AI is getting faster, cheaper, and more capable every quarter. But the problems it's creating inside research and insights teams? Those aren't really about the tools.
Think about what happens when polished-looking output moves through an organization without anyone truly owning it. Decks nobody read closely. Reports where the AI filled in conclusions a human never stress-tested. Emails that sound authoritative but wouldn't survive a single follow-up question.
The speed is real. So is the damage.
Andrew Reid, CEO and Founder of Rival Technologies, tackled this head-on in a recent piece for Entrepreneur: "Most Teams Are Using AI the Wrong Way." His argument is a useful gut check for anyone in insights who's been riding the AI wave without pausing to ask what they're actually putting their name on.
Here's the thing that gets buried under all the hype. Generative AI tools — Claude, ChatGPT, pick your poison — are probabilistic. They're designed to produce plausible outputs, not accurate ones. They predict the next word based on patterns in training data. That's what makes them fluent, fast, and honestly kind of impressive.
That's also what makes them unreliable without a human in the loop.
A Harvard Business Review piece on AI in market research makes a related point: the promise of AI in research only holds when human judgment is part of the process, not bolted on as an afterthought. At Rival, Andrew notes, many AI integrations are purpose-built for tightly scoped tasks with defined guardrails, which keeps the randomness in check. But the general-purpose tools most of us use daily have no such guardrails. That's where discipline matters most.
Put this on a sticky note. Tattoo it on your forearm. Whatever works for you.
Andrew's central argument comes down to one standard:
"Use AI however you want, but every word leaving your inbox, Slack or presentation should still feel authentically yours. If you cannot defend a sentence out loud, you should not send it."
The example he gives is painfully familiar. A presentation circulates through a company, triggering hours of downstream analysis and discussion, until it emerges that the person who sent it hadn't actually read the full deck. The AI generated it. Someone forwarded it. An organization acted on it.
Nobody owned it.
That's the accountability gap, and it compounds fast.
For insight professionals, this goes beyond a productivity problem. Research from Rival Technologies on AI slop found that when colleagues encounter obviously AI-generated work, they don't just discount the output. They discount the person behind it, and then start questioning the underlying data. Trust erodes fast and rebuilds slowly. Ask anyone who's sat through a research debrief where the client clearly stopped believing the numbers.
Here's a counterintuitive truth insight teams need to hear: AI doesn't automatically save time. When output skips proper review, the cleanup work that follows often exceeds whatever time the tool saved in the first place.
Andrew puts it plainly: "AI is not a solve-my-problem button. In many cases, rushing to use AI without scrutiny creates more cleanup work later than doing the thinking properly the first time."
He also flags a failure mode every heavy AI user has experienced. A session starts strong, the output feels sharp and useful, and then quality quietly drifts. LLMs are built to keep generating. Stopping to question their own reasoning? Not part of the design. That part is on you.
His fix: build in deliberate pauses. Summarize what's been produced. Reset if needed. Re-examine the assumptions before pushing forward.
This maps directly to what insight leaders have been telling us. Jason Jacobson, Senior Director of Consumer Insights at Sekisui House, framed it well: "The number one skillset of an insight professional is their ability to think. What AI has enabled us to do is use that thinking time more effectively."
The unlock isn't offloading your thinking. Protecting more time for it is.
This is the part of Andrew's piece that should make every insights leader a little uncomfortable.
He describes a habit of interrupting polished presentations with pointed questions. Not because they look weak, but because they look too clean.
Confident narratives and sharp-looking charts used to signal that someone had done the work. Now they signal that someone had access to a decent AI tool.
What he's actually testing when he pushes back on a deck:
That last question is particularly important. As we've written about before, the researcher role isn't disappearing, it's evolving. AI handles synthesis at scale. When used right, it can help to improve your storytelling techniques. But what carries the most weight is interpretation: connecting findings to business context, knowing what matters and why, catching what a pattern-matching model simply can't see. Keeping humans genuinely in the loop is what makes insights credible when they land in a boardroom.
To be clear: Andrew isn't sounding an alarm against AI. In fact, he has said before that researchers and leaders need to be proactive about AI adoption. The leverage is real, the speed gains are meaningful, and the upside for teams who use it well is significant. The goal is a culture where experimentation is encouraged and accountability is clear.
"Teams should feel comfortable exploring these tools without fear of getting everything wrong. At the same time, expectations need to be clear: if you send it, you own it."
Our Market Research Trends 2026 report found that 46% of researchers expect their AI budgets to grow this year. The investment is accelerating. The teams that get the most out of it won't be the ones generating the most output. They'll be the ones who pair AI speed with genuine human accountability.
That combination is rarer than it should be. Which also makes it a competitive advantage.
The accountability gap is the real AI risk for insight professionals right now. Not displacement, not data quality in the abstract. The specific, quiet erosion of trust that happens when AI-generated work travels through an organization with nobody truly owning it.
Andrew's standard is simple and it works: if you can't defend a sentence out loud, don't send it. That bar doesn't slow down good AI use. It's the condition that makes it sustainable.
Use AI freely. Own everything you send. That's it.
Read the full piece on Entrepreneur: Most Teams Are Using AI the Wrong Way — Here's How Smart Leaders Avoid Costly Mistakes
Got questions about insight communities and mobile research?Chat with one of our experts
No Comments Yet
Let us know what you think