Kimberly Bromann runs research at HubSpot with one other person, inside a company of roughly 9,000 employees. At Quirk’s New York 2026, she described what that used to look like. The team was too slow for the market. Its qualitative data was too shallow, its analysis painfully manual. A lot of lean research teams will recognize that description.
The fix wasn’t more headcount. It was an always-on program called the “Growth Gabby Listening Post,” powered by AI. It runs autonomous interviews with intelligent probing and turns hundreds of transcripts into queryable, decision-ready reports. Her verdict: easier than training an intern. “It really has shifted the cognitive overhead as well,” she added.
That shift, cognitive overhead traded for capacity, is showing up at a scale bigger than one team. Eight in ten insights professionals now say insights operations plays a significant role inside their organization. That’s per the 2026 GRIT Insights Practice Report, up from a standing start just one year ago. Fraud detection tools have moved the same way, from nice-to-have into embedded standard practice: 70 to 88% of teams across every segment now use them regularly. Work that used to sit on a person now sits on infrastructure. (We covered what that shift looks like on the data quality side in Data Quality Gets a Framework, and a Public Scoreboard)
CodexMR’s own numbers describe the same shift at study level: 80% faster survey programming, and a full quant study timeline compressed from 12.5 weeks to 10.5 weeks, spec to delivery. One DIY client saw 60% faster operational readiness and 90% less programming effort. Roughly two weeks handed back, per study.
The question, the one both events raised this year, is what a team does with that reclaimed research team capacity.
What Leaders Do With the Time They Protect
Shourav Sen, VP of Strategic Insights and Analytics at Colgate-Palmolive, protects time for this even inside a mandatory stage-gate process. He reserves research team capacity for portfolio evolution, innovation roadmaps, category growth, the work that doesn’t fit neatly into a project brief. “Our job is to influence with impact,” he said at Quirk’s New York. “The job does not end by delivering a presentation or a report. It is influencing the decision.”
Maria Amenta, Corporate VP of Research at New York Life Direct, reserves 15 to 20% of her team’s annual capacity for what executives aren’t thinking about yet. That’s not slack in the schedule. It’s a deliberate allocation, decided in advance, protected from the next urgent request.
Some of that protected time goes toward work AI still can’t do well. James Wycherley of the Insight Management Academy put it plainly at the event: “I don’t think AI is great at looking around corners yet, but it’s fabulous at doing lots of scenario testing if you put in different parameters.” Two Mondelez researchers, presenting with Mintel’s Emily Matterson, described cutting an 18-month innovation cycle down. The method: pairing early consumer signals with product intelligence, what brands do and how markets respond. Lisa Balban, Senior Director of Insights and Analytics at Mondelez, called it an iterative process they move through “very, very quickly.” The speed didn’t replace her team’s judgment. It gave them more runway to use it.
The Warning Inside The Same Data
Not every organization spends reclaimed time the same way, and the industry’s own research flags the risk. Rick Kelly, Chief Strategy Officer at Fuel Cycle, makes the case in his 2026 GRIT commentary. This is Jevons’ paradox arriving in market research. When something gets more efficient, total use tends to go up, not down. His conclusion: “There will be more research-like activity everywhere.” Faster doesn’t automatically mean freer. It can just as easily mean more studies crammed into the same calendar.
Kelly’s own prescription follows directly: “Anxiety does not come from AI itself. Anxiety comes from ambiguity. So the prescription is not to slow down. It is to get clearer.” The GRIT report is blunter about who’s exposed: “those most at risk are not the slow adopters of AI, they are the efficient executors with no clear control point.” Speed without a plan for the time it buys back isn’t a strategy. It’s a faster treadmill.
Wrap Up
That’s the argument for treating speed as an input, not the output. CodexMR’s platform removes the operational weight: survey programming, QA, data processing. That turns the time question into something a team decides on purpose, not a byproduct it hopes shows up. What Sen, Amenta, and Bromann described at the team level is the same shift GRIT measured industry-wide. Research team capacity is becoming something organizations design for, not something that happens to them.
The teams protecting 15 to 20% of their capacity for the unassigned question didn’t get there by accident. They decided it mattered before the time existed to spend on it.
