Building an Agent Is the Easy Part

• Agentic AI is evolving from assistance to autonomy. Marketing teams are moving toward more connected, orchestrated workflows.
• Effective agents need a strong foundation. Reliable data, advertising intelligence, and governance are essential.
• Intelligent decisions drive performance. An agent’s value depends on its ability to improve advertising outcomes.
• Scaling AI requires organizational change. Teams must rethink workflows, responsibilities, and skills.
• Building and buying will go hand in hand. Custom agents and existing technology will play complementary roles.

You don’t have to look far to find predictions about the agentic future of advertising.

AI agents will run campaigns, optimize budgets in real time, and fundamentally reshape customer journeys. Gartner predicts that 60% of brands will use agentic AI to enable one-to-one interactions by 2028. And according to BCG, 96% of CMOs already believe AI is driving an end-to-end transformation of their function.

The technology is moving incredibly fast.

But transformation is proving harder.

The same BCG research found that 42% of CMOs are still using generative AI primarily to assist people with discrete tasks. Fewer than a third have progressed to agent-led workflows. Just 8% have multiple agents operating autonomously.

The real story of agentic transformation is unfolding between what is technologically possible and what organizations can actually operationalize.

From assistance to autonomy to orchestration

It helps to think about agentic transformation as a journey instead of a destination.

The first stage is assistance. AI helps a marketer do something: analyze performance, summarize data, create content, identify opportunities, or recommend actions. The person remains the operator.

Then comes autonomy. Instead of simply recommending what should happen, AI can complete the work. An agent might identify an underperforming campaign, determine the appropriate response, and execute the change within predefined parameters.

The third stage is orchestration. Multiple specialized agents and systems work together across workflows that span planning, activation, optimization, measurement, and beyond, with humans increasingly setting goals, defining constraints, and managing exceptions rather than executing every individual task.

Our relationship with technology changes at every stage.

Giving AI permission to summarize a report is very different from giving it permission to move $1 million of media budget. Connecting multiple agents across an enterprise introduces another level of complexity.

Marketing organizations aren’t progressing through these stages uniformly either. One team might already be building agents while another is experimenting with ChatGPT. Even within the same team, individuals can have radically different levels of AI maturity.

That makes transformation as much an organizational challenge as a technology one.

Building versus buying is becoming the wrong question

The maturity curve raises another question: how companies want to access the technology.

Some will continue working primarily through advertising platforms. Others will build their own agents and experiences. Increasingly, the answer will be somewhere between the two.

Building your own is understandably appealing. Foundation models have made it easier than ever for companies to create an agent and shape the experience around their own processes, with the potential to reduce their reliance on traditional advertising software and the costs that come with it.

But we’re quickly discovering something important:

Building the agent is often the easy part.

Making that agent genuinely good at marketing is considerably harder.

There are four challenges marketing leaders should be thinking about.

1. Give agents the right foundation

An agent is only as useful as the information and capabilities it can access.

Marketing data remains extraordinarily fragmented across publishers, retail media networks, ad platforms, analytics environments, commerce systems, customer data, and internal business intelligence.

Connecting it once isn’t enough. Connections change. APIs change. Permissions change. Taxonomies need normalizing. Data needs governing.

Salesforce’s 2026 research found that 98% of marketers encounter barriers to personalization, with data among the most common problems. Agentic AI doesn’t magically remove that problem. If anything, autonomous execution makes getting the foundation right more important.

2. Don’t confuse automation with performance

There is a big difference between an agent that can change a campaign and one that knows which change will improve it.

Advertising is a game of marginal gains played at enormous scale. A percentage-point improvement in an optimization decision can translate into millions of dollars.

Therefore, the next competitive question won’t be, “Can my agent take this action?”

It will be: How intelligent is the decision behind the action?

Years of forecasting models, experimentation frameworks, optimization algorithms, and advertising expertise don’t suddenly become irrelevant because an LLM can call a publisher’s API.

3. Prepare for the operational burden

Running agents in production requires ongoing maintenance, oversight, and clear ownership.

Who determines what an agent can do? What happens when something goes wrong? How are actions audited? How do organizations control model and token costs? Who maintains the connections and underlying tools as they change?

This is one reason we’re seeing growing demand for Forward Deployed Engineers — people who sit between technology and the business and turn AI capability into working production systems.

For marketing leaders, this means planning for the people, skills, and resources needed to keep agents working effectively long after the initial build.

4. Redesign the organization, not just the technology

As agents take on more work, some of the hardest questions are about the people managing that work.

If an agent can monitor and optimize thousands of campaigns simultaneously, what does the campaign manager do?

Who builds agents? Who approves them? Who owns their performance? Who decides when humans need to intervene?

And how do you democratize AI without creating hundreds of disconnected agents following different rules?

For agencies, these questions become even more complicated. Standards and governance have to work across teams, clients, and potentially thousands of users.

Agentic transformation therefore can’t simply be an IT project or a series of AI pilots. It requires companies to rethink workflows, responsibilities, skills, and ultimately the shape of the marketing organization.

Start with the work, not the agent

For marketing leaders figuring out what to do next, three questions offer a useful starting point.

First: what work should disappear? Look at the outcome you’re trying to achieve and ask which manual steps an agent could remove. Consider what your team could do with that time back.

Second: what would you need to trust AI with that work? Think through the data, intelligence, permissions, governance, and human oversight required. These should be part of the design from the beginning.

Third: what should you own? Decide where building your own technology creates a meaningful advantage and where existing infrastructure can give your agents the capabilities they need. For many organizations, the answer will involve a combination of both.

At Skai, we see our role as helping companies make that transition from wherever they are today. That means opening up our data, advertising intelligence, and execution capabilities so teams can draw on them through our platform, in their own agents, or through a combination of the two. It also means bringing human expertise to help with implementation, adoption and the changes to how teams work.

Building an agent is an accessible starting point. Turning it into a reliable part of an advertising operation requires sustained investment in the technology and the people around it. The opportunity for marketing leaders is to shape that future deliberately, deciding what to automate, where to build, and how to equip their teams to lead.

Ready to learn more? Skai is here to help you better understand what agentic transformation actually looks like inside your own operation. Contact us today.

---

Frequently Asked Questions