Read time: 6 mins
Ask whoever ran your last research program one question: who actually answered. Not the sample size. Not the demographic breakdown. Who, specifically, and how do you know they were paying attention, telling the truth, and hadn't already answered the same questions for someone else last month.
Most vendors can't answer that with a straight face. That's not an accusation about their honesty. It's a gap in what the industry has decided is worth checking.
Everyone Already Has AI. That's Not the Question Anymore.
A couple of years ago, the differentiator was whether a research platform used AI at all. That's over. Every serious platform runs on it now, the same way every serious agency runs on email. Asking about it is like asking a law firm if it has a printer.
The harder question replaced it fast: can you actually trust what came out the other end. Not the formatting. Not the narrative. Not the confident chart sitting on slide twelve of the deck. The signal underneath it, before anything got summarized, cleaned up, or made to look decisive.
That shift changes what agencies and consultancies are really being paid for. Speed and scale used to be the pitch, and AI made both of those nearly free, which means neither is worth a premium fee anymore. What's left to sell is judgment, and judgment isn't something you can install. It gets built slowly, in the least glamorous part of the process, the part that never makes it into a pitch deck: who was actually recruited, how tired they were, whether they were who they claimed to be, and what got quietly thrown out before the deck ever reached a client.
Jay Thordarson
VP, Research Services
The Logit Group
“The vendors worth partnering with are the ones who can answer one simple question without hesitation: who actually answered your study?”
The Part Everyone Keeps Forgetting Is Human
Here's what most of the AI-in-research conversation misses entirely. The technology is genuinely excellent at working with data. It has no way of knowing the data was thin to begin with. Feed it a genuinely engaged, verified sample and it finds real signal fast. Feed it a burned-out respondent rushing through a survey to collect an honorarium, or someone who's already answered four nearly identical studies this quarter, and it will still find a pattern. It just won't mean what the deck says it means.
A rushed respondent usually isn't lying outright. They're doing something quieter and harder to catch: giving the fastest plausible answer instead of the true one. Multiply that across a sample and you get results that are technically valid and substantively empty, confident numbers sitting on top of shallow attention. No model flags that on its own, because each individual answer looks fine. It only looks wrong once someone notices the person behind it stopped really thinking three questions earlier.
This hits specialist research hardest, because there are so few real specialists to go around. Physicians, senior IT buyers, category experts, the same relatively small group gets recruited across dozens of competing studies within the same few months. Some of them stop responding to invitations at all, which quietly reshapes who's left. Others become fast, practiced survey-takers who know exactly what a good answer sounds like without engaging much at all. A smaller number aren't who they claim to be, coached or self-taught just well enough to get past a screener built for expertise they don't actually have.
None of that throws an error. It produces a deck that looks exactly as credible as an honest one, right up until someone asks the wrong question and it doesn't hold.
What This Actually Costs
Smaller decisions can absorb this kind of noise without much damage. A modest campaign tweak built on a slightly hollow insight rarely sinks anything on its own. The math changes completely for the decisions agencies exist to inform in the first place, a launch, a repositioning, a brand's answer to a shifting category. At that level, the real risk isn't being wrong. It's being confidently wrong in the same direction as competitors who fished from the same overworked pool of specialists and never thought to check either.
AI has made this easier to hide, not harder. A polished summary of rushed, hollow, or partly misrepresented responses reads exactly like a polished summary of the real thing. The technology doesn't disclose its own inputs. Someone has to actually go check, by hand, before the deck ever gets built.
What We Actually Check For
This is the layer Logit Group has built its name on, not because it's exciting, but because it's the layer that decides whether a client can defend a finding when someone senior pushes back on it. Live verification instead of a checkbox. Tracking whether a specialist pool has already been tapped by two or three other studies that quarter. Noticing when answers are technically correct but strangely rehearsed, then following up in a real conversation instead of trusting a form. None of it is glamorous. All of it is the difference between a study that survives scrutiny and one that only looked like it would.
Agencies and consultancies holding their ground as AI commoditizes everything downstream of data collection are the ones treating human attention and honesty as seriously as they treat the analysis layered on top of it. Everyone else keeps shipping faster, cheaper reports that look identical to the trustworthy ones, right up until someone finally asks who was actually in the room.
Ask that question before the study starts, not after the deck's already been presented.
Go back to that first question now: who actually answered your last study. If you got a straight answer without hesitation, you're already working with someone who takes this seriously, and that's worth knowing. If you got a pause, a reassurance about sample size, or a redirect back to the findings themselves, that pause is the answer. It just isn't the one anyone wants to say out loud.
The vendors who can hold that question without flinching are the ones worth building a real partnership around. Everyone else is selling you a deck and hoping you never ask what's underneath it.
What Comes Next
So far in this series, I started with how AI is reshaping fieldwork in Faster, Smarter, More Human. Then I pulled back the curtain on respondent commitment. And now, in Mission Control, we have looked at how chaos is steadied when studies go sideways.
There is still one more layer that is often overlooked, and that is the conversations with our clients in the middle of it all. Many people see these conversations, but what they do not always see is how difficult they can be. The words matter as much as the fixes. Choosing your words carefully will add hours of sleep to your life.
My next article will take you inside those conversations. The ones where a client is waiting for answers, and the study has gone sideways. Where the words you choose can either calm the room or light it on fire. I'll share how we manage expectations without sugarcoating, how a well-timed bit of humor can take the air out of a tense call, and how trust is either strengthened or shattered in those moments.
About The Author
Jay Thordarson
VP, Research Services
Jay is an accomplished market research professional with extensive experience in global qualitative and face-to-face research.
