In conversations with marketing teams, I rarely have to convince anyone that something in their way of working is inefficient. They usually know exactly where the problems are.
They know which approvals take too long, which reports are rebuilt by hand, where teams are copying information between systems, and which processes depend on somebody remembering how things were done last time. Increasingly, they also have people experimenting with AI on their own and finding genuinely useful ways to work faster.
The difficult part comes afterward.
Someone has to turn those observations into an internal case for changing how the work gets done. That means asking for budget, involving IT or procurement, addressing governance concerns, and explaining why the organization should invest now rather than wait another six months.
This is where progress often stops.
The proposed investment has a visible price. The current process does not. Its cost is spread across salaries, delays, duplicated work, missed opportunities, and people spending their time compensating for systems that do not work particularly well together.
Waiting can therefore feel like the cheaper option even when the company is already paying for it.
The cost is already there
A simple calculation makes the direct cost easier to see.
Imagine ten people each spending six hours a week on recurring coordination, manual reporting, publishing preparation, approval chasing, and rework. At a loaded hourly cost of €45 across 52 weeks, that represents €140,400 in annual capacity:
10 × 6 × €45 × 52 = €140,400
That is an illustrative model, not a benchmark or a promise that €140,400 can be removed from payroll. Any serious calculation should use observed hours, actual working weeks, and the company’s own employment costs.
Its purpose is to put a number beside work that normally disappears into the calendar.
The direct cost is only part of it.
Slow processes affect how quickly a company can respond, test, publish, and learn. Disconnected workflows make it harder to maintain consistent decisions across markets and teams. Skilled people spend time on copy-and-paste work, chasing approvals, and reconstructing information that should already be available.
There is also an opportunity cost. Every hour spent rebuilding a report is an hour that cannot be spent understanding what the report means. Time spent locating the current asset cannot also be spent improving the campaign.
That becomes more important as AI increases the amount of work teams can potentially produce.
AI makes inefficient workflows more visible
Many of the teams I speak with are already using AI somewhere. Sometimes the company has introduced it formally. Sometimes individuals have simply found tools that make part of their job easier.
That is useful progress, but it can also reveal the next problem.
An AI assistant may draft copy in seconds while the approval still takes two weeks. It may summarize campaign performance while somebody still has to assemble the underlying data by hand. An agent may prepare regional variations while nobody has established which product information is approved for use in each market.
Bain’s 2026 Automation and AI Pathfinder Survey describes the accumulated handoffs, redundant approvals, and workarounds inside processes as workflow debt. Among the companies in its survey that measured AI cost savings, nearly 40% reported reductions below 10%, while data access and integration was the most frequently cited barrier to progress.
That should temper some of the larger AI promises, but it should not become an argument for waiting.
It is an argument for choosing the problem more carefully.
A company does not need to begin with a sweeping AI transformation. It can start with one workflow where the cost, delay, or quality problem is already visible and find out whether technology can improve it.
Why companies still hesitate
There are rational reasons to be cautious.
Changing a working process creates disruption. New systems have implementation costs. Data needs to be prepared. Permissions have to be considered. People need training, and somebody has to deal with the period when the old and new ways of working overlap.
AI adds another concern because employees quite reasonably wonder what automation means for their own jobs.
Those concerns deserve an answer. They should also be weighed against what happens if the organization keeps the current process for another year.
That is the conversation I often see companies struggle to have internally. The risk of changing something is visible and concentrated around a decision. The risk of leaving it alone arrives in small pieces every day.
Those pieces accumulate.
Competitors learn faster. Good employees spend more time on routine coordination. Individual teams build their own workarounds. AI tools appear independently across the organization because people are trying to solve immediate problems themselves.
Eventually, waiting produces its own transformation, just without much design behind it.
Small pilots make the decision easier
The companies that move forward do not necessarily begin with larger budgets or greater confidence. They make the next decision small enough to test.
A useful pilot might focus on one recurring campaign workflow and one strategic objective. Perhaps the goal is to reduce the time from an approved brief to publication, improve the quality of regional variations, lower reporting effort, or reduce the amount of work repeated during approval.
The current process is measured first. The pilot gets an owner, a budget, and a review date. The organization agrees in advance what evidence would justify expanding it and what would make it stop.
That is a much healthier use of experimentation than launching an AI initiative because everyone else appears to have one.
The pilot itself is not the problem. Experimentation is how an organization replaces assumptions with evidence.
Problems begin when pilots are disconnected from a business objective, or when successful experiments remain isolated because nobody has decided how the surrounding work should change.
This is becoming especially important with AI agents.
One agent may draft content. Another may analyze performance. Another may retrieve product information or prepare a regional adaptation. Each can be useful individually while still creating another fragmented collection of tools if the organization has not decided how they work together, what context they share, where human judgment remains necessary, and who owns the resulting process.
The larger opportunity therefore sits in the orchestration of people, processes, information, and agents, rather than in accumulating more agents.
Saving time is only useful if you know what to do with it
BCG’s 2026 AI at Work research found that 42% of regular frontline AI users reported saving at least eight hours a week. At the same time, 66% said they received limited or no guidance on how to use the time they saved.
That is an important part of the investment case.
If a reporting process falls from ten hours to four, six hours of capacity have become available. The value comes from what the organization does next: more analysis, more testing, better planning, stronger creative work, or another priority that previously had no room.
Without that decision, productivity gains can simply disappear back into meetings, messages, and other routine work.
AI investment therefore needs a connection to a strategic outcome. Higher quality, higher revenue, lower operating cost, faster learning, or another defined objective gives the recovered capacity somewhere useful to go.
Make the decision while you still have room to learn
The distinction I see is not between companies that understand AI and companies that do not. Most marketing leaders understand by now that something significant is changing.
The difference is what they do with that knowledge.
Some organizations choose a manageable problem, run a controlled experiment, measure what happens, and use the evidence to decide what comes next. They may scale the investment, redesign the process, or stop the experiment altogether. Any of those decisions can be rational.
Others continue waiting for certainty.
The difficulty is that certainty usually arrives late. By then, competitors have learned from several rounds of experimentation, employees have built their own habits and tools, and the company is making decisions under pressure rather than by choice.
There is no requirement to bet the marketing organization on AI tomorrow.
There is a strong case for identifying one expensive workflow, understanding what it already costs, and testing whether there is a better way to run it.
Doing nothing is still a decision. It simply has a budget that is harder to see.