Can AI Fix Modern Advertising?

In 2015/16, I recall a conversation I had with my mentor about data. Our concern was that data was becoming commoditized and the need for clean, targeted universes was getting diluted. Direct mail and addressable media had to be right, because every mail piece cost postage and production. That made custom selects and real models a must have. Digital makes impressions nearly free, so the precision needed in traditional media became less important. Segments got named for discoverability, not accuracy. They compete for a media planner’s attention, not for their relevance.

AI is speeding up that trend. It scales with whatever you feed it and produces impressive looking output. A model trained on segments built to be discoverable is only as good as the data it was fed. The results look better than ever and the underlying data is worse. The concern we saw back in 2015 came to fruition.

Programmatic is based on mass audience marketing. A large universe is needed to drive a higher match rate, so that more ads can be delivered. A highly targeted universe is smaller by nature, but more responsive. The smaller universe will drive a lower match rate, which affects scale. Programmatic runs on scale. Precision targeting doesn’t.

Digital trades responsive lists for reach. One of the biggest questions digital faces is reporting and attribution. In my five years in digital marketing, clients wanted to see strong lower funnel results from targeting that was built for awareness. The targeting never matched the KPI.

This is the question that will define the next era of marketing. In the mid-2000s, I worked with large financial mailers. Their internal policies started to preclude the use of age, income, marital status and general life-stage events, all of which were triggers for financial decisions. The fear was discriminatory lending practices. The result was that gross response dropped and the ROI became prohibitive for future campaigns. Their rules became their own downfall.

In 2026, states like NJ and CA are passing laws limiting the use of PII, and large corporations are adopting restrictive policies of their own. The industry-wide conversation is whether to even carry names in those states, or to use channels that require personalization at all. That is a factor in direct mail and other print channels seeing a decrease, while anonymous channels like banners, OTT and audio see the dollars move to them.

There are 40 years of evidence that personalization drives response. We’re walking away from it for reasons that have nothing to do with its efficiency.

Marketing needs to unironically look back to what worked best in the past. A client’s CRM list is their best marketing tool. It allows for cross-promotion and is the anchor for any lookalike modeling.

Lead-gen forms and other self-reported sources are strong intent signals. If someone is actively looking for auto insurance, home insurance or a product, that is the prospect you want to contact in the next 14 to 30 days. Overlaying that data with demographics, permissible-use Zip+4 credit attributes and transactional data creates rich pools of prospects, and those can be used cross-vertical as well.

It also means a return to data modeling. As more marketing becomes mass appeal, building regression and two-step models is what separates a data shop from its competitors.

The question asks about intelligent targeting and I hear ethical marketing. The line has to be drawn when you know the decision is murky at best, but profitable. You make the ethical decision because it is the right one. Sometimes that costs you. That’s where the line lives, and where you have to choose what side you want to be on.

AI can be an amazing tool, and data will always be the anchor that marketing is built on. The one thing that never changes is the human making those decisions. A quote I have used before: “Marketing works best when intelligence guides it, innovation supports it, and ethics keep it defensible.”