The marketing problem I keep returning to begins with perfectly serviceable AI work, which can be more dangerous than obvious failure. A file can arrive looking as if it has already been argued through, revised, approved, and placed into a campaign, despite being assembled before anyone has reached a view about the customer or the company. It has the outer texture of a decision, and that texture is often enough to make a team stop asking questions.
I have worked in several marketing departments, and I have never seen the work change at this speed. What worries me is less the disappearance of particular tasks than the easy substitution of activity for judgment. The pressure now comes from every direction at once: more assets, more channels, more reporting, more applications, more tool mandates, and a professional expectation that all of this can be carried by a smaller group of people because much of the surface work is now automated.
The reassuring story has the machines removing drudgery and returning us to strategy. That outcome is available, though it rarely emerges by default. The usual result is a denser workflow in which the time saved by a tool is immediately claimed by another request, another output, another dashboard, or another expectation that the team should now be able to cover a function it previously did not own. Most organizations use efficiency to raise the volume requirement instead of creating room for better work.
That is how content becomes a problem. Once a company can make twenty versions of something with little friction, it becomes harder to defend the time required to make one version genuinely good. The temptation is to call this testing. Sometimes it is. More often, it is a way of avoiding the uncomfortable earlier decision about what the message is, what the audience already believes, and what the company can credibly promise.
The marketers I trust most are becoming more willing to kill work. They are willing to say that a campaign has no angle, that a social calendar is merely filling space, that a brief contains the answer it was supposed to discover, or that a new channel has appeared because somebody wanted evidence of progress rather than because the business needed it. This can sound negative inside an organization that has turned production into a virtue. I think it is a form of care. Every weak asset sent into the world makes the strong ones easier to miss.
The argument should start with the workflow
I am tired of conversations about whether AI is good or bad for marketing. The interesting question is whether a particular use makes a real decision easier, or simply makes the department appear more current.
AI is genuinely helpful when it removes the preparation around a piece of work that still requires human attention. It can organize a messy research set, produce a competent first draft, identify a reporting anomaly, check a calculation, create a range of mechanical variations, and save the energy that would otherwise be spent getting to the point where someone can think. The problem begins when the generated output acquires authority just because it is fluent.
A weekly report is a good example. No sensible person should want a marketer to spend hours assembling the same recurring figures by hand. But a report that can summarize the figures is not the same thing as a report that understands why they changed. The gap between those two things contains the work. Somebody still has to decide whether a decline reflects demand, a platform quirk, an exhausted audience, a worse lead source, an attribution issue, or nothing worth reacting to yet.
That distinction should govern the budget. I would rather see a team pay for one tool that saves real preparation and gives an experienced person more time to investigate than pay for a sprawling stack that produces a great deal of presentable output and no sharper decisions. AI has its own payroll, hidden in the review, correction, prompting, data handling, brand risk, and handoffs that accumulate around the subscription fee. It is easy to ignore that payroll when the output looks immediate.
The companies that get this right will probably look a little less enthusiastic than their peers. Their use of AI will be specific, unglamorous, and tied to a workflow that somebody can explain without reaching for a slogan. They will use it to clear the desk, then do the work that cannot be cleared from the desk.
A metric is usually an argument from someone else’s system
I have become much less interested in the confidence of a channel report. The number may be accurate within the small world that produced it, yet still tell a misleading story about the business. That is especially true when the platform is optimized to make its own activity look meaningful, which is not a scandal so much as a fact of the arrangement.
Marketing departments often get into trouble by treating a clean number as a conclusion rather than an invitation to investigate. A cheap click can be a cheap click. A surge in engagement can be a surge in engagement. Neither tells us very much until it reaches the customer, the sales process, the retention curve, or some other point where the business actually has to live with it.
The more senior version of the job involves resisting the pressure to turn every result into a story of success. Sometimes the honest answer is that a number is incomplete, the test was inconclusive, or the channel has produced an event that deserves monitoring rather than a budget increase. I think people underestimate how much trust this builds inside a business. A marketer who is willing to preserve uncertainty when it is warranted becomes more useful than one who can always make the slide feel optimistic.
