Know Your Dealer: Where is Your Speed Coming From?

Do you know where your speed is coming from?

Seemingly understanding your customer is faster than ever. Marketers now have access to tools that promise insight in hours instead of months. Thousands of conversations can be conducted simultaneously, themes apparently identified almost instantly, and “outputs” delivered in a format that looks really snazzy and ready to drop into a deck. What once required logistics, fielding, and careful analysis can now happen within a single workflow. That speed is sexy, and it can be addictive.

But where is that speed coming from? Do you know how that speed is possible? Can you tell the good speed from the bad? Do the dashboards show you where your reputation is at risk? There is very good speed, but we need to be able to identify it.

That’s because speed in consumer understanding has never been neutral. It comes from replacing or modifying something in the process, and those changes determine whether the output strengthens decision-making or erodes it. A lot of the current speed trend is coming from changes that aren’t obvious to the untrained eye looking at the “insights” it produces. Some elements are adjusted—and some parts are removed entirely.

In some ways this isn’t new. When I taught quantitative analysis at NYU in my Global Market Research course, I would say to my students, “SPSS is not going to pause and say, ‘Hold up, are you sure that's what you want to do? Something looks funny.’” Of course not. It’s going to spit out a number—an “answer.” SPSS gives you the output. It’s up to you to know you set it up right.

This is all still true in 2026, but the stakes are higher. You have to trust that the “experts” (who, in some cases, are 22-year-olds devoid of experience in consumer insight, research or even deep conversations with other humans) set it up right. And these experts don’t think about cuts to the process, but efficiency (for efficiency’s sake).

The first efficiency is AI moderation. Instead of asking the same questions in the same way, many AI systems adapt in real time, shifting their line of questioning based on each respondent’s answers. Dynamic responses custom-tailored to each participant? What could be better? At the level of a single conversation, that can feel like a clear upgrade. It’s more fluid, more natural. That’s useful! But it's also where things can get slippery.

When each person is effectively answering a different set of questions, there’s no longer a shared frame of reference. And without that, comparison isn’t possible. What’s more, aggregation gets shaky, and the neat pattern you found won’t hold up when you roll out your strategy to millions.

Efficiency #2 sees us replacing the participants with AI. Some tools generate responses based on existing digital behavior rather than asking people directly, meaningfully. It creates the perception of scale—fast. It feels expansive. It feels right in its confidence and speed. You can ask the “participants” anything at any time because they’re always “on.”

But the “existing digital behavior” it uses is shallow internet filler. That means the response you get isn’t a nuanced reaction to a specific topic in a specific moment. It’s a reconstruction—a duct-taped together version of a lot of slop people posted with little thought.. And it’s not the same thing as understanding.

Marketing decisions aren’t made in the abstract. They’re made in response to concrete things—messages, products, choices. You need to know the psychology of how people will react to that thing in that moment, based on their needs. Inference can point you to the past. It can’t predict the future.

A third "efficiency" is AI in analysis. A very useful tool that doesn’t check its own work. While it organizes language well, surfaces themes, and even packages them in a way that looks campaign-ready, it doesn't verify whether any of it actually holds up. That step—pressure-testing patterns, testing for inconsistencies, confirming what’s real—that’s where a lot of the rigor has traditionally lived. It’s also one of the easiest things to compress when speed becomes the only goal.

With each of these efficiencies, the output is ostensibly clean and cohesive—but the cost of that shiny output is imprecision somewhere else. Speed alone isn’t proof, and coherence isn’t accuracy.

The role humans have played in research was never just about asking questions or summarizing answers. It was about holding the whole structure together: making sure inputs were consistent, responses were comparable, and conclusions were actually grounded in something real. That structure still matters. AI doesn’t remove the need for that structure, nor can it solve for it.

When the structure doesn’t hold, the output will not tell you. Just as I taught my graduate students learning SPSS: The output doesn’t announce that it’s worthless. In fact, the output looks great!

And that’s where the massive risk is, and why you need to know your dealer. How are they using AI? Is it just for speed, or have they built a new structure that includes speed as well as depth and assurance. This is where the future lies. We can build a new structure and a new process.

If you’re a marketer, you’re not thinking about your customers’ needs while creating campaigns for no reason. You’re doing it to inform real decisions: decisions about what to design, what to say, where to spend huge sums of money. Any insight you use to make these decisions needs to hold up when the stakes are real. When you are deploying your reputational capital on a massive strategy that you need to have a double sided ROI.

AI use is a given. It can make everything better. It can remove friction without weakening thinking. And depth at speed is possible, but only if we use AI as an additive element—not a replacement—and build a new structure around its magic.