What is Casey in Salesforce Service Cloud?

Casey is a ready-made Agentforce agent for customer service, previously known as Help Agent. It answers customer questions based on knowledge articles, documents and web pages and can work in channels such as chat and voice.

What does the Service Rep Assistant do?

It supports human service reps during a case by reading customer sentiment and case context in real time and creating or adjusting an action plan as the conversation changes.

Is the Salesforce native contact center available in Germany?

According to Salesforce, the native contact center became generally available in more than 30 European countries on 16 October 2026. Check the details for your country and edition with Salesforce.

Is Casey suitable for B2B service in manufacturing?

For standard questions with clear answers, such as order status, documentation or return rules, yes. For technical diagnosis, warranty decisions or goodwill, Casey should hand over to a human or ask for approval.

Service Cloud: Casey and the Service Rep Assistant

Salesforce is reorganising its service offering around two kinds of help: Casey, an AI agent that answers customers directly, and the Service Rep Assistant, which supports human service reps during a case. In October 2026 Salesforce also made its native contact center generally available in Europe. For manufacturers with B2B service, these launches can shorten response times and relieve service teams, but only if the knowledge behind them is reliable and the rules for handing over to people are clear. Here is an overview of the launches and an assessment from the client side.

What did Salesforce launch in Service Cloud?

On 13 October 2026 Salesforce summarised ten Service Cloud launches (opens in a new tab). The most relevant for enterprise service organisations:

Native contact center in Europe. Generally available since 16 October 2026 in more than 30 European countries, with Australia and New Zealand following in November. It combines voice, digital channels, CRM data and AI agents in one system.
Service Rep Assistant. Reads customer sentiment and case context in real time and creates or adjusts an action plan for the rep as the conversation changes.
Slack swarming. The Service Rep Assistant can open a dedicated Slack channel for a difficult case, bring in experts and share the case context. This is in open beta.
Workforce engagement management. Planning for hybrid teams of people and AI agents, with real-time monitoring of schedule adherence, forecasts and scheduling suggestions in natural language.
Customer signals and topic clustering. Live sentiment and insights across cases, voice and chat, and automatic ranking of the main contact reasons.
New channels. WhatsApp with sales and payment inside the conversation, and RCS for rich media in native message threads.

Salesforce also introduced new Service Cloud editions, Core, Advanced and Max, which bundle Service Cloud, contact center, workforce management, Slack and Tableau Next with a fixed amount of consumption credits. The announcement does not say which edition includes which AI features, so this needs to be clarified with Salesforce before planning.

Who is Casey?

Casey is the new name for what Salesforce previously called Help Agent: a ready-made Agentforce agent for customer service. It was part of the wave of named agents Salesforce released in September, as Salesforce Ben reported from Dreamforce (opens in a new tab).

According to the demos, setting up Casey is deliberately simple:

Add Salesforce Knowledge and drag in articles, PDFs and web pages as sources.
Connect Casey to channels, including voice through a phone number.
Test the agent live on the same screen.

Two related functions make Casey more interesting for enterprise service:

Agent approvals. When Casey reaches a policy limit, for example a discount or a goodwill gesture above a threshold, it can route an approval request to a human rep. The rep approves or rejects it in the console.
Agent Optimizer. Analyses conversations, ranks contact reasons and can turn existing flows and Apex classes into actions the agent can use.

How does the Service Rep Assistant differ from Casey?

Casey talks to the customer. The Service Rep Assistant works next to the human rep. It follows the conversation, recognises sentiment, keeps track of the case context and suggests the next steps. Salesforce also showed a rebuilt service console that prioritises work by customer value and churn risk, not only by the age of a case.

For many B2B service organisations, the assistant is the easier starting point. The rep stays responsible for the answer, and the AI saves time on searching, summarising and documenting. Mistakes are caught by a person before they reach the customer.

What does this mean for B2B service in manufacturing?

Service in manufacturing differs from consumer service. Customers are companies, questions are often technical, and the stakes are higher: a machine that stands still costs the customer money every hour. That shapes where Casey and the assistant fit.

Good candidates for Casey:

Order and delivery status, if the data in Salesforce is current.
Requests for documentation, manuals, certificates or spare part information.
Standard questions about return rules, service hours or contract terms.
Collecting the right information before a case reaches a technician, such as serial number, error code and photos.

Better handled by people, supported by the assistant:

Technical diagnosis of complex faults.
Warranty and goodwill decisions, which should go through approvals with clear thresholds.
Escalations from key accounts, where the relationship matters more than speed.

The contact center launch is relevant for German manufacturers who run their hotline on a separate telephony system today. A native contact center brings calls, cases and customer data together. Whether it fits depends on the existing telephony, the requirements for data storage and the integration with ticketing in other systems.

What has to be in place before going live?

From an architecture point of view, four things decide whether Casey and the assistant help:

Reliable knowledge. Casey answers from knowledge articles and documents. Outdated, contradictory or unowned articles lead to wrong answers. Every article needs an owner and a review date.
Clear approval rules. Define which decisions Casey may take alone, which need approval and up to which amount. Write them down before the first conversation.
Clean data. Order status, asset data and contracts must be current in Salesforce or reliably available from the ERP. Otherwise the agent gives confident but outdated answers.
Measurable goals. For example, the share of requests solved without a rep, first-contact resolution or time to the first answer, measured before and after.

These points apply to every AI agent, which is why I argue for putting business value and data quality first. They also explain why AI features belong in the target architecture and not only in a pilot.

How should a service organisation start?

A pragmatic path:

Start with the Service Rep Assistant for one team, so people stay in control while the organisation learns.
Clean up knowledge for the most frequent contact reasons, using topic clustering to find them.
Introduce Casey for one narrow channel and topic, for example documentation requests in the customer portal.
Set up approvals for every decision with financial impact.
Measure and expand only when the numbers support it.

Service Cloud is moving fast, and choosing the right building blocks is part of architecture and partner governance. If you want to discuss how these launches fit your service organisation, send me a message.