Service records and AI have different jobs
Zendesk's Customer Service offering brings requests, channels and support work together. The ticketing system provides the record and workflow; AI agents aim to handle eligible requests through knowledge and actions; Copilot assists human teams during their work. These products are related, but the buyer should still ask what each component is responsible for. An automated resolution, a suggested response and a ticket assigned to a person represent different outcomes. Signet uses this distinction to avoid treating every AI feature as the same type of work. The official pages describe the capabilities, and the customer's deployment determines how they are configured and used. A product evaluation can therefore start with one recurring request and follow it through the appropriate path. Which component reads the knowledge, which component can take an action, and which component asks a person to decide? That exercise connects a list of features with the service process the company actually wants to operate, including the points where a human remains responsible.
Internal requests need permissions and an owner
Employee Service applies the service model to employees who need help from internal teams. A request about a workplace account or an internal process can require information across several systems, and access to that information must follow the employee's permissions. Zendesk's employee offering and May 2026 announcement describe internal service agents and connections to familiar work tools. The announcement identifies source-level permissions as part of the intended employee experience. Availability notes belong alongside that announcement because a product vision and a generally available capability are different stages. Signet interprets the buyer's task as deciding which internal requests can be handled consistently, which require an approval, and which should reach a named team. An employee interface becomes useful when it respects those decisions. It can reduce the effort of finding the right entrance, but the organization still owns its policies and knowledge. Applicants can make this concrete by describing an internal request they helped route or resolve and the permission boundary they had to respect.
KRAFTON shows the work behind automation
The company-published KRAFTON case describes support across brands, languages and locations. It reports approximately one million annual inquiries, a team of more than thirty agents and thirteen supported languages in that customer's described operating scope. KRAFTON used macros, automations and triggers to handle work consistently, and reporting to examine brands, languages and inquiry types. The case then describes adding an AI agent and improving its use with a dedicated internal team and Zendesk guidance. The operational lesson is the sequence of setup, observation and adjustment. Signet reads this as evidence of implementation work rather than a shortcut from purchase to business results. The case reports an eighteen percent reduction in processing time and a fifteen percent reduction in agent costs in that customer context. Those figures describe KRAFTON's published experience. A different buyer should define its own baseline, request mix and review period before comparing results. For an applicant, the useful preparation is to explain how an operational change was adopted, measured and revised after people began using it.