On September 29, 2026, Salesforce published its Dreamforce recap for Marketing Cloud, built around five ideas: campaigns that improve themselves, adaptive channels, B2B signals turned into pipeline, visibility into what AI answer engines say about your brand, and "headless" marketing reachable through MCP. If you run Salesforce Marketing Cloud Engagement (SFMC) today, there are two ways to get this wrong: swallow the pitch whole, or wave it away. Here we walk through each announcement, separate what is concrete from what is still a direction, and list the work worth starting now so you are ready when these capabilities land.

What Salesforce actually announced

First, the facts as Salesforce presents them. The source article gives no availability dates, so keep that in mind before you plan anything around them.

1. Campaigns that improve themselves

The Campaign Agent assembles a campaign, recommends audiences and creates personalized content. After launch, it evaluates customer context and intent, prioritizes between competing campaigns and suppresses messages that should not go out. This is the most significant piece: the trade-offs between competing campaigns, which are still largely manual, would shift to the agent.

2. Adaptive channels

Salesforce lists Adaptive Web: Quick Setup, Personalized Paths in Flow, RCS messaging and Mobile Flash Sending. The common thread is an experience that reacts to behavior instead of following a fixed journey.

3. B2B, AI visibility and headless marketing

On the B2B side, Account Discovery shows engagement across buying committees, while the Piper and Hunter agents qualify inbound interest and pass context to sales reps. On the brand side, Palmata and Contentful aim to help you understand and influence what answer engines say about you. Finally, Headless Marketing exposes Marketing Cloud to external tools through MCP servers.

Our take: a clear direction, not a big bang

If you are on SFMC Engagement, nothing forces you to migrate tomorrow. But the message fits the recent Marketing Cloud Next releases: the platform keeps moving toward Data Cloud, agents and continuous orchestration rather than scheduled campaigns.

The key insight: an agent is only as good as your data, your consent records and your suppression rules. AI amplifies the quality of your foundations, good or bad.

Why your data decides everything

An agent that "suppresses messages that should not go out" needs to know who must not be contacted. In many SFMC accounts, that knowledge is scattered across exclusion lists, Data Extensions, consent fields and rules buried in automations.

Centralize your suppression rules

A good first exercise is to build a single suppression Data Extension, fed by a SQL query in Automation Studio. Here is a baseline to adapt to your own data model:

SELECT
  c.SubscriberKey,
  c.EmailAddress,
  'Consent withdrawn' AS Reason
FROM Consent_Master c
WHERE c.EmailConsent = 'N'

UNION

SELECT
  s.SubscriberKey,
  s.EmailAddress,
  'Unsubscribed' AS Reason
FROM _Subscribers s
WHERE s.Status IN ('Unsubscribed', 'Held', 'Bounced')

Whether or not you ever adopt the Campaign Agent, this table becomes your single source of truth for exclusions and saves you from costly mistakes, particularly around GDPR.

Measure marketing pressure

Arbitrating between campaigns also means knowing how often people are really contacted. This query on the sends data view gives you a first 30-day indicator:

SELECT
  SubscriberKey,
  COUNT(*) AS Sends30d
FROM _Sent
WHERE EventDate >= DATEADD(day, -30, GETDATE())
GROUP BY SubscriberKey
HAVING COUNT(*) > 8

The contacts that surface here are your candidates for a frequency cap, which an agent would need to respect or which you will keep enforcing yourself.

RCS and adaptive flows: what to test first

RCS is appealing because it brings rich messages (carousels, buttons) to mobile. But every extra channel also adds complexity to consent and measurement. Start with one narrow use case, such as an order follow-up, with a holdout group and a success metric defined before you send.

For personalized paths in Flow, apply the same discipline: document the branches, name your decision rules clearly and keep a baseline journey to compare results against.

MCP and headless marketing: opportunity and caution

Exposing Marketing Cloud through MCP lets people work from the tool or agent they prefer without losing context. That is attractive for ops teams juggling many tools. It is also a governance question: who can read which data, who can trigger which action, and with what audit trail? Define least-privilege roles and log actions before you open the door to agents.

A practical action plan

In the next 30 days

Audit your suppression lists and consent data, centralize them, then measure marketing pressure per contact. These three jobs pay off with or without an agent.

In the next 90 days

Pick one pilot use case (RCS or an adaptive journey), define how you will measure it and prepare a governance brief for any agent access. Want a second pair of eyes on your account? You can get in touch with us here.

Key takeaways

Nothing is dated. Salesforce gives no availability timeline, so treat these announcements as a direction, not a roadmap.

Data comes first. Consent, suppression and contact frequency decide whether any campaign agent succeeds.

Centralize suppressions. One SQL-fed Data Extension reduces the risk of non-compliant sends.

Test small. Pilot RCS and adaptive journeys on one narrow use case, with a holdout group.

Govern MCP access. Least-privilege roles and logging before you open anything up to external agents.