This also requires a degree of institutional nerve. Different teams will naturally defend the metric that protects their own work, and the dashboard can make that defense feel scientific. The most useful leaders create a shared view of the decision rather than a competition among self-flattering reports. They ask where the demand came from, what happened after the lead, and what result would justify moving real money.
The internet has made directness more valuable
The most sensible response to noisy platforms is a little more conservative than marketing culture usually allows. A business needs ways to be found, understood, and trusted that do not depend entirely on a feed, a trending format, or an opaque system of distribution. Paid media and social work still matter, though no single platform deserves the right to define what counts as attention.
I would take a reliable email program, accurate local information, useful product pages, strong customer reviews, responsive service, and repeat customers over a month of frantic social output almost every time. Those assets do not feel glamorous because they do not lend themselves to a screenshot. They do something better: they make the business easier to trust when a person is actually deciding whether to buy.
Organic social still has a place, though I think many teams have been encouraged to ask too much of it. It can make a company feel alive, give customers somewhere to ask questions, and build a relationship with people who already have a reason to care. It becomes less effective when it is treated as free mass distribution, particularly when the answer to every disappointing post is to produce more posts.
The same restraint applies to brand voice. I have no objection to humor, self-awareness, or a brand that sounds like someone who has been on the internet. The trouble starts when a company mistakes recognizably online language for a relationship with its audience. A joke can be technically competent and still feel commissioned. An executive can borrow the right vocabulary and still reveal that they have no contact with the product or the people who use it.
Good voice comes from proximity. It is easier to write when the company has listened to customers, knows what irritates them, understands its own limitations, and can say something that has not been copied from the prevailing tone of the feed. This sounds obvious, but a great deal of marketing work is devoted to avoiding that proximity because it is easier to generate language than to find out what people actually mean.
“More with less” has become a way to avoid priorities
I have some sympathy for broad marketing roles. They can make a person formidable. Someone who understands creative, distribution, conversion, measurement, and the commercial logic of a business is much more interesting than someone who only knows one platform well.
But the marketing job description has become a place where companies hide their refusal to prioritize. A role that includes paid media, content, social, creative briefing, analytics, email, partnerships, landing pages, and reporting can be a real ownership opportunity only when the person has authority, access to information, a credible budget, and permission to make tradeoffs. Without those conditions, it is a list of functions that management prefers not to staff.
AI is making this worse because it gives leaders a plausible explanation for why a smaller team should absorb more. The logic usually runs from a true premise, that a tool can speed up a task, to a false conclusion, that a person can therefore own the entire surrounding system. The missing ingredient is attention. A marketer can use automation to draft a brief and still need time to understand the offer, choose the audience, review the work, interpret the result, and explain the decision to the people who control the budget.
I would choose a team with permission to say no over one that is merely large or technologically advanced. The better team lets a new responsibility displace an old one, keeps measurement close enough to the business to discipline the conversation, and makes every urgent request compete with the rest instead of adding it permanently to the job.
Hiring has become another version of the same problem
The application process now contains the same abundance problem as marketing itself. A candidate can generate polished materials quickly, while a hiring manager receives far more polished materials than they can seriously read. Both sides can use automation to simulate care, and both sides then complain that the process feels impersonal.
Longer application rituals will not repair the hiring process. Candidates are right to be selective with the unpaid time they spend, especially after months inside opaque hiring systems. The application that matters ought to contain some evidence of a mind at work: a decision made under constraint, an explanation of what changed after the decision, a question that reveals a real understanding of the company, or a small piece of analysis that could not have been written for every employer at once.
Hiring managers have a parallel responsibility. If a company wants a careful application, it should make the role legible and create some reason to believe that a person will read what comes back. A short realistic problem can reveal much more about judgment than an elaborate take-home exercise, and it does so without pretending that a free consulting project is an interview.
The professional advantage in marketing is becoming harder to package because it is increasingly a matter of discernment. The best people will still know the platforms and tools, but their real value will appear in the work they refuse, the metric they question, the false certainty they puncture, and the customer relationship they protect. Cheap production has made that kind of judgment more visible, even as it has made it much harder to practice.