makersmarx

Disclosure: I work at Iterable. The opinions here are my own, and I've tried to give each platform a fair read based on its public announcements.

This fall, nearly every major customer engagement platform held its big launch within a few weeks of the others. Dreamforce in mid-September, Klaviyo's K:BOS, Iterable's Fall Product Release on September 24, and Braze Forge at the end of the month. Read the announcements side by side and a pattern shows up quickly. The 2025 pitch was "AI helps you write the email." The 2026 pitch is "AI remembers the customer and acts on it."

As a Solutions Architect, I care less about the keynote slogans and more about the architecture underneath. So here's how the four bets compare, and what each one means for teams designing a modern marketing stack.

1. Iterable: Closing the "Context Gap"

Iterable's announcement is titled "Iterable Gives Marketing Tech a Memory", and the framing is about what it calls the context gap: the customer data, behaviors, conversations, decisions, and outcomes that are usually scattered across separate systems.

The Bet: A "continuous intelligence loop." Customer context informs each interaction, each response creates new intelligence, and that intelligence shapes what happens next. As CPO Nick Beil put it: "Customers remember every interaction they have with a brand. The technology behind those experiences should, too."

What shipped: Native RCS and WhatsApp that use the same real-time customer context as the rest of the platform. An Integrations Hub and native Salesforce Sync. Journeys Live Data, which brings real-time signals into running journeys. And the Nova agent family, including Nova Conversations, a new Monitoring Agent that spots anomalies and recommends fixes, and an upgraded Analytics Agent.
The proof points: Babylist reports reclaiming 50% of its CRM team's campaign bandwidth with Nova Intelligence. HelloFresh uses Journeys Live Data to run several experiments inside one global journey, "no custom middleware, no profile bloat."

The Architect's Take: Iterable is betting on depth of context inside the platform. It's a natural next step from the Spring Nova release I covered earlier this year. The value is that every channel, conversation, and agent reads from and writes to the same customer understanding, so learning compounds instead of resetting with each campaign.

2. Braze: Decisioning With Guardrails

At Forge 2026, Braze pushed its AI from content creation into campaign operations.

The Bet: Let agents make per-customer decisions, and wrap those decisions in governance.

What shipped: BrazeAI Decisioning Studio Go, self-service agents that pick timing and creative per customer. Agentic Standards, an automated QA layer that checks campaigns against your rules. BrazeAI Operator Connect, built on the Braze MCP Server, which lets tools like Claude and ChatGPT act inside Braze while keeping user-level permissions. Plus Conversational Agents (beta) across WhatsApp, SMS, RCS, and web.
The caveat: Much of this launched in beta, with general availability dates projected for October.

The Architect's Take: Braze's standout idea is governance as a product. Agentic Standards admits something the industry doesn't say enough: autonomous campaigns need automated QA, or speed becomes risk.

3. Klaviyo: Going Headless

Klaviyo's K:BOS story was about opening the platform to agents wherever they run.

The Bet: The CRM becomes infrastructure, and the interface becomes whatever AI tool the marketer already uses.

What shipped: A headless release exposing 260+ MCP tools and 490+ APIs to Claude, ChatGPT, and other AI systems. Earlier in the year, Composer and Customer Agent were positioned as agents that share one real-time profile, where each one's actions enrich the data the other uses. Customer Hub also gained onsite web personalization.

The Architect's Take: Klaviyo has the most aggressive surface area play. It's a strong fit for e-commerce teams who want their AI assistant, not a dashboard, to be the primary workspace.

4. Salesforce: The Enterprise Data Layer, Everywhere

At Dreamforce 2026, Salesforce kept expanding Agentforce and pushed its data and governance model outward into tools like Claude and Slack.

The Bet: Your CRM data, permissions, and workflows should follow you into any AI surface. For marketing, that shows up as "headless marketing," which exposes Marketing Cloud functionality through MCP and APIs.

The Architect's Take: Salesforce's strength is breadth: sales, service, and marketing on one data model. The trade-off is the usual one, which is that enterprise breadth comes with enterprise complexity.

The Convergence: Three Things Everyone Agrees On

Strip away the branding and the four platforms agree on more than they disagree:

MCP is now table stakes. Every vendor here ships an MCP server or a headless layer. I wrote about this in The Power of MCP, The Command-Line Marketer, and The Multi-Server Marketer. What was an early experiment in the spring is a standard checkbox by the fall.
Conversational channels are the new frontier. RCS and WhatsApp showed up in nearly every launch, which is exactly the shift I predicted in Beyond the Green Bubble. Messaging is shifting from one-way pushes to two-way conversations, and conversations need memory to work.
Agents need shared context. Whether it's Iterable's intelligence loop, Klaviyo's shared profile, or Salesforce's data layer, everyone agrees that agents working from separate data will give customers a disjointed experience.

Where They Actually Differ

The real question isn't whether a platform has memory. It's where the memory lives and who learns from it.

Headless-first (Klaviyo, Salesforce): The platform is a data and action layer. Intelligence lives largely in the external AI tools the marketer brings.
Governance-first (Braze): Agents make decisions, and the platform polices them.
Context-first (Iterable): The platform itself accumulates understanding across channels, conversations, and outcomes, and it still exposes that context outward through MCP.

None of these is wrong. But for architects, the choice affects everything downstream. If the intelligence lives outside the platform, you own the job of keeping it consistent. If it lives inside, you need to trust and audit what the platform is learning.

Conclusion: Design for Compounding

My advice to teams evaluating the stack in 2026 is to ask one question in every demo: "Does what the system learned yesterday make tomorrow's message better, automatically?" Copywriting assistants are a commodity now. The lasting advantage comes from compounding context, where every send, reply, and conversion makes the next interaction smarter.

That's the race all four of these companies are running. It's going to be a fascinating 2027.

Want to dig deeper? Read Iterable's full Fall 2026 announcement and compare it with the Braze Forge and Klaviyo K:BOS recaps.