Key facts

• Generative AI is the AI you chat with: it writes, researches, summarises and analyses when you prompt it. Agentic AI takes a goal and does the work, planning, using tools and checking itself until it’s done.

• Agentic AI is built on generative AI, so this isn’t a choice between them. It’s a journey, and if your team already uses generative AI well, you’re already on it.

• There are four types of agentic AI, and they build on each other: agentic assistant, automation, agent and agentic system.

• For a marketing leader the shift is about your role and your team: you move from doing the work to orchestrating the agents and people who do it together.

• Most teams are strong with assistants and just starting on the rest, which is exactly where the opportunity is.

Generative AI and agentic AI get talked about as if they’re two competing things. They aren’t. Generative AI is the AI you chat with. Agentic AI does the work. And the second is built on the first, so the real question for a marketing leader isn’t which one to pick. It’s how far along the journey from one to the other your team has already travelled, and what that does to the way you lead.

Here’s what each one is, the four types of agentic AI you’ll actually use, what genuinely changes for you as a leader and how to take your team with you.

What is generative AI?

Generative AI is the AI you chat with: it writes, researches, summarises and analyses when you prompt it. It’s reactive and genuinely broad, and it waits for your next instruction before it does anything more.

This is the AI most marketing teams already use every day. You ask, and it answers: a draft subject line, a campaign concept, a summary of a long report, a quick piece of research, a first version of a landing page. It’s the chat and chatbot layer, and it’s more capable than people give it credit for. Its limit isn’t what it knows, it’s that it waits for you. It does one thing at a time and then stops until you prompt it again. You’re still the engine carrying the work from one step to the next, and generative AI makes each step faster.

What is agentic AI?

Agentic AI is the AI that acts. You give it a goal, and it plans, uses tools, checks its own work and keeps going until it meets the standard you set. It pursues an outcome rather than waiting for the next prompt.

Agentic AI changes the relationship. Instead of handing it a single task, you hand it a goal and a clear sense of what good looks like, and it works towards that goal on its own. The way I teach it, an agent has a brain, actions and tools: the model is the brain that holds context, the actions are the steps it takes and the tools are its hands. It plans the steps, calls the tools it needs, reviews its own output and keeps going until the job is done to your standard. That loop, pursuing a goal with limited supervision, is what makes something genuinely agentic. It’s also why the same underlying model can feel so different depending on how it’s set up: a chatbot answers you, an agent delivers an outcome for you.

So is it generative AI vs agentic AI, or both?

Both, and that’s the good news. Agentic AI is built on generative AI, the model is the brain inside the agent, so this isn’t a choice between them. If your team already uses generative AI well, you’re already on the path.

It’s tempting to frame this as generative versus agentic, but the more useful truth is kinder than that: you don’t have to choose. Agentic systems run on generative models, and the model is the reasoning engine inside the agent, so one grows into the other. The way I teach it, it’s a journey. You start with an assistant, you move to automation, then to agents and finally to an agentic system where people and agents work as one team. If your team is already fluent with generative AI, that isn’t a standing start, it’s the first stage already behind you. So the question isn’t which to pick. It’s how far along the journey you are, and what a good next step looks like for your team.

The four types of agentic AI

There are four types of agentic AI, and they build on each other: agentic assistant, automation, agent and agentic system. They’re a journey from staying in control of each task to designing a system that works alongside your team.

This is the structure I keep coming back to, because it tells you where you are and what’s next. Each type does a little more on its own than the one before it.

You can’t really skip stages, and you don’t need to. Most marketing teams are fluent with assistants and are just beginning with automations and agents. That’s a strong place to be, because the next step is always clear: take something you already do well with an assistant and turn it into a repeatable workflow.

What do you build these on?

The four types are about how much the system does for you. What you build them on is a separate question, and the tools sort by how hard they are to use against how much they do.

It’s worth keeping these two ideas apart, because the market blurs them. The tools run from AI assistant apps (Claude.ai, ChatGPT, Custom GPTs, Gemini), to pre-built agent platforms (Relay, Relevance AI, Lindy, Copilot Studio), to custom automation (n8n, Zapier, Make), to agentic coding and open source (Claude Code and the rest). Two agentic building tools matter most for a marketing leader. Claude Code is powerful and impactful, the serious build path. Claude Cowork is the one business people, not just engineers, can actually build on, which is why it has spread so fast. There’s a fuller, worked walk-through of the use cases these support in the companion piece on agentic marketing use cases.

