Salesforce Service Cloud Voice (SCV) is a native integration between Amazon Connect telephony and Salesforce CRM that brings phone calls directly into the agent workspace — eliminating context-switching and enabling real-time AI assistance at every point in the call. If you're building or modernizing a Salesforce contact center, SCV is the most transformative capability available today.

What is Service Cloud Voice?

Service Cloud Voice embeds an Amazon Connect softphone directly inside the Salesforce Service Console. When a call comes in, the agent sees the caller's full CRM record automatically — account history, open cases, recent purchases, and sentiment data — all before saying hello. AWS Transcribe converts the conversation to text in real time, Einstein AI analyzes the transcript to fire Next Best Action recommendations, and when the call ends, AI generates a complete case summary automatically.

SCV Architecture: How It Works

The architecture has three layers. The telephony layer (Amazon Connect) handles call routing, IVR logic, queue management, and the voice stream. The intelligence layer (AWS Transcribe + Einstein) converts speech to text in under 300ms and runs NLP analysis to detect intent, sentiment, and suggested actions. The CRM layer (Salesforce) surfaces the right data, logs everything automatically, and enables agents to take action without leaving the workspace.

Real-Time Transcription Setup

Real-time transcription is powered by AWS Transcribe Streaming, processing audio in 8-second chunks and returning transcribed segments to the Salesforce UI via a streaming API. Configure a Voice Call Record in Salesforce, connect your Amazon Connect instance via the SCV setup wizard, and enable the Transcription feature in your contact center configuration. The transcript appears live in the agent workspace as the call progresses.

Einstein Next Best Action Mid-Call

Einstein Next Best Action (NBA) uses the live transcript as a real-time signal to fire contextual recommendations. Configure NBA strategies using Flow-based recommendation logic and connect them to the voice transcript via a Platform Event subscription. When Einstein detects trigger phrases — "cancel", "escalate", "competitor", "discount" — the relevant recommendation appears instantly in the agent's workspace.

Sentiment Analysis and Caller Mood

SCV integrates with Amazon Comprehend for real-time sentiment analysis on the transcript stream. Sentiment scores update continuously and are stored on the Voice Call record. Build supervisor dashboards with live sentiment heatmaps, trigger automatic supervisor alerts when sentiment drops below a threshold, and report on sentiment trends by agent, queue, and time period.

Post-Call AI Summaries

When a call ends, Einstein generative AI reads the full transcript and produces a structured case summary — reason for contact, actions taken, resolution status, and recommended follow-up. Written directly to the Case record, eliminating 3–5 minutes of post-call wrap-up per agent. At 50 calls/day per agent, that's 250+ minutes of productivity returned daily.

Skills-Based Call Routing with SCV

SCV routes calls through Amazon Connect routing profiles, which map to Salesforce Omni-Channel queues. Configure agent skills (language, product expertise, tier level) in Amazon Connect, then use Contact Flow blocks to match caller attributes to required skills. The result: a Spanish-speaking enterprise customer with a billing issue routes directly to a bilingual billing specialist — automatically, every time.

Salesforce Marketing Cloud Journey Builder is the most powerful customer journey orchestration tool available to enterprise marketers. Unlike batch-and-blast email tools, Journey Builder treats every subscriber as an individual — moving them through personalized paths based on real-time behaviors, CRM data, and AI-driven decisions.

Journey Builder vs Automation Studio

Journey Builder is individual-centric and event-driven — one subscriber moves through their own unique path based on behavior. Ideal for welcome series, nurture flows, post-purchase sequences, and lifecycle management. Automation Studio is batch-oriented — processing large groups on a schedule, running SQL queries, importing files, triggering batch sends. Most enterprise implementations use both: Automation Studio prepares and segments audiences, Journey Builder orchestrates the experience.

Entry Sources: Getting Contacts Into Your Journey

Journey Builder supports seven entry source types. Data Extension Entry — contacts enter when added to a specified DE, populated by Automation Studio SQL. API Entry — fires when a real-time event occurs in an external system (form submission, purchase, app event). Salesforce Data Entry — triggers on CRM record changes via MC Connect. Choose your entry source based on latency: API entry fires in seconds, DE entry on a schedule.

Decision Splits: The Brain of Your Journey

Use Engagement Splits to branch on email opens, clicks, and conversions. Use Random Splits for A/B testing message variants. Use Einstein Engagement Splits to let AI predict which contacts are likely to convert and route them to your highest-value path. Stack multiple splits to build sophisticated segmentation logic without code.

AMPscript Personalization Inside Journeys

AMPscript renders server-side at send time with access to the subscriber's data context. The most common pattern is Lookup() — retrieving a value from a Data Extension based on a matching key: %%=Lookup("ProductDE","ProductName","ProductID",@productId)=%%. Use LookupRows() for multiple records. Use IF/ELSE blocks for conditional content rendering. Cache results in variables to avoid multiple database calls per render.

