Jamf and Ravenna bring Mac tech support to Slack with AI agents connected to Jamf Pro

Jamf has published on its corporate blog (21 September 2026) an article describing the integration between Ravenna, an AI-based help desk service, and Jamf Pro, its Apple device management platform. The pitch is simple: employees shouldn't need to understand device management to fix a problem with their Mac, and Ravenna's AI agents can handle those requests directly in Slack, drawing on data that already exists in Jamf Pro.

The workflow described starts from a typical case: an employee reports that an application won't open. Before sending a generic troubleshooting guide, the Ravenna agent can check in Jamf Pro whether the expected software and configuration profiles are actually installed on that machine, as well as review hardware details and the device's management status. With that information, the agent tailors the conversation to the specific case instead of applying a standard answer that might not match the real configuration of that fleet of machines.

One point the article stresses is that the instructions are not improvised by the AI: they come from the support team's own runbooks, which are connected to Ravenna. For each agent it is possible to define where it should look for information, which tools it can invoke and when it should hand the conversation over to a person. This addresses a known limitation of generic AI assistants: a reasonable recommendation for someone else's infrastructure may be inadequate for the particularities of a specific device estate.

When a human technician takes the case, Ravenna displays the requester's device information in a sidebar: model, macOS version, available storage, FileVault status and last check-in date, visible while reading the reported issue. It is an integration designed to speed up mixed work between AI and people, not just to fully automate support.

For recovery and security requests, the article details three steps that can be built into a Ravenna workflow: locating the right Mac among the devices assigned to an employee (relevant if they have more than one registered), remotely locking a Mac reported as lost, stolen or compromised via a command sent through Jamf Pro, and retrieving the FileVault key escrowed in Jamf Pro so an employee can regain access to their encrypted machine. Jamf stresses that it is the organization that decides which conditions must be met before these actions are carried out: a device lookup can simply run, while key recovery can be configured to wait for approval. Ravenna allows each of an agent's tools to be marked as automatic, subject to the requester's confirmation, or dependent on authorization from a designated person.

For more specific needs not covered by the standard actions, the article mentions Foundry, described as a beta feature that allows custom functions to be created by describing in natural language what is needed, using the existing connection to Jamf Pro to build and test them. Once published, a function can be invoked by an agent or used within a workflow, and is reusable across different contexts. It is also possible to roll back to a previously published version if a change causes problems, and the logs of each step of the workflow (what ran, what it returned and where it failed, in case of error) remain available for review.

The article closes with a mention of a related session at JNUC 2026 (Jamf Nation User Conference), Jamf's user conference, without detailing further content about it.

It is worth noting that this is a promotional piece from the vendor itself (Jamf, in collaboration with Ravenna) hosted on its corporate blog, so it describes the product's capabilities in favourable terms and includes no adoption figures, pricing, regional availability or comparisons with other AI-assisted technical support solutions. Within the broader trend of AI assistants applied to internal IT support, the case illustrates a pattern that recurs across the sector: using AI agents not as autonomous replacements for support staff, but as a layer that queries existing operational data (in this case, Jamf Pro's device inventory and status) and executes predefined actions within configurable guardrails, leaving it to each organization to decide how much autonomy to delegate to the AI for sensitive actions such as remotely locking a machine or handing over an encryption key.