Agentic Marketing, Defined by Who Owns the Loop

Agentic marketing is marketing in which software agents own the loop of observing customers, deciding what to do and acting on it, inside guardrails a person set. The vendor definitions ranking on page one come from flow builders and journey suites, so they place the agent inside the workflow. The honest test is who owns the decision.

We build agentic marketing for ecommerce, so read this as informed but interested. The incumbents’ definitions are quoted in their own words and treated fairly. The test in the last section before the FAQ is ours, and it is the one a buyer needs.

What agentic marketing means

Strip the word back to its root. An agent is something that acts on its own toward a goal. Marketing has always had a loop: observe what customers do, decide what to say to whom, act by sending it, then observe again. Marketing automation put the act step on rails and left observe and decide with people, who expressed their decisions in advance as segments, triggers and branches. Generative AI made the drafting inside that loop faster. Agentic marketing is the point at which the loop itself, observe, decide and act, moves from the person to the system, with the person setting the goal and the guardrails and judging the results.

That is a narrower definition than most of what ranks for the term, and the narrowness is the point. An agent that drafts a campaign for a person to send has not taken over the loop. An agent that picks the best send time inside a journey a person designed has not taken it over either. Both are useful. Neither is agentic marketing in the full sense, because the decision about who gets what, and whether they get anything, is still made weeks earlier by a person in a builder.

Who is writing the definition on page one

Search the phrase and the first results are a consultancy, a business school, an ad platform and vendors of marketing software. Salesforce’s guide describes the split as marketers who “Set the strategy, define the guardrails, and hand off execution,” with agents that “execute, adapt, and optimize in real time.” Its definition of an AI marketing agent is a system that will “autonomously reason through data, make decisions, and execute marketing tasks like segmentation, personalization, and campaign activation,” with the caveat that “The human needs to teach the agent.” ActiveCampaign now sells its plans under the heading “Autonomous marketing plans,” and its AI page describes what that delivers as “Campaigns drafted. Issues caught. Insights delivered.”

Read those carefully and a pattern appears. In both definitions the agent lives inside the journey, the campaign or the workflow that the platform already sells. The strategy is human, the journey is human, and the agent executes and optimizes within it. That is a real and valuable step, and it is also exactly the definition a company would write if its product were a journey builder. Nobody is being dishonest. The category is simply being defined by the companies with the most to protect, and their definition keeps the human-designed workflow at the center.

What changes at each step toward agentic

It helps to see the incumbents’ definition as one stop on a line rather than the destination.

• Agents that execute inside a workflow. Send-time optimization, predictive segments, subject-line and content drafts, gap-filling campaigns. The person designs the journey; the agent tunes it. Klaviyo, ActiveCampaign, Braze and Salesforce all describe this step.
• Agents that build the workflow. The person states a goal in plain language and the agent assembles the journey for approval. ActiveCampaign’s “builds an entire campaign for you,” Salesforce’s “describe your goals in natural language,” Klaviyo’s Composer, which will “build the audience, draft the content, and map out the send strategy,” and Braze’s “let AI agents build on-brand content and campaigns for you” are this step. Faster to build, still a journey, still one design applied to a segment.
• Agents that replace the workflow. There is no journey to approve, because the agent decides per customer and per moment whether to act at all and what to do. The person sets guardrails and reads results. This is the step where the loop changes hands, and it is what the word agent means.

Examples of agentic marketing in ecommerce

The clearest examples are the interventions every Shopify brand already runs as flows, run instead as decisions.

• Cart abandonment. A flow sends the same three-email sequence, with the same discount, to everyone who abandons. An agent decides for this shopper whether a reminder is warranted, whether a discount is needed at all, which channel, and when, and it learns from what happens.
• Win-back. A flow fires at day 60 for everyone. An agent watches for the signal that a specific customer is cooling and acts on it, which for some customers is day 20 and for others never.
• Quiet hours and frequency. A flow enforces a global cap. An agent holds a message for a shopper who was already contacted yesterday and would convert better on Thursday.
• Flash sales. A campaign blasts a list. An agent chooses who hears about it, on which channel, and who is spared because they bought at full price on Monday.

Every one of those is a decision nobody wrote a rule for, which is the practical test of agentic: could a person have written this rule in advance? If yes, it is automation, however clever. If no, the system decided.

What agentic marketing is not

It is not a chatbot, though a chatbot can be an agent for support. It is not ChatGPT drafting a newsletter, which is generative AI at a person’s direction. It is not a predicted send time, a churn score or an AI subject line, which are features inside a flow. And it is not a workflow builder with an agent that fills gaps in your calendar, which is what ActiveCampaign’s “spots upcoming moments and builds an entire campaign for you” describes. All of these are worth having. None of them moves the decision about who gets what out of the builder.