MC Connect: Syncing with Salesforce CRM

Marketing Cloud Connect creates a bidirectional sync between SFMC and Salesforce CRM. Use synchronized Data Extensions to pull Contact, Lead, Opportunity, and custom object data into Marketing Cloud. Configure Synchronized Send Relationships to update CRM records when emails are sent, opened, or clicked. A Lead Status change to "Qualified" can instantly fire a high-priority nurture sequence.

Salesforce Omni-Channel is a native routing engine that intelligently assigns work — cases, chats, calls, leads — to the right agent at the right time based on skills, capacity, and priority. Unlike simple queue-based routing, Omni-Channel understands agent workload in real time and makes assignment decisions that optimize for both customer experience and operational efficiency.

How Omni-Channel Routing Works

When work enters Omni-Channel, the routing engine evaluates three factors simultaneously: queue membership (which agents can handle this work type), agent skills (which available agents have the required skill set at the required proficiency level), and capacity (which agents have available capacity within their configured workload model). The engine selects the best-matched agent and pushes the work item to their workspace instantly.

Skills-Based Routing Configuration

Three steps: First, define Skills in Salesforce Setup — "Spanish Language", "Billing Specialist", "Tier 2 Support". Second, assign skills to agents with proficiency levels (1–10). Third, configure Routing Configurations to require specific skills at minimum proficiency thresholds. When work requires "Billing Specialist at 7+", only agents with that skill at that level are eligible.

Capacity Models: Tabs vs Percentage

Tab-based capacity limits agents to a fixed number of concurrent items (e.g., max 3 chats). Simple and easy to understand. Percentage-based capacity assigns a weight to each channel — a voice call consumes 100% of capacity while a chat consumes 33%, allowing agents to handle one call OR three chats. Percentage-based is more sophisticated and better reflects real-world workload.

Supervisor Console

Real-time visibility across all active agents, current work items, queue depths, and wait times. Supervisors can monitor conversations silently, whisper coach agents during live chats, barge in to take over interactions, and manually reassign work items. Configure alerts for queue wait times exceeding SLA thresholds — the console flags these automatically in red.

Digital Engagement Channel Setup

MIAW embeds a Salesforce-native messaging window with bot deflection, file sharing, and rich messaging. SMS requires 10DLC registration — a US carrier compliance requirement that takes 2–4 weeks. Facebook Messenger connects via a Facebook App webhook to Salesforce. WhatsApp requires a WhatsApp Business Account connected through a Salesforce-approved partner like Vonage or Twilio.

Salesforce Agentforce represents the most significant shift in enterprise AI since Einstein. Where Einstein Bots followed scripted decision trees, Agentforce uses the Atlas Reasoning Engine — a multi-step planner that interprets natural language intent, retrieves live data, executes real Salesforce actions, and adapts to unexpected inputs mid-conversation.

What is the Atlas Reasoning Engine?

Atlas operates in a Plan → Retrieve → Act → Verify loop. When a user sends a message, Atlas parses the intent and identifies sub-goals. It retrieves relevant data via RAG from Knowledge Articles, CRM records, and configured data sources. It selects and executes Actions (Flows, Apex, APIs). Finally, it verifies the outcome meets the original goal and either returns a response or loops for another reasoning cycle.

RAG Grounding: How Agentforce Stays Accurate

Retrieval Augmented Generation (RAG) separates Agentforce from generic LLM chatbots. Instead of relying purely on model weights (which hallucinate), Agentforce retrieves verified data from your Salesforce org at inference time. Configure Data Sources — Salesforce objects, Knowledge Articles, files, external APIs — and Agentforce queries them during reasoning. Retrieved data is injected into the LLM prompt as grounding context, dramatically increasing accuracy.

Agent Topics and Actions

Topics define agent scope declaratively: "This agent handles order inquiries for orders placed in the last 90 days." Actions are what the agent can execute: invoke a Flow, call Apex, query a record, send a message, create a case. Each action is discrete, logged, and auditable. The agent autonomously decides which actions to take in what sequence to fulfill the user's goal.

Einstein Trust Layer

Enterprise-grade AI safety: data masking (PII scrubbed before any external LLM call), toxicity detection (harmful output filtered before reaching users), zero data retention (Salesforce guarantees prompts are never stored by the LLM provider), and audit trails (every AI interaction logged for compliance). This makes Agentforce viable for regulated industries — healthcare, finance, legal.

Human Handoff with Full Context

When Agentforce escalates, it transfers the entire session context. The receiving human agent sees a complete transcript of the AI interaction, the customer's intent as interpreted by Atlas, all actions already taken, and the recommended next step. Average handle time drops significantly because the human never re-gathers information the AI already collected.

