Salesforce launches Agentforce 360 with Slack-native agents, voice agents, and context indexing

Salesforce launches Agentforce 360 for enterprise AI agents - built for Slack, voice, and real customer context

Salesforce has introduced Agentforce 360, a platform to build, test, deploy, and manage enterprise AI agents across Slack, voice, and web channels. For customer support leaders, this means faster setup, tighter control, and agents that can work where your teams and customers already are.

What's new and why it matters for support

• Agentforce Builder (redeveloped): Create and test agents with natural language. Brief the agent, preview responses live, then export via API or JSON to plug into your stack.
• Agent Script: A portable JSON spec for agent behavior, built for human readability. Easier to review, govern, and share across teams.
• Atlas Reasoning Engine controls: New configurable settings to tune how agents reason and make decisions.
• Slack-native interface: Interact with agents directly inside Slack alongside Salesforce apps. Includes connectors for third-party file storage so support teams can pull context without switching tools.
• Agentforce Voice: Phone-ready agents that can be tuned to your brand's tone, interruption rules, and call duration for a better caller experience. Built on Salesforce AI Research's "eVerse" for low-latency streaming and natural flow.
• Contact center integrations: Works with CCaaS platforms including Amazon Connect, Five9, Genesys, NiCE, and Vonage.
• Context Indexing: Automatically uses PDFs, images, flowcharts, tables, and more as context. Governed through Data 360 so security and privacy policies carry through.

Proof points support leaders will care about

• Production usage: Salesforce reports Agentforce has already handled 1.8 million customer conversations internally.
• Containment at scale: Heathrow's "Hallie" agent reaches a 90% chat resolution rate without human handoff, using CRM, loyalty, booking, Wi-Fi, and live flight data. Accessible via WhatsApp and targeting up to a 40% improvement in digital contact efficiency.
• ROI focus: Salesforce highlights the common "pilot purgatory" problem and ties Agentforce 360 to enterprise data so agents answer with context, not guesses.

How this fits into your support stack

• Slack for internal workflows: Let agents triage tickets, fetch knowledge, and trigger automations where your teams already collaborate.
• Voice for customer calls: Stand up branded, interrupt-aware phone agents and route to live agents via your existing CCaaS.
• Context Indexing for answers that stick: Bring unstructured content into scope (FAQs, PDFs, policy docs, diagrams). Govern access with Data 360 to respect roles and permissions.
• Portable behavior: Use Agent Script to standardize how agents act across channels and teams.

Practical steps to put this to work

• Pick 3-5 high-volume intents (order status, refunds, password resets, appointment changes). Design the agent to own these end-to-end.
• Audit data sources your agents need: product policies, troubleshooting guides, macros, known issues, annotated chats. Convert stray PDFs/images into Context Indexing.
• Pilot in Slack first. Use the live preview in Agentforce Builder. Collect edge cases from frontline agents and refine fast.
• Define voice rules before going live: tone, interruption thresholds, escalation paths, max handle time, and compliance text for recordings.
• Wire up governance: map roles to Data 360, set redaction for PII, and log every agent action for QA.
• Measure the right metrics: containment rate, first contact resolution, AHT, CSAT, deflection, and time-to-update content after policy changes.
• Keep humans in the loop: easy handoff, agent summaries for faster pickup, and daily review of failure modes.

Release timeline

• Available this month: Agentforce Voice and Context Indexing.
• Beta in November: Agent Script, the new Agentforce Builder, and Agent Graph.
• Rolling releases will continue through early 2026.

Questions to ask your team (and Salesforce)

• What customer intents will we fully automate vs. triage and escalate?
• Which data sources must be included in Context Indexing to avoid wrong answers?
• How will we monitor reasoning quality in Atlas and enforce escalation rules?
• What PII handling and audit logs are in place across Slack and voice channels?
• How will we measure cost per resolved contact and justify expansion?

Bottom line for support

Agentforce 360 brings agent creation, context, Slack workflows, and phone service under one roof. If you have clear intents, clean knowledge, and strong guardrails, you can move from pilot to production without stalling out.

Upskill your support org on AI agents

If you're planning an Agentforce rollout or building AI-assisted support, explore practical training paths built for support roles:

• AI courses by job function
• Popular AI certifications

The announcement comes a day ahead of Dreamforce 2025 at the Moscone Center in San Francisco.