For years, marketing automation has been about helping teams execute faster. Agentforce for marketing points to something more consequential: a future where AI agents take on parts of execution and decisioning, while marketers move further upstream into strategy, architecture, governance, and business outcomes.

Marketing automation has gone through several major shifts over the past two decades. We moved from batch campaigns to triggered communications, from static lists to behavioral segmentation, and from individual sends to increasingly sophisticated customer journeys. Each evolution removed some of the manual work involved in getting a campaign from an idea to execution.

Yet the fundamental operating model remained largely unchanged. Marketers designed the strategy, defined the rules, built the audience, created the journey, configured the decisioning, launched the campaign, and then analyzed the results. The technology helped marketers execute the plan.

Agentforce for marketing introduces the possibility of a different model, one where agents can increasingly participate in building, deciding, optimizing, and executing the plan itself.

Dreamforce 2026 will offer a closer look at where Salesforce is taking this next. 

In our recent pre-Dreamforce leadership roundtable, we explored what this transition could mean for lifecycle marketing teams, particularly as Salesforce continues to bring agentic capabilities into marketing workflows. The conversation pointed to a future where AI, far from replacing marketers, will shift their focus to areas where human judgment adds value.

How Agenforce changes marketing strategies 

From AI assistance to agent-led execution

The first wave of generative AI in marketing has largely been about assistance. It can help write copy, generate SQL, identify errors, summarize information, or accelerate repetitive tasks. These capabilities have obvious productivity benefits, but they don’t fundamentally change who owns the workflow. The marketer remains responsible for deciding what needs to happen and translating that decision into a series of actions inside the marketing platform.

Agentforce for marketing starts to challenge that model.

The discussion around Salesforce’s agentic capabilities is increasingly focused on areas such as campaign creation, segmentation, journey creation, decisioning, personalization, and optimization. Agents can increasingly take on portions of the work themselves. 

Lesley Higgins captured the distinction particularly well during the discussion, describing the evolution as moving toward “handing the keys over to the agent.” The agent may eventually be trusted to make some of the decisions that previously required a marketer to manually configure the platform.

That changes the role of the marketing automation team.

If an agent can build the segment and construct much of the journey, the marketer’s value increasingly comes from defining the objective, establishing the experience, determining the constraints, interpreting business context, and deciding where human intervention is required. The role moves further upstream, from operating the machinery to designing the system that operates the machinery.

The journey is becoming less prescriptive

One of the more interesting implications of agentic marketing is what happens to journey orchestration itself.

Traditional journey design assumes that marketers need to anticipate customer behavior in advance. We create branches for different scenarios, define wait periods, establish rules, and attempt to account for as many customer paths as possible.

That approach has served marketing well, but it also has an inherent limitation: customers rarely behave exactly as the journey designer expects.

Agentic decisioning introduces a different approach. Instead of attempting to explicitly define every possible path, marketers can establish the experience, business objective, available actions, and guardrails, while allowing an agent to determine the appropriate next action based on customer behavior and context.

During the roundtable, JB Bitra described this as a move toward journeys becoming less prescriptive. The system has more flexibility to determine how the customer should progress through it. Capabilities such as journey decisioning and next-best-action approaches are examples of this broader direction.

Lifecycle marketers may spend less time mapping every possible branch and more time defining the objectives and experience within which the agent operates.

Customer 360 is the starting point, not the complete decisioning context

There is a tendency to assume that if an organization has a unified customer profile, it is ready for AI-driven decisioning. In reality, an agent needs considerably more context than a customer record.

During the roundtable, Lesley highlighted the broader set of information an agent may need to make an appropriate marketing decision: active campaigns and offers, inventory, eligibility, contact frequency, consent, business priorities, permissions, and the specific actions the agent is allowed to take.

Consider a simple example. Knowing that a customer recently purchased a product is useful. Knowing that the customer recently purchased it, is not eligible for the next offer, has already received three communications this week, has opted out of a particular channel, and is currently part of another promotional campaign is far more useful. That distinction becomes critical when an AI system has the authority to act.

Agentic marketing requires decisioning context as well as customer data.

This also elevates areas that have historically been treated as operational concerns. Consent management, permissions, data quality, suppression logic, and business rules become part of the foundation that determines whether an agent can safely operate.

The more autonomy organizations give their agents, the less room there is for fragmented data and undocumented business logic.

Data quality becomes an execution issue

For years, data quality has been discussed primarily in the context of analytics and reporting. Bad data produces bad dashboards, inaccurate segmentation, and unreliable insights.

