96% of CMOs Say AI Is Transforming Marketing. Only 8% Run Agent-Led Campaigns.

Marketing leaders rarely agree on anything as completely as they now agree on AI. When BCG surveyed 300 CMOs for its 2026 report, 96% said AI is driving an end-to-end transformation of their function.

The next number in the same survey is 8%. That's the share running campaigns in which multiple AI agents operate autonomously. Another 42% use generative AI only to help people with discrete tasks, which describes a lot of teams using a chatbot for subject lines and an image tool for ad visuals.

BCG isn't alone in seeing a gap. Two other large 2026 surveys point the same way, although they measure different things, so the numbers shouldn't be read as one trend line.

Different questions, same pattern: wide agreement on where marketing is going, and much thinner evidence of teams actually operating that way.

What "agentic" actually means

The word gets stretched, so it helps to be specific. A generative AI tool helps a person produce something faster. An agentic workflow observes a signal, makes a decision, takes an action and adjusts based on the result, with people reviewing the system instead of approving each step.

Imagine a buyer visits a pricing page twice in a week. A tool can help a marketer write a better follow-up email. An agent could spot the visits, choose the relevant campaign, tailor the message, send it at a sensible time and change its approach depending on the response.

The first makes a person more productive. The second changes what a team of the same size can do. BCG says that in its recent client engagements with leading CMOs it has seen cost efficiency improve by 20% to 30%, marketing ROI triple and campaign cycle times improve tenfold. Treat those as what BCG has observed with selected clients, not as survey averages or a typical result. The report itself separates engagement observations from survey data.

Why teams get stuck at the tool layer

They're measured on savings

40% of CMOs told BCG that their organisation mainly holds them accountable for cost savings and efficiency. That pushes AI budgets towards use cases that are easy to defend: cheaper content, faster image production, automated reporting. You can show the cost per asset before and after.

Redesigning how campaigns run is harder to justify. It costs more, takes longer and shows up in revenue numbers marketing may not fully own. Choosing the safer project is a rational response to the scorecard. If the scorecard stays the same, the choice will too.

The decision rights are unclear

94% of CMOs say CEO expectations of marketing have risen significantly in the past two years. One insurance CMO told BCG: "Our Board and CEO are challenging us to move faster on our AI agenda."

Pressure from above doesn't always come with authority. Roughly half of CMOs say marketing leads AI investment decisions within their function, while 72% of CEOs, in separate BCG research, see themselves as the main decision maker on AI across the enterprise. Agents that touch the CRM, paid media, email and the website need access across all of those systems. A CMO can usually approve a new tool alone. Connecting the data underneath often takes people they don't control. So tool adoption moves quickly while system change slows down.

The data isn't ready

Salesforce found that 98% of marketing teams using AI report at least one data-related barrier to personalisation, and that the average marketing organisation has seven data sources to integrate to support agentic marketing. An agent working from partial data makes confident decisions on partial information, faster than a person could.

Gartner's Ewan McIntyre described the risk this way: "CMOs invest in AI tools faster than they build the data foundations, processes, governance and talent required to scale them." The same report says 70% admit their internal processes aren't yet mature enough to scale the technology.

Money isn't the main constraint

Spending is already significant at larger companies. 43% of the CMOs BCG surveyed said their AI investment in marketing passed $15 million this year, up from 28% a year earlier. Gartner found that AI takes 15.3% of marketing budgets on average, in budgets that barely grew. Better-equipped organisations put 21.3% into AI.

BCG sorts CMOs into Leaders (32%), Followers (26%) and At-Risk (42%). The size of the At-Risk group matches the 42% who use generative AI only to assist with discrete tasks. To us, that suggests the dividing line is how the work is organised, more than which tools have been bought.

Why this can't wait forever

The outside world is shifting too. 91% of B2C CMOs and 76% of B2B CMOs told BCG that AI-moderated, no-click discovery is already reshaping their funnels. In Salesforce's survey, 88% are in the process of optimising for AI-driven search such as ChatGPT and Google's AI Overviews.

Competition could also come from new entrants. "If established brands don't build this first, new agentic-native attacker brands will do so," said BCG's Mark Abraham, a coauthor of the report. A beauty CMO in the research put it more personally: "I am racing to avoid a future where we lose ground to agent-native beauty startups."

What the India data adds

Global averages can hide regional differences. In BCG's report, 28% of Asia-Pacific CMOs put agentic commerce in their top three priorities, compared with 23% in North America and 13% in Europe.

Coverage of the survey's India findings adds that 57% of Indian respondents say AI is funded by the marketing function, against 47% globally, and that 53% of Indian CMOs project 5% to 9% incremental topline growth, against a global average of 43%. No India sample size is given in that coverage, so treat these as a signal of ambition. If marketing holds the AI budget in many Indian companies, the question of who owns the operating model is a practical one, and possibly easier to settle.

What to do about it

Finish one workflow before starting ten. Pick one journey stage or campaign type, such as re-engaging lapsed customers. Redesign it around agent decisions and define exactly where a person reviews the work. One complete workflow will teach you more than AI assistance sprinkled across twenty tasks.

Map the data before choosing the agent. List every system the workflow needs, who owns each one and how fresh the data is. If an agent will send email based on purchase history, and purchase data arrives in a weekly export, you've found your first project.

Write down who can decide what. Before anything is built, agree who approves access to customer data, who sets the limits on what an agent may do on its own and who answers when it gets something wrong. These conversations are slow, and they're slower after the build.

Change one number on the scorecard. If the team is judged only on cost per asset, it will keep producing cheaper assets. Add a measure the workflow should actually move, such as time from signal to response, or revenue from the journey you redesigned. Then take that case to the CEO or CFO.

Keep humans where mistakes are expensive. Autonomy doesn't have to be all or nothing. Let agents run low-risk steps, and keep people on pricing, regulated claims and anything that could embarrass the brand. Widen the limits as the results earn it.

A note for smaller teams. These surveys lean towards large companies, and Gartner's sample includes many firms above $1 billion in revenue. A 10-person team doesn't need an operating model redesign. But the same order applies at any size: clean, connected customer data first, one workflow next, and more autonomy only once it has proven itself.

Where this leaves marketing leaders

The honest reading of the 2026 data is that most marketing teams are still early, including many that describe themselves as transforming. There's nothing embarrassing about that. The risk lies in spending more each year on tools while the scorecard, the approvals and the data stay exactly as they were.

Nobody needs to chase the 8% figure for its own sake. Treat it as a reminder that buying AI is the easy part, and that the harder work is deciding how your team should run once you have it.

(For the practical side of day-to-day AI use, see using AI in your marketing without sounding like everyone else.)

Is your team stuck between AI tools and agent-led workflows? AgniCorp Media helps growth teams find the structural blockers and plan what to change first.