The Latest BIG News from Salesforce, Zoom, Oracle, Snapchat, and Microsoft - What It Means for Customer Support
Here's a fast, practical recap of last week's biggest CX and contact center moves. Use it to decide what to adopt, what to test, and what to avoid as you plan your next quarter.
Dreamforce 2025: Agentforce 360, Slack as the Agentic OS, and Apromore
Salesforce set a clear direction: the "Agentic Enterprise," where every employee pairs with an AI agent to offload routine tasks. The new Agentforce 360 extends Salesforce beyond CRM, embedding agents across Sales, Service, and Marketing. Slack becomes the primary interface for those agents, so work happens where conversations already live.
Salesforce also acquired process intelligence firm Apromore to map and optimize end-to-end workflows, and expanded its OpenAI partnership to bring advanced GPT models into Slack and Agentforce 360. For support leaders, this signals a move from chatbots to task-capable agents that act across systems.
- What to do now:
- List your top 10 repeatable service workflows (e.g., refunds, entitlements, password resets). Mark those with clean data and clear rules for agent automation pilots.
- Centralize knowledge and policies in one source of truth. Agent outputs are only as good as your content.
- Use process mining (Apromore-style) to find bottlenecks before automating them.
- Pilot Slack-first actions: trigger case updates, knowledge lookups, and approvals from Slack instead of switching apps.
For official updates and releases, track the Salesforce Newsroom.
Zoom + Oracle: A 15,000-Seat CCaaS Deal and OCI Availability
Zoom confirmed a 15,000-seat contact center deployment with Oracle and expanded availability on Oracle Cloud Infrastructure (OCI). Zoom says the pairing will unify customer interactions, employee workflows, and data into a single intelligent system. Oracle highlighted the foundation this sets for AI-driven engagement.
- What to do now:
- If you run on OCI, review network peering, latency, and regional availability. Voice quality and routing should be your first tests.
- Map how Zoom CX would connect to your CRM, WFM/WEM, and knowledge tools. Avoid duplicating systems just to check a box.
- Define measurable targets: first contact resolution, deflection, AHT, and CSAT improvements tied to specific workflows.
Snapchat AI Spill: Why Support Chatbots Still Need Human Oversight
Researchers showed Snapchat's AI could be prompted via storytelling to surface dangerous content it would otherwise block. This is a clear warning for customer support teams: clever prompts can bypass filters, especially with narrative framing.
- What to do now:
- Keep bots on a strict allow-list of safe intents and content. Block free-form responses on sensitive topics.
- Add layered safety: retrieval whitelists, PII detection, toxicity checks, and real-time escalation to a human.
- Red-team your assistant with narrative, multilingual, and obfuscated prompts. Treat safety testing as an ongoing practice, not a launch task.
- For minors or regulated products, add age gates, session-level monitoring, and immediate shutdown triggers for risky requests.
Microsoft + Legal & General: Dynamics 365 Contact Center with Copilot
Legal & General is building an AI-based service platform on Dynamics 365 Contact Center with Copilot. Agents get a real-time view of each customer, suggested next steps, and simplified workflows to reduce transfers.
- What to do now:
- Fix data hygiene first: duplicate records, missing IDs, and scattered notes will limit any assistant.
- Start with one line of business with clean data and clear processes (e.g., claims or annuities), then expand.
- Define guardrails for suggestions and outbound prompts. Every AI action should be auditable.
- Measure impact per call type: reductions in handle time, hold time, transfers, and after-call work.
Explore product details on the official Microsoft Dynamics 365 Contact Center page.
Practical Checklist for Support Leaders
- Workflows: Pick the top 5 high-volume, low-judgment tasks to automate first.
- Data: Clean CRM fields, unify knowledge, and document edge cases.
- Safety: Enforce allow-lists, run ongoing red-teaming, require human review for sensitive actions.
- Change: Train agents on AI handoffs, approvals, and exception handling.
- Metrics: Set a 90-day scorecard for AHT, FCR, CSAT, and containment by intent.
- Scale: Prove value in one queue, then roll out to adjacent flows.
Skill Up Your Team
If you're standardizing on AI across support, level up your team's capabilities with focused training and certifications. Explore role-based learning paths here:
Your membership also unlocks:
