The success of Salesforce’s new AI agents stands and falls with your data foundation. With Agentforce, Salesforce is building artificial intelligence firmly into the Service Cloud. That brings new tools for service agents, administrators and management. But before the AI takes work off your hands, it reveals how good your existing automations really are.
What is Agentforce from Salesforce?
Agentforce is Salesforce’s platform for AI agents that automate customer interactions directly within your existing architecture. Unlike general-purpose language models, these agents work directly with your existing workflows and company data. Built-in guardrails prevent the AI from inventing facts, so-called hallucinations. That protects sensitive customer data. The agents take over recurring tasks and relieve your team.
The foundation: Agentforce 360 and Data 360
Under the name Agentforce 360, Salesforce bundles the AI agents, Slack and Data 360 (the former Data Cloud). All applications use the same models and access the same harmonised data layer. Through the “Intelligent Context” feature, the agents also read unstructured content such as PDFs or diagrams to resolve complex issues. Slack serves as the surface where your team and the AI agents communicate with each other.
New AI colleagues for customer service
At Dreamforce, Salesforce introduced “Casey”. This pre-configured customer service agent was trained on real Salesforce help documentation. In live demos, Casey applied return policies for different countries on its own, without users having to refine their input.
Then there is the new Agent Optimizer. It scans past service conversations, sorts contact reasons by volume and automation potential, and automatically converts existing Flows and Apex classes into actions for the agent.
Why technical debt becomes visible now
This automatic conversion is exactly where consulting is needed. When the Agent Optimizer picks up your existing automations, it inherits their weaknesses. Poorly configured Flows or outdated Apex code slow the AI down directly. Before you activate Agentforce, your Salesforce org belongs on the test bench. At Aviando we see a clean data foundation and the reduction of technical debt as a mandatory prerequisite for using AI.
Tools for service agents and team leads
Salesforce has reworked the service console. Work assignment now prioritises tickets by customer lifetime value and the customer’s churn risk, no longer just by the age of the ticket.
Omni-Channel gets the new work type “Agent Approvals”. When an AI agent such as Casey hits a policy boundary, it hands the case over to a human. The agent immediately sees the full context: conversation summaries, order history and interaction details are already loaded before they actively take over the case.
To manage these hybrid teams, Salesforce introduces the Command Center for Service. It provides real-time insight into ongoing conversations, including sentiment analysis. On that basis, the accompanying Schedule Manager creates staffing plans within seconds.
AI-assisted administration with AIforce
With AIforce, administrators will describe their requirements in natural language. A skill file collects parameters such as channels, queue structures or support numbers. A coding agent, demonstrated at Dreamforce with Claude Code, translates that into Metadata API calls and creates a manifest for sandbox deployment. That makes it easier for new administrators to get started. You still have to define the business requirements precisely, though.
The technology delivers the tools, but your system architecture determines the quality. We analyse your Flow inventory and identify technical debt in your org. Talk to us to prepare your Salesforce environment for Agentforce.
Sources
- https://mind-force.de/vertrieb/dreamforce-salesforce-stellt-agentforce-vor/
- https://comselect.de/agentforce-360/
- https://www.cloudcomputing-insider.de/salesforce-agentforce-360-dreamforce-2025-a-c02e313139c18307d0c94cda8bc796e2/
- https://www.salesforceben.com/top-service-cloud-announcements-at-dreamforce-26/
- Dreamforce 2026 Live: highlights (salesforce.com)
