A working paper by Nihal Kingre

Abstract

Every marketing course teaches Maslow, AIDA, and the four Ps. Almost none teach what happens in the three seconds before someone actually opens their wallet. This paper looks at two things happening in my own practice right now: first, how AI research tools like Notebook LM are changing the mechanics of how a strategist builds a point of view, and second, the psychological wiring that decides whether a brand message lands or dies, the part that never makes it into a syllabus because it doesn't fit neatly into a framework. The argument is simple. The tools got faster. The human reading of a customer's mind did not get easier. If anything, speed without depth is now the bigger risk in the room.

1. The Research Stack Has Changed, the Job Hasn't

A year ago, building a brand research document meant opening fifteen browser tabs, a Google Doc, and hoping I remembered where I read that one stat about Indian Gen Z spending habits. Today the workflow looks different. I drop a stack of PDFs, competitor decks, and old client briefs into Notebook LM, ask it to find the contradictions between sources, and get a synthesis in the time it used to take me to format a slide.

That's a real shift. But here's what I keep noticing: the tool is excellent at finding what's already been said. It is not good at noticing what's been left unsaid. Notebook LM will tell me five sources agree that "Gen Z values authenticity." It will not tell me that "authenticity" has become a word brands use to avoid saying anything specific, or that the actual driver underneath isn't authenticity at all, it's the fear of being caught liking something uncool.

AI-assisted ad ops has the same pattern. Tools that auto-generate ad variations, predict CTR, and optimize spend in real time have compressed weeks of manual A/B testing into hours. I've used this to test headline structures across client accounts faster than I ever could manually. But the machine optimizes for the click. It has no opinion on whether the click was earned through genuine resonance or through a cheap dopamine trick that burns trust three purchases later. That distinction, between a click that builds a customer and a click that just rents attention for a day, is not something any current ad platform measures. It's something a strategist has to hold in their head.

So the research stack changed. The job, reading what a number actually means about a human being, did not get automated. It got more important, because now there's more noise to read through.

1.1 What AI Research Tools Are Actually Good At

Three things, in my experience:

• Cross-referencing volume. Finding the same insight repeated across twelve sources in minutes instead of days.
• Surfacing contradictions. When one report says D2C CAC is rising and another says it's stabilizing, the AI will flag both instead of quietly picking one.
• Compressing first drafts. Turning raw research into a structured outline that a human then has to argue with, not accept.

1.2 What They're Not Good At

• Reading silence. What a brand's customers don't complain about publicly because the complaint feels embarrassing to say out loud.
• Sensing tone shifts. The difference between a five-star review that's genuinely warm and one that's polite because the customer feels guilty leaving anything lower.
• Knowing when a stat is technically true and practically useless. "73% of consumers want sustainable packaging" tells you almost nothing about what they'll actually pay for it.