The agentic AI era is here. Where should CMOs focus right now?

What are CMOs actually navigating right now?

François Gaumond:

The honest answer is that most marketing leaders I work with are caught between two speeds: leadership and business are demanding faster go-to-market, more personalization, and leaner operations through GenAI and agentic AI simultaneously, while those same leaders are managing real organizational constraints: technology backlogs, new data governance requirements, and a genuine shortage of people who have actually scaled GenAI and agentic inside a marketing function before.

Not to mention, all of this is happening while there is the expectation to continue to deliver business-as-usual.

That gap between ambition and execution shows up in the research. Deloitte’s State of Generative AI in the Enterprise found that more than 70% of organizations are experimenting with or deploying generative AI, with marketing among the leading adopters.¹ But experimentation is not transformation.

What we’re seeing in practice is a lot of good intention, strategic thinking documents, and a whole lot of isolated prompting. We’re seeing people using AI in their day-to-day, but not in a systematic way that is embedded into their overall process.

I was reading Anthropic’s AI Fluency Index, which reinforces this. They found that the most common expression of AI fluency is augmentative: people are treating GenAI as a thought partner rather than delegating work entirely. While 85% of users iterate and refine AI outputs, they are far less likely to question the reasoning or identify missing context, particularly when the model produces something that looks complete.²  

The percentage of organizations that have genuinely coordinated agents across their marketing function to deliver repeatable, scalable value is still very low. That creates a specific frustration. Leaders see what’s possible, they’re getting pressure from above, but they don’t have a clear operational roadmap and risk sign-offs to close the gap.

The hype makes it harder, not easier, to figure out where to start.

The organizations that will pull ahead are the ones that stop treating this as a technology adoption problem and start treating it as a new operating system. Those require very different solutions.

Why this wave of AI feels different for CMOs

Lauren Bradeen:

I’ll add some context that I think a lot of CMOs will relate to: marketing leaders aren’t new to transformation. In the last 15 years alone we’ve navigated the digital marketing revolution (I still remember the DMP. RIP), the first wave of AI, a decade of big MarTech investments, and now generative and agentic AI.

Each shift asked something new of us, but this one feels categorically different. The speed, the hype, and the genuine potential are unlike anything we’ve seen before, and that combination puts marketing leaders in a uniquely uncomfortable, but exciting, position.

On our end, we’re so excited for the outcomes. Our vision is clear. But a meaningful amount of the real agentic progress is outside my direct control. We all know that in a large global organization, governance and risk don’t move at the pace of technology.

I’ll be honest: that’s tough to swallow at times. And yet the stakes of getting it wrong are real, and I think they’re underappreciated in a lot of the agentic conversation.

Think about what’s actually at risk.

• Client and employee data flowing through AI systems that weren’t designed with your specific compliance requirements in mind.
• Brand voice being replicated at scale by agents that haven’t been properly trained on your guardrails producing content that’s off-message or factually wrong and published before a human catches it.
• IP exposure when agents draw on third-party content without clear indemnification policies.
• Regulatory risk that varies by market, which in a global organization means your governance framework has to work across jurisdictions simultaneously.
• Emojis and em dashes... everywhere.

Any one of those is a serious problem. In combination, they can do lasting damage to client trust and brand equity that no campaign can undo.

So I hold both things at once: genuine urgency and genuine care about getting it right.  

The organizations skipping governance are accumulating risk they haven’t priced yet. And despite all of that, there is still so much we can do within what’s available today. I’m lucky to work at Deloitte in that regard. We’re at the forefront of this, there is real capacity to be unlocked right now for higher-value work, and honestly, we’re having fun while we do it.

There’s still significant value to unlock

François Gaumond:

That’s an important point. There is still significant value accessible to teams today, even while the larger governance and infrastructure questions are being resolved.

As we’ve partnered with Lauren’s marketing team, we started by understanding their go-to-market process and its actual barriers. They’d just come through a significant transformation, so the strategy, KPIs, targets, artifacts, and MarTech architecture were all documented. They had examples of what great looks like and what doesn’t.

From “lonely agents” to orchestration

Lauren Bradeen:

About 18 months ago, we built a strategy anchored on efficiency, growth, and team impact. At the time, we didn’t yet have Deloitte Digital involved, so we were forging our own path.

We started with tools. What exists that can materially improve how we deliver? Our internal creative agency was the first place we started. We now have access to inventory we never could have imagined with stock imagery and our production time has significantly decreased.

Then we started with prompting and actually built a prompt book for each step of our go-to-market process, from the brief, to media plans, to insight generation from a report. It was essentially a manual version of an agent, telling the marketer what to prompt an LLM.

Then we built a few standalone agents, mainly for content and captions. But to borrow a term from my friend Michael Letsche, they were “lonely agents”: isolated AI capabilities that weren’t connected to anything broader. We were doing everything we knew how to do and we were seeing solid impacts, but it still felt like we were fingerpainting compared to what agents and agentic can do.

So we engaged Deloitte Digital, and now we have a clear path to real agentic orchestration. The structured documentation we’d built, including examples of our strongest briefs and media plans, became the grounding foundation for the agents we’re building. That’s the part that often gets skipped.  

Foundational work matters

François Gaumond:

The platform environment is evolving rapidly to support exactly this, both internally and externally. Anyone who has been at a recent tech conference would have heard about all kinds of innovations and agents.

The organizations that have done the foundational work, as Lauren’s team has, will be positioned to move very quickly as governance is resolved and capabilities can be onboarded with speed.

The killer basics: What should a CMO actually do this quarter?

Lauren Bradeen:

Three things.

1. The team

Find the people on your team who are obsessed with this. Everyone has some. They will be a bigger unlock than any formal plan.

You can start a taskforce or put them into roles, but the people who are testing, learning, and testing tools just because they love it are critical. Pair this with a cross-functional group of marketers embedded in design and testing. If you don’t design it to be useful from the start, your agents won’t be adopted.

2. Strategy and use cases

GenAI shouldn’t be standalone. It should be embedded into your strategy and operating model.

Get clear on how you will measure success. Document your processes, artifacts, and data sources. This builds your context layer, which you need before you scale anything.

Anchor your GenAI strategy to your operating model, not a technology roadmap. B2B and B2C teams face different constraints. Don’t borrow someone else’s use cases. Start with what slows you down most and focus there.

3. Design governance in from the start

Brand integrity, IP risk, accuracy, and compliance are not things to retrofit.

Human-in-the-loop doesn’t mean just reviewing and pressing send. It means owning the inputs and strategically assessing what comes back. As marketers, we need to own our creativity and judgment or else marketing plays to the average instead of breaking through.

A call to action for CMOs

François Gaumond:

First, stop waiting for the technology to be fully ready and start investing in organizational readiness. Operating model design and change management take time, and they need to start now.

Second, ignore the noise around definitions. Whether something is called an agent or agentic matters less than whether it delivers real value.

Third, in parallel with near-term use cases, look at your entire operating model and how it could be reimagined with agentic AI. That’s where long-term transformation happens.

Navigating your own AI journey?

Connect with Deloitte’s marketing leaders to get unstuck.