AI's Next Opportunity: Deciding Where the Money Goes

A while ago, a marketing friend vented to me: proposals get written faster and faster, assets keep piling up — but what his boss asks for never changes. One line:

Was the money spent in the right place?

That stopped me for a second. Notice something? These past two years, every AI conversation has been about "faster": ten copy variants in minutes, posters at the click of a button, video scripts in bulk. On the production side, AI has already become a blood-red ocean.

But the question that's actually worth money hides at the other end: not "how fast can we make things" but "where should the next dollar go."

Money Is Getting Tighter

First, a few numbers.

A survey of 300+ US marketing leaders found that over the past 12 months, marketing spend grew just 1.7% — the slowest since 2021. Harsher still: marketing budgets have fallen to 9% of company revenue.

What does that mean? For every $100 the company takes in, it's only willing to spend $9 on marketing.

The pie isn't just shrinking — the hand holding the knife is trembling. In times like these, if you're still using AI as a "content printing press," you're basically begging for spare change while holding a golden bowl.

Companies have run their own numbers too: they estimate that by 2029, more than half of marketing activities will be carried out by AI. Note — "carried out," not merely "assisted."

So what exactly should AI carry out?

The spending decisions.

From "Making Assets" to "Choosing Channels"

What does "deciding where the money goes" mean? Let me tell you what's happening right now.

What did ad buying used to look like? Plan set at the start of the month, channel budgets sliced up, run for a month, review at month's end, adjust next month. Problem spotted? Wait — wait for the next cycle.

And now? Ad delivery systems adjust on the fly: bids running high get auto-suppressed, an underperforming creative gets auto-swapped, when channel A's ROI picks up the budget auto-shifts over. No one's sign-off required.

That's the heart of the change. Marketing used to be a contest of budgets; next it's a contest of intelligence.

Whoever's algorithm understands the user better — what they've watched, how the last campaign performed, what the competition is doing, what context this person is in right now — gets to place money on the single most precise target.

Netflix is a great example. They use predictive models to flag subscribers at risk of churning ahead of time and target retention efforts; they even personalize thumbnails with AI based on each person's viewing habits. Same show — the cover you see and the cover I see may not be the same image at all.

And this isn't Netflix's exclusive club. Retailers are running store-level personalized promotions, consumer goods companies are honing demand forecasts with it, B2B teams are using it to rank account priorities. At the door of every industry stands the same opportunity.

When Agents Step In, More Than Efficiency Changes

One step further, and things get even more interesting.

Programmatic advertising — machines buying and selling ad inventory automatically, in real time — passed $271 billion in US transactions last year, more than nine-tenths of digital display advertising. Machines buying ads is old news.

The news is that "agents" have arrived.

What's an agent? Not a tool that helps you work — a system that can make its own decisions and act on them. It designs its own A/B tests and changes its own variables when results come in; it throws hundreds or thousands of "headline × image × call-to-action" combinations into simultaneous testing and shifts budget to whichever wins; it even runs a round of optimization decisions every few hours — something humans used to do once a week.

One agency worked exactly this way. The results: customer acquisition cost down 60%, sales cycle shortened by 40%.

Impressive? I'd say so.

IAB has a forecast: in 2026, four in ten video ads will be created from scratch by generative AI, or run through it once.

But every coin has its flip side.

Before Full Autonomy, Put Humans on Watch

An agent making a decision for you every few hours — sounds thrilling. But while you're asleep at 2 a.m., what exactly did it decide?

That's what I really want to flag.

AI will make marketing faster and cheaper — no argument there. But however strong the tool, buying it and plugging it in doesn't produce results by itself. The team has to understand what it's doing, someone at the company has to watch it for drift, ethical red lines have to be drawn in advance — don't let automation burn through years of accumulated brand trust overnight.

At the end of the day, a successful AI transformation moves people, process, technology, governance, and measurement together — it isn't "buy a system and call it done."

The marketing leaders of the future won't compete on hand speed, but on who can orchestrate AI, data, and human creativity together.

Machines handle calculating fast; humans handle thinking right.

Back to my friend. He later deleted the slide in his deck about "how many assets AI can make us" and replaced it with a new set of questions: which decisions can be handed to machines, which must stay with humans, and how do we verify that the money saved on efficiency actually turned into growth.

I think he deleted the right slide.

When the pie shrinks, what's valuable was never the speed of the knife — it's the hand that knows where to cut.