Agentic marketing raises the stakes.

When an employee makes a mistake because they misunderstood a data point, there is usually an opportunity for another person to catch it before execution. When an autonomous system is making decisions at scale, the same underlying data problem can be repeated across thousands or millions of customer interactions.

That makes the quality, accessibility, and structure of marketing data an execution concern.

As Matt Kelly noted during the discussion, AI is only as good as the customer context and data behind it. If the underlying foundation is fragmented or poorly governed, introducing agentic workflows can create and amplify problems. 

For organizations with long-established Marketing Cloud environments, this is particularly relevant. Years of journeys, SQL, data extensions, custom processes, suppression logic, and workarounds may have created a system that works, but is difficult for an agent to understand.

This is why the transition to agentic marketing is as much an architecture exercise as it is an AI exercise.

The new marketing skill: knowing what to delegate

The most important capability for marketing leaders may be learning how to determine what should be delegated to an agent and what should remain under human ownership. 

That distinction will vary by organization.

An agent may be well suited to generating segments, adapting journeys, identifying patterns, optimizing send times, or handling repetitive operational tasks. Strategic positioning, customer experience principles, commercial priorities, brand decisions, and high-impact exceptions may continue to require human judgment.

The challenge is delegation with accountability. 

That requires marketing leaders to define objectives, permissions, thresholds, escalation points, and guardrails before giving agents greater autonomy. In an agentic environment, governance cannot sit outside the operating model. It becomes part of the architecture itself.

From reporting events to anticipating events

Another potentially significant change is the role AI can play in campaign intelligence. Marketing teams have become extremely good at reporting what happened. We can measure opens, clicks, conversions, engagement, journey progression, and a long list of other metrics.

But do those signals always tell us about what will happen next?

During the discussion, Matt raised the possibility of agents evaluating active campaigns and helping marketers understand how those campaigns may perform going forward, potentially supporting decisions around whether to continue, change, or stop an initiative.

This points toward a more consequential use of AI in marketing. Instead of simply reporting that a journey generated a particular engagement rate, the system could increasingly help marketers understand whether the journey is contributing to the business outcome that matters.

Lesley also highlighted the opportunity to move beyond an overreliance on opens and clicks and use richer data and AI-driven synthesis to understand whether a journey is actually preceding outcomes such as purchases.

That shift, from activity reporting to outcome-oriented decisioning, may ultimately prove more valuable than campaign efficiency alone. 

The next interface for marketing may not be Marketing Cloud

What happens when marketers no longer need to work directly inside the marketing platform for every task?

Lesley described a future in which a marketer could have a conversation in Slack or another working environment, describe the audience and campaign requirements, and have an agent translate that request into the underlying work across Data 360 and Marketing Cloud. The marketer could then review the proposed segment, creative, journey, and execution plan. 

JB discussed the role that MCP and headless approaches could play in enabling this type of interaction, including use cases where Salesforce capabilities can be accessed through conversational interfaces. If this direction continues, the marketing platform becomes less of a destination and more of an underlying execution layer.

That has implications well beyond user experience. It changes how marketing operations teams think about skills, documentation, architecture, process design, and even how organizations structure their marketing technology environments.

What to watch for Agentforce for marketing at Dreamforce 2026

With Agentforce becoming an increasingly important part of Salesforce’s broader vision for the agentic enterprise, the most interesting Dreamforce announcements could be the capabilities that demonstrate how much autonomy marketing agents can realistically take on.

A few areas are particularly worth watching:

• More autonomous campaign execution: How far can agents move beyond assisting marketers and actually create, modify, and optimize campaigns?
• Dynamic journey decisioning: How much of the customer journey can be determined dynamically based on behavior, context, and business objectives?
• Richer decisioning context: How deeply can Agentforce combine customer, campaign, business, consent, and operational data to make decisions?
• Headless marketing experiences: Can marketers increasingly describe what they need conversationally and have agents translate that intent into work across Marketing Cloud and Data 360?
• Governance and accountability: As agents become more capable, how will organizations establish the permissions, guardrails, approval mechanisms, and accountability required to operate them safely at scale?

These questions will tell us more about the maturity of Agentforce for Marketing than any individual product announcement. We explored all of these questions in greater depth during Mavlers’ Lifecycle Marketing in the AI Era: A Pre-Dreamforce Leadership Roundtable, featuring Matt Kelly, Lesley Higgins, and Jyothsna “JB” Bitra.

Tune in to the full roundtable discussion here.