The Friday Note #13: The Agent Stack

This week: Salesforce put a buying agent on WhatsApp. Google's monetizing AI Mode at 29% of commercial queries. The infrastructure for agent-mediated commerce is assembling faster than anyone expected—and it's reshaping what "go-to-market" means.

📖 The Shift

Commerce Is Moving to Chat

Salesforce dropped their new B2B Commerce suite this week, and the headline feature is an AI buying agent that operates through WhatsApp and SMS. Not a chatbot that answers questions—an agent that actually handles purchasing conversations, from product discovery to order completion.

This isn't a concept anymore. It's shipping.

The implications are significant. For years, B2B commerce has been stuck in a weird place—too complex for simple e-commerce, too transactional to justify high-touch sales for every order. The AI agent sits in the middle: sophisticated enough to handle nuance, scalable enough to handle volume.

And it's happening on messaging apps, not websites. That's the part that should get your attention. The buyer journey is migrating to WhatsApp, SMS, Slack—places where people already spend their time, where the agent can meet them instead of asking them to navigate a portal.

🏗️ The Stack

What Agent-Mediated Commerce Looks Like

The infrastructure is coming together in layers. Understanding the stack helps you see where the opportunities—and the disruption—are concentrated:

Layer 1: Conversation Channels

WhatsApp, SMS, Slack, Teams. The surfaces where agents engage buyers. Salesforce, Meta, and Microsoft are all racing to own this layer.

Layer 2: Agent Intelligence

The models and reasoning engines that power agent decisions. GPT-4, Claude, Gemini—plus the specialized fine-tuning for commerce contexts.

Layer 3: Commerce Orchestration

Catalog, pricing, inventory, fulfillment. The Salesforce Commerce Cloud, SAP Commerce, Shopify Plus layer that agents need to execute transactions.

Layer 4: Enterprise Data

Customer history, contract terms, account preferences. The CRM and ERP systems that give agents context about who they're talking to and what they can offer.

The companies that win will be the ones who connect all four layers seamlessly. A WhatsApp agent that can check inventory, apply contract pricing, and reference past orders—that's powerful. A WhatsApp agent that can only answer product questions? That's a chatbot.

💰 The Monetization

Google's Already Selling Ads

While everyone's debating whether AI search will kill advertising, Google's been quietly building the ad infrastructure. The latest data: ads now appear in 29% of commercial AI Mode queries.

Think about what that means. Nearly a third of the time someone asks Google's AI about a product or service, they're seeing ads. Not organic results filtered through an AI summary—actual paid placements inside the AI experience.

The SEO playbook isn't dead; it's evolving. But the paid search playbook? That's very much alive. The question isn't whether AI search will have ads—it's whether your ads are showing up in the AI context.

The bet that AI would create an "ad-free" discovery experience is looking increasingly naive. The platforms need revenue. The ad infrastructure will follow the attention, wherever that attention moves.

💡 What This Means for PMM

Three Implications

1. Channel Strategy Gets Harder

When buying happens on WhatsApp, your website analytics don't capture the journey. When AI agents mediate discovery, your content might influence the outcome without ever showing up in a click stream. Attribution is about to get messier.

2. Content Needs Agent-Readable Structure

AI buying agents don't read the way humans do. They need structured product data, clear pricing logic, explicit feature comparisons. The SEO discipline of the last decade—structured data, schema markup, clean information architecture—becomes even more important.

3. The "Human Touch" Becomes a Positioning Choice

If most competitors go agent-first, the ability to offer human interaction becomes a differentiator, not a liability. For complex, high-value sales, "talk to a person" might be the premium positioning.

📰 Quick Hits

This Week in AI + Marketing

EU AI Act Article 50 Takes Effect Today

Transparency requirements for AI-generated content are now law in Europe. Disclosure obligations for marketing content start getting enforced.

Open-Source Marketing Agent Stack Hits GitHub

Digital Marketing Pro: 158 skills, 25 specialist agents, EU AI Act compliance built in. The agentic marketing toolkit is going open source.

Deloitte: Agentic AI in B2B Ecommerce "Still a Long Ways Away"

Despite the hype, Deloitte's assessment is that fully autonomous buying agents are further out than the vendor pitches suggest. Infrastructure is coming; full autonomy is not.

Frictionless B2B Buying Closes Deals 31% Faster

Dentsu research: 10-point improvement in buyer experience correlates with 14% uplift in opportunity value. The business case for agent-assisted commerce is becoming quantifiable.