Agentforce Case Triage and Resolution

Client and industry

A retailer based in the United States received customer requests through its website, email, and phone. Its support process relied on Salesforce to store customer data, but staff still had to establish what each request concerned, whether it was a simple question or a problem requiring investigation, and which specialist or department should handle it. The workflow also needed to distinguish cases suitable for an initial AI resolution attempt from those requiring human attention. The Agentforce implementation included multilingual support in English and Spanish.

The challenge

Manual triage across communication channels

Requests from the website, email, and phone needed manual review before they could be assigned to a specialist or department. Analysis of the customer’s business process identified triage and categorization as a major demand on support staff’s time.

Customer and product context needed investigation

Staff had to distinguish straightforward questions from problems requiring investigation, check customer data in Salesforce, and identify the product or service delivered to the customer. This work delayed assignment to the appropriate team.

Knowledge searches consumed support time

Human support agents searched the knowledge base for relevant articles and instruction sets before working on a request. Business-process analysis identified this search effort as another major source of time spent before resolution could begin.

Human effort spent on requests suitable for AI

Alongside complex problems, support teams received requests that could be addressed using existing knowledge and instructions. Without separating these cases during triage, human agents remained the first point of handling for both routine and uncertain issues.

Our solution

Routine Automation implemented an Agentforce workflow to support case triage and an initial resolution attempt for suitable requests. It analyzes requests recorded in Salesforce from website, email, and phone interactions, using the available request details and customer records. The agent was configured to support customer requests in both English and Spanish.

Requests recorded in Salesforce

Website, email, and phone requests enter as Cases with the available customer information.

Agentforce · automated triage

Understand, classify, and route

Agentforce supports English and Spanish, reviews the request and customer records, identifies the issue and product or service where a clear match exists, and determines the handling path. Replaces manual initial review, categorization, and routing assessment.

Eligible for AI handling → Agentforce queue

Agentforce · initial resolution

Find guidance and attempt resolution

Agentforce retrieves relevant knowledge and follows configured instructions to try to solve supported requests. Replaces manual knowledge searches and initial handling for eligible cases.

Unresolved or uncertain → human-led queue

Resolved within scope → record outcome and close

Unclear or specialist issue → human-led queue

Human specialist handles exceptions

Specialists receive cases needing human attention, including unsuccessful AI attempts, with the available context and attempted steps.

Human specialist resolves the case and records notes

Support feedback improves guidance

Human escalation notes inform instruction changes that are reviewed and tested before use. Updated guidance supports future Agentforce handling.

With English and Spanish support, Agentforce takes over manual request review, classification, routing analysis, and knowledge lookup for eligible cases. Human teams handle exceptions and review instruction changes.

Business value and results

The retailer reports a 75% reduction in human time spent categorizing cases. For 75% of cases, the first response is delivered within minutes, against a standard one-hour first-response SLA. These results address the initial triage bottleneck and give customers an earlier response while specialists handle cases requiring human attention.

Figures are reported by the retailer. The one-hour comparison is the standard first-response SLA, rather than a measured previous average. Autonomous resolution rates and total support workload reduction have not been quantified.

75% less human categorization time

The retailer reports that Agentforce takes over categorization work accounting for 75% of the human time previously spent on this task, reducing the manual effort required to prepare incoming cases.

First response within minutes for 75% of cases

The retailer reports that 75% of cases receive a first response within minutes, against a standard one-hour first-response SLA. This measures the initial response, rather than the time to resolve the issue.

Routine requests receive an AI first pass

Eligible cases reach Agentforce for an initial resolution attempt, allowing human specialists to focus on requests that require their attention.

Relevant guidance ready for investigation

Agentforce retrieves applicable knowledge and instructions, giving support staff a starting point and taking over the initial knowledge search for eligible requests.