Marketing Cloud has two server-side scripting languages. AMPscript is purpose-built for email personalization and renders inside emails at send time. SSJS (Server-Side JavaScript) runs in CloudPages and Landing Pages, enabling complex application logic and external API calls. Knowing when to use each is a core Marketing Cloud developer skill.

AMPscript: Purpose-Built for Email

AMPscript renders server-side at send time with access to the subscriber's data context and the sending Data Extension. Most common pattern — Lookup(): %%=Lookup("ProductDE","ProductName","ProductID",@productId)=%% fetches a product name from a DE at render time. Use LookupRows() for multiple records. Cache results in variables rather than calling Lookup multiple times — each Lookup is a separate database query.

SSJS: JavaScript for CloudPages

SSJS runs on Marketing Cloud's server when a CloudPage or Landing Page loads. Use it for: processing form submissions and writing to Data Extensions via Platform.Load("Core","1.1.1"), making HTTP API calls with HTTP.Get() and HTTP.Post(), complex data manipulation that would be verbose in AMPscript, and building interactive web experiences that bridge Marketing Cloud and external systems.

Combining Both Languages

The most powerful patterns combine both. A SSJS CloudPage collects form data, validates it, calls an external API, and writes a record to a Data Extension. Journey Builder triggers from that DE update and sends an AMPscript-personalized email that dynamically renders content from the same DE. SSJS handles application logic, AMPscript handles email personalization — they complement each other perfectly.

Salesforce Data Cloud is a real-time customer data platform (CDP) natively embedded in Salesforce. It ingests data from any source, resolves fragmented customer identities into a unified profile, enables real-time segmentation, and activates that data across every Salesforce product and external system. It is the foundation of every modern Salesforce AI and personalization strategy.

Identity Resolution: The Core Value Proposition

Most enterprise businesses have fragmented customer data — the same person exists as a Lead in Salesforce, a subscriber in Marketing Cloud, a user in their mobile app, and a record in their ERP. Data Cloud's Identity Resolution engine stitches fragments together using deterministic matching (exact email/phone matches) and probabilistic matching (behavioral and demographic similarity scoring). The output is a Unified Individual — one master profile consolidating every data point across every system.

Data Streams: Ingesting Any Source

Data Cloud ingests data via Data Streams — connectors to Salesforce CRM, Marketing Cloud, web analytics, mobile apps, external databases, and cloud storage. Each Data Stream maps source fields to the Data Cloud canonical schema of standard objects (Individual, Contact Point Email, Engagement, Product). Once ingested, data is available for segmentation and activation within minutes.

Calculated Insights: Real-Time Aggregations

Calculated Insights are SQL-based metrics computed continuously across your unified data: total lifetime purchase value, average days between purchases, product category affinity score, churn risk index. These metrics update in near real-time as new data arrives and are available as profile attributes for segmentation. Einstein models consume Calculated Insights as features, improving prediction accuracy significantly versus raw event data.

Salesforce Digital Engagement enables asynchronous digital channels — Web Chat, SMS, Facebook Messenger, WhatsApp — to route through Omni-Channel into the Salesforce agent workspace. Each channel has different licensing requirements, setup complexity, and compliance obligations. This guide covers all four channels end-to-end.

MIAW (Messaging for In-App and Web)

MIAW is Salesforce's modern web messaging framework, replacing legacy Live Agent Chat. MIAW supports persistent conversations (chat history survives page refresh), rich media (file attachments, images, carousels), bot integration via Agentforce, and asynchronous messaging. Embed via the Embedded Service Deployment — a JavaScript snippet that loads a configurable chat launcher on your website. MIAW conversations are stored as Messaging Sessions, convertible to Cases via Flow when escalated.

SMS with 10DLC Compliance

10DLC is a US carrier requirement for business SMS. Before any business SMS via Salesforce Digital Engagement, you must: register your brand with The Campaign Registry (TCR), register your campaign use cases (marketing, customer care, 2FA), and have numbers approved by each carrier. Non-compliant SMS sees 30–50% filtering rates. Registration takes 2–4 weeks — plan accordingly before your go-live date.

Facebook Messenger Integration

Requires a Facebook App with Messenger permissions, a webhook to receive message events, and a Messaging Channel in Salesforce Setup. Messages flow through the webhook into Salesforce as Messaging Sessions and route via Omni-Channel. Configure supervisor flagging via a custom action in the Messaging Session layout — agents can flag conversations for review without interrupting the interaction flow.

WhatsApp Business via Digital Engagement

Requires a WhatsApp Business Account (WABA) approved by Meta, connected through a Salesforce-approved partner (Vonage or Twilio). Configure Message Templates for proactive outbound messages — WhatsApp requires pre-approved templates for any message sent outside the 24-hour service window. The 24-hour rule: only template messages can be sent more than 24 hours after the customer's last inbound message.