The agentic stack for ecommerce

Ecommerce ran on a SaaS stack of tools people operate: a data layer, a decision layer expressed as segments and flows, and an execution layer of email and SMS. The agentic stack keeps the data layer and replaces the middle. Relvino runs the loop Observe → Decide → Act on a store’s owned channels: it watches live shopper signals, decides offer, timing and channel for that one shopper in about 80 milliseconds, and executes across email, SMS and on-site. Guardrails are set once, including margin floors, channels and brand voice, and inside them 100% of flows run without a human in the loop.

The decisions start from priors rather than from zero, because the engine is a Large Retail Model trained on 7M+ data points across 10K+ retailers and 1.78M shoppers. Brands replacing an incumbent see 2–6× ROI in 30 days and up to 10× revenue uplift year over year, and because the agent decides whether to send at all, the registered efficiency claim is 80% less spam and lower send costs. One cleared example: POV Beauty, 2X fewer emails, same revenue. That last number is the one that separates agentic marketing from automation with better drafts. Automation makes sending cheaper. An agent that owns the decision sends less.

Automation, generative AI and agentic marketing at a glance

• Who designs the journey · Marketing automation: A person, in a builder · Generative AI in your platform: A person, with AI drafts · Agentic marketing: Nobody; there is no journey
• Who decides per customer · Marketing automation: The rules the person wrote · Generative AI in your platform: The rules the person wrote · Agentic marketing: The agent, in ~80ms, inside guardrails
• What the human does · Marketing automation: Builds and maintains flows · Generative AI in your platform: Reviews drafts and tunes flows · Agentic marketing: Sets guardrails, judges results
• What improves results · Marketing automation: More flows, more tests · Generative AI in your platform: Better copy, better send times · Agentic marketing: The policy learns from every send
• Proof that fits it · Marketing automation: Open and click rates · Generative AI in your platform: Time saved drafting · Agentic marketing: Side-by-side revenue in a 14-day pilot

How to tell whether a product is agentic

Three questions, asked of the product rather than the brochure. First, is there a journey to approve? If the answer is yes, a person still owns the decision and the agent works inside it. Second, what is the unit of decision, a segment or a shopper? Segments are rules written in advance; a shopper is a decision made now. Third, what does the bill measure? Contacts stored and sends made are meters for a tool a person operates; a system that owns the loop can be priced against what it produces. The buying checklist that follows from these is in agentic marketing platform, the four tiers of agents are sorted in AI marketing agents, and the email-channel version is in autonomous email marketing.

Frequently asked questions

Which agentic AI platform is best for marketing?

It depends on which step you are buying. If you want agents that execute and draft inside journeys your team designs, the platform you already use likely ships that: Salesforce, Braze and Klaviyo all describe agents working inside their workflows, and ActiveCampaign describes its AI as building campaigns for you. If you want the loop itself handed to the system, so that there is no journey to approve, you are buying autonomous decisioning, and the right way to pick one is a bounded side-by-side pilot with revenue as the score, not a demo.

What are the 7 types of AI agents?

The textbook taxonomy lists simple reflex agents, model-based reflex agents, goal-based agents, utility-based agents, learning agents, hierarchical agents and multi-agent systems. In marketing terms, a flow is a simple reflex agent (if this trigger, then that send), a send-time optimizer is model-based, and an autonomous decisioning system is a learning, utility-based agent: it chooses the action with the best expected outcome for that shopper and updates from the result.

What is an example of agentic?

A system that decides, for one shopper who just abandoned a cart, whether to send anything, whether a discount is needed, which channel to use and when, and then sends it and learns from the outcome, with nobody having written a rule for that shopper. The contrast is a cart-abandonment flow, where a person wrote the sequence in advance and every abandoner receives it.

What are some examples of how agentic AI can be used in marketing?

On owned channels: cart and browse abandonment decided per shopper, win-back timed to each customer’s own cooling signal, frequency and quiet hours held per person rather than as a global cap, and flash sales sent only to those who would not have bought anyway. Inside platforms: agents that draft campaigns, fill calendar gaps and tune send times. In paid media: bidding and budget agents. The first group replaces the flow; the others work inside it.

Is ChatGPT an agentic AI?

On its own, no. ChatGPT drafts, analyzes and answers at a person’s direction and does not act on customers. It becomes part of an agent when it is wired to tools and goals and allowed to take actions, which is what agent frameworks and platform agents do with models like it. For marketing, the test is whether anything reaches a customer without a person sending it.

How long does it take to migrate from Klaviyo to Relvino?

About 30 minutes for the technical cutover: authenticate the sending domain, install the Shopify app, and historical data ingests on its own. Nothing is rebuilt, because there are no flows to recreate. Revenue runs on a different clock: run a 14-day pilot with Klaviyo paused rather than cancelled, and compare the two side by side before deciding.