What changes for you as a leader?

The change is your role and your team: you move from doing the work to orchestrating the people and agents who do it together, and your job becomes leading them through it. That’s the real work, and it’s more about people than technology.

With generative AI you’re still doing the work, just faster. With agentic AI the work gets done by a system you design and direct, so your job shifts from producing the output to orchestrating the agents and the people who produce it together. Once an agent can do part of the marketing job, it stops being a tool that sits beside the team and starts being part of the team. That’s why the first thing to rethink is the org chart, not the tool list. And it’s why the part that stays firmly yours is the input: the judgement, the brand, the fresh point of view the agents work from. AI can generate endless options from your input, but it’s your input, and finding input your competitors don’t have is the new marketing job.

The second half is your team, and this is the human part. The shift is emotional as much as technical. People often feel it as a threat before they feel it as a help, and they look to you for a sense of where it’s going and what good looks like. In one of my sessions a senior marketer put it honestly: “My role is changing. Tell me exactly how my role will look in one year.” I couldn’t give them certainty, and that’s the honest answer. What I could tell them is true for everyone: the capabilities of AI roughly double every seven months, so no one can hand you a fixed map. You move with it, you embrace it, and you lead from the front. Adoption stalls far more often on fear than on skills, so the leaders who do this well lead the change on purpose: they set the direction, show what good looks like and bring the team with them.

The encouraging part is that this gets easier every month. The tools keep getting simpler and the capability keeps climbing, so the barrier isn’t capability any more. It’s whether marketing leads the shift or waits for someone else to. It’s worth knowing the stakes: Gartner predicts that more than 40% of agentic AI projects will be cancelled by the end of 2027, usually because they were deployed without that leadership, not because the technology failed. The leaders who bring their people with them are the ones who land it.

When should you use generative AI or agentic AI?

Use generative AI for one-off work a person will review. Use agentic AI for repeatable work where you can define what good looks like and trust the system to reach it. The clearer your definition of done, the more agentic you can go.

A simple way to decide. If the task is a single piece of work that a person will check before it goes anywhere, generative AI is the right tool, and it’s probably already in your stack. If the task is repeatable, has clear success criteria and you can describe what good looks like well enough to trust the result, it’s a candidate for an agent. The best place to begin is wherever your definition of done is easiest to write down, because that definition is exactly what the agent works towards.

Generative AI and agentic AI at a glance

Frequently asked questions

Is ChatGPT generative AI or agentic AI?

Mostly generative. ChatGPT is the chat layer: you prompt it and it responds. It can take some agentic steps in its agent mode, but on its own it’s an assistant that answers you. Agentic AI holds a goal and acts towards it across steps without a prompt each time.

What are the four types of agentic AI?

Agentic assistant, automation, agent and agentic system, and they build on each other. An assistant gives you faster output on a single task with you in control. An automation runs a repeatable multi-step workflow on a schedule or trigger. An agent pursues a goal and adapts. An agentic system is people and agents working as one team. Most teams are strong on assistants and just starting on the rest, which is where the opportunity is.

Is agentic AI just generative AI with extra steps?

It’s built on the same kind of model, so in one sense yes, and the extra steps are the whole point. A generative tool produces an output and waits. An agentic one pursues a goal: it plans, uses tools, checks its own work and keeps going until the goal is met. The model is similar. What changes is that it acts rather than waits.

Does agentic AI use the same models as ChatGPT or Claude?

Yes. The large language model is the reasoning engine inside the agent. The stronger the model’s reasoning, the better the decisions the agent can make on its own.

Where should a marketing leader start?

Start with one repeatable workflow where you can write a clear definition of done, and lead your team into it openly. Begin with the assistants you already have, then move up to automations and agents as your confidence and your guardrails grow.

Join the agentic journey

Generative AI made content cheap to produce. Agentic AI makes the workflow, and the shape of your team, the thing you design. The leaders who do well over the next couple of years won’t be the ones with the cleverest prompts. They’ll be the ones who understand the difference well enough to redraw the team around it, who keep their own judgement and point of view as the input the agents work from and who lead their people through the change with them.

That’s the work of becoming agentic, and it’s what we do with senior marketing leaders inside the Agentic CMO Accelerator.