Salesforce Agentforce turns one-way marketing into two-way commerce with AI

Salesforce adds two-way AI commerce to email, SMS, and WhatsApp

Salesforce's latest Agentforce Commerce update turns one-way marketing blasts into two-way conversations across email, SMS, and WhatsApp. Customers can reply directly, and AI agents can answer questions, negotiate offers within guardrails, personalize recommendations, take payment, and place orders.

The goal is simple: remove friction between interest and purchase. For support and CX leaders, that means fewer dead ends, faster resolutions, and a new path to revenue inside channels you already manage.

Why support and CX leaders should care

Mass sends get replies. Even a 1-2% response rate on a million emails is 10,000 conversations. Historically, your team can't read them all, let alone act on them quickly.

Constellation Research analyst Liz Miller put it bluntly: "It's solving the manpower issue of sorting through 6,000 email replies to get to the five that matter to experience." That's the job-surface intent, resolve blockers, and capture orders-without burning your team out.

What's new in Agentforce Commerce

• Two-Way Email, SMS, and WhatsApp: Connect a commerce-enabled AI chatbot to message replies. It can clarify offer terms, check product availability, suggest bundles, and complete orders.
• Guided Shopping: Conversational analytics that steer the agent's next step and trigger targeted promos to close the sale.
• Contextual Search (open beta): Search results shaped by the active conversation, so customers see what actually fits their intent.
• Marketing and Commerce Connected Journeys: Use signals like email signups or cart re-engagement to trigger timely conversations and offers.

Availability: parts of this are live in the Winter '26 release, more features land in February. Contextual Search is in open beta.

Shift in buyer behavior you can act on

Salesforce reports shoppers spent 35% more time on brand sites during the 2025 holidays versus the year before. Traffic to AI-search channels like ChatGPT doubled. Nitin Mangtani, who leads Agentforce Commerce, reads this as a signal: customers want deeper product context and reassurance right before purchase.

For support leaders, that's your moment. Rich answers, clear policies, and fast paths to checkout turn hesitation into conversion.

Operating model: how to make this work without chaos

• Intent playbooks: Define top intents (clarify offer, compatibility, shipping, returns, price match). For each, set approved responses, data sources, and allowed concessions.
• Offer rules: Document stackability and exclusions. The AI must know what can or can't be combined-no exceptions.
• Escalation design: Route high-value, high-risk, or confused intent to humans. Set wait-time and sentiment thresholds for handoff. Log everything back to CRM.
• Payments: Use secure payment links or hosted pages. Never collect card data in free text. Tokenize and audit.
• Consent and compliance: Respect opt-ins, frequency caps, and regional rules. For WhatsApp, follow platform policies (WhatsApp Business Policy).
• Tone and brand: Create a short style guide: voice, refusal patterns, apology language, and how to handle edge cases.
• Product truth: Connect to live inventory, shipping SLAs, and return policies. No stale data.
• Metrics: Track reply-to-conversion rate, average order value from replies, time-to-first-response, containment (no human needed), CSAT, refund rate, and margin impact from discounts.
• Team upskilling: Train agents to supervise AI, approve exceptions, and fix content gaps fast. A focused path helps (AI upskilling for support and CX teams).

Quick pilot plan (30-60 days)

• Pick one campaign (e.g., a coupon on a narrow product set) and one channel (Two-Way Email).
• Limit intents to the top five, with clear approval limits for discounts and substitutions.
• Enable secure payment links and confirmation messages. Test refund/return flows.
• Instrument metrics and set weekly reviews to adjust prompts, policies, and offers.
• Expand to SMS/WhatsApp after you hit target conversion and CSAT on email.

Common failure points to avoid

• Price inconsistency: AI offers that don't match cart logic cause churn and chargebacks.
• Unsupported promises: Shipping or return claims the warehouse can't fulfill.
• Fraud risk: Overly generous concessions without identity or order history checks.
• Privacy leaks: Collecting PII or payment details in open text channels.
• Thread fragmentation: Separate systems for email, SMS, and WhatsApp with no unified history.

What good looks like

• Customer replies "Can I stack this 10% coupon with free shipping?"
• AI checks rules, confirms stackability, suggests a related add-on, shares updated total, and sends a secure payment link.
• Order placed, confirmation sent, CSAT captured. If edge case appears, seamless handoff to a human with full context.

Strategic takeaway

Liz Miller's advice lands: pause for a moment and plan. Two-way channels force new playbooks, new content, and tighter collaboration between marketing, commerce, and support. Get the rules right, wire in live data, and let AI handle the high-volume routine while your team focuses on the few conversations that truly move the needle.

Salesforce previewed these updates at NRF '26 (Retail's Big Show). If you're evaluating, start small, measure hard, and expand only when the experience holds up under load.