Only 6% of marketers see AI impact.

Bain surveyed 1,397 senior marketing and finance executives and found that 95% of marketing organizations have adopted AI tools.

Only 6% said AI was impacting performance.

I wanted to dig deeper into this because marketers clearly aren't struggling to adopt AI.

Other studies show a bigger problem.

Supermetrics surveyed 435 marketers. 80% said they were under pressure to adopt AI, but only 6% had fully embedded it into their workflows.

McKinsey found that 90% of CMOs are experimenting with AI, while fewer than 10% have scaled it or captured value across marketing workflows. Only 28% of marketers said their companies were changing teams and workflows around AI.

So I think the problem is how we're defining AI adoption.

A marketer using ChatGPT every day is one form of adoption.

A marketing workflow being redesigned because AI can handle multiple steps is another.

Many companies have done the first. Not many have done the second.

Most AI adoption is still happening at the task level.

A marketer needs to analyze something, so they use AI. They need research, so they use AI. They need to work through a dataset, prepare a campaign, or summarize customer feedback; they use AI.

That can save time for the person doing the work. But marketing performance is rarely determined by one task.

Take campaign optimization.

A change in performance has to be detected. The data needs to be collected. Someone has to identify what caused it, decide what to change, get any required approval, make the change, and then track the performance.

If AI cuts the analysis from an hour to ten minutes, that stage improved considerably.

But if the data still has to be collated manually, the recommendation waits for approval, and someone else has to make the campaign changes later, the complete workflow hasn't improved by the same amount.

This is where individual productivity and company performance start to seperate/.

McKinsey found nearly 60% of marketers now use AI several times a week, but fewer than 10% are capturing value across end-to-end workflows. 

It describes most current implementations as ‘AI being layered onto existing processes’ rather than the processes themselves being redesigned.

That’s the problem here… AI can make several people faster without making the overall process much faster.

A workflow has dependencies between each step.

AI might produce an analysis immediately, but it can't act on data it can't access.

It might recommend a campaign change, but it can't move the process forward if it doesn't have permission to make or stage that change.

It can complete one person's work, but the output still needs to reach the next part of the process in the correct format and with context.

The limitations therefore aren't always in the AI. They can be in everything surrounding it.

Supermetrics found only 33% of marketers said they could effectively activate their data. 52% said an external data team defines their data strategy, while 36% identified connecting marketing data as a major area for improvement.

The WFA found a similar pattern: AI is already used by 80% for content creation and 72% for research and insights, but only 46% use it for workflow automation or autonomous agents.

Using AI within a single task usually requires little change.

Building it into a workflow means connecting the data, systems, rules, and approvals involved in completing that process.

That's a more challenging implementation.

Go back to the campaign example.

The task-level version starts when a marketer asks AI for help.

They open the tools, collect the data, give AI the context, and ask it to investigate the performance change.

AI does a good analysis. Then the marketer takes over again.

Workflow implementation starts when the performance change happens.

The system detects it, pulls the relevant campaign and conversion data, checks the likely causes, and prepares the recommended action.

Changes within agreed limits could be staged automatically. Anything affecting budget or carrying more risk could go to the marketer for approval.

Once approved, the change can be made and its performance monitored against the original problem.

The marketer hasn't been removed from the process. Their involvement has moved to the parts where a decision or professional expertise is required.

That's a different use of AI, and it gives you something to measure.

Because that's the easiest implementation. 

You can give a marketing team ChatGPT tomorrow.

You don't need to redesign a process, connect five systems, or decide which actions AI should be allowed to take.

The WFA's data shows how adoption has developed. 80% of major brands now use AI for content creation, 74% for ideation and production, and 72% for research and insights. Workflow automation is at 46%.

The early use cases fall within the work marketers already do.

Workflow implementation asks different questions:

Those aren't questions about which AI tool to buy. They're about how the marketing operation works.

If a team already has high AI usage, getting more people to use AI isn't going to help.

I'd pick one important marketing workflow and look at it from the initial trigger to the final outcome.

For example:

Performance declines > cause identified > action decided > change deployed > result measured.

Then look at every step.

Then track the workflow before changing it.

If it currently takes three days to go from detecting a campaign problem to deploying a fix, that's the benchmark.

If the AI version reduces that to four hours, you have a clear operational result.

Then you can look at whether that speed also changed acquisition cost, conversion rate, or whatever business metric the workflow was designed to improve.

That's a better way to look at AI adoption than how many employees used ChatGPT this month.

The next stage is finding the marketing workflows where AI can change how work gets done, then measuring whether the complete process performs better afterwards.