Zendesk—one of the key benefits of our partnership with Ringover. The fewer times agents have to switch between tools, the more visible the impact on performance metrics, often more so than with an isolated AI feature. Native Contact Center is a relevant option for teams that do not yet have a dedicated telephony solution, rather than a reason to replace an existing integration that is already working well.
Zendesk’s Action Builder now includes five new connectors: HubSpot, Chargebee, Stripe, Calendly, and GitHub. These allow CRM, billing, scheduling, and development actions to be triggered directly from a Zendesk workflow.
What does this change in practice? This is arguably the most underestimated update of the month. It is not about resolving the ticket itself, but everything that happens afterward (updating the CRM, issuing a customer credit, scheduling a follow-up appointment). This is exactly the type of connector that reduces repetitive manual work for agents without relying on a chatbot or generative AI.
Should you activate every Zendesk update as soon as it is released?
No. Every new feature should address an identified need within your organization. Activating a feature simply because it is available, without a clear use case, dilutes your teams’ attention and makes your instance more difficult to maintain.
Do predictive AI agents replace human agents?
No. They are designed to handle low-value, repetitive requests and free up time for complex cases that require human judgment. This is the approach we see among most of our clients who have successfully deployed AI.
How do I know whether my Zendesk instance is ready for these updates?
It depends on your Zendesk edition (Suite Professional and above for certain features) and the current state of your existing documentation and CRM. A quick audit can help determine this before any deployment.
Zendesk acquired Forethought in early 2026. What does this change?
Zendesk customers can now purchase Forethought capabilities as an additional module. Forethought AI agents by Zendesk autonomously respond to and resolve customer requests, while also helping support administrators deploy and manage AI agents across the customer experience. Depending on their capabilities, Forethought can identify customer intent, automate support tasks, assist agents, extract information, and assess quality.
July 2026 confirms a broader trend at Zendesk: fewer isolated features and more connectors designed to remove friction between tools. For a CX team, the goal is not to activate everything, but to identify the one or two updates that address a genuine, clearly identified pain point.
SEIF supports customer service teams with the audit, configuration, and adoption of these Zendesk updates, including through our certified Zendesk training programs. The focus is not just on technical activation, but on the organizational decisions behind it.
Want to know which of these updates makes sense for your organization? Talk to a SEIF expert.
This article is based on Zendesk’s official updates released in July 2026, enriched by SEIF’s hands-on experience in the field.