By John Bird, EVP, Infotools / Image courtesy of Infotools
AI has become the answer to almost every question. Most of the ideas are sensible enough. Faster reporting, smarter search, automated summaries and easier access to historical studies all have obvious value. What's changed is that almost every platform now tells a similar story, which makes it much harder for buyers to work out where the real differences lie.
Working at a company that's spent more than 35 years building software for market researchers changes what you look for during those conversations. AI is already helping research teams save time in plenty of practical ways, but it has also made selecting a platform more complicated than it used to be.
Long-established vendors are introducing AI-driven capabilities while new companies appear almost every month promising conversational analysis, automated reporting or instant answers from years of survey data. The technology is moving quickly, but software demonstrations have become much easier to produce than confidence in how those tools will perform once they become part of an organization's research operation.
Product demonstrations only tell part of the story
I've watched plenty of vendor presentations over the years, and they usually achieve exactly what they're designed to do. The presenter asks a carefully selected question, the platform produces an answer in seconds, and everyone starts thinking about the hours that could be saved.
One comment I hear repeatedly from prospects and clients comes much later. The demonstration looked great, but once the platform was implemented, everything became more complicated. Extra work appeared that nobody expected. Costs increased. Features that looked straightforward during the sales process turned out to require custom development or manual workarounds. More than one client has joked that they have a bit of "software PTSD" after being burned by previous implementations.
A research program that's been running for eight or ten years rarely looks like the clean dataset used in a demonstration. New markets have been added, questionnaires have changed, suppliers have come and gone, weighting has been revised and exceptions have accumulated because the business needed them. Those details rarely appear during a product demonstration, but they become part of everyday work once the platform goes live.
I spend less time looking at how quickly a platform answers a prepared question and more time understanding how it behaves when those complications appear. Discussions move to weighting, statistical testing, trend continuity, respondent-level data, and whether researchers can inspect the evidence behind an AI-generated answer before it reaches the rest of the business. Those conversations usually reveal how much market research experience sits behind the software.
Market research isn't just another dataset
Many of the companies entering this space have deep AI expertise but relatively little experience working with market research data. We saw something similar when business intelligence platforms such as Tableau and Power BI started finding their way into research teams. They solved many business reporting problems, but supporting weighting, multi-select questions, statistical testing and the other requirements of market research often took considerably more work than people expected.
The same pattern is starting to emerge with AI. Market research data may look structured on the surface, but it carries years of decisions about questionnaire design, weighting, coding, harmonization and reporting. Preparing that information for reporting and analysis requires people who understand how studies fit together, which variables should be standardized and where comparisons are valid. Those decisions still rely on market research expertise, and they're becoming an important part of how organizations evaluate software providers.
Choosing a platform now requires different questions
The latest GRIT report found that brand-side research teams reported more staff reductions than increases for the first time since Greenbook began tracking the measure, while almost nine in ten expect AI and automation to play a bigger role over the next year. That’s why software decisions are receiving more scrutiny. Teams need technology that genuinely reduces manual work without introducing new uncertainty into the research process.
When evaluating platforms, researchers need to understand how they handle respondent-level data, what happens when multiple suppliers contribute to the same program, whether historical studies can be compared without losing important context, and how straightforward it is to trace an AI-generated summary back to the underlying data. Those discussions usually reveal much more than another feature list.
Ipsos recently found that many people feel both excited and nervous about AI. I hear much the same thing from research leaders. They're already seeing where AI can remove repetitive work, but they also want to understand where answers came from, whether they can verify them, and whether the people building the technology understand the realities of market research. Those questions have become part of almost every software evaluation, and they're often the ones that separate a polished demonstration from a successful implementation.
About the Author: John Bird currently serves as an Executive Vice President for Infotools (www.infotools.com). His experience spans B2B and B2C work, and he has conducted research programs in over 70 countries. He is focused on fueling curiosity and moving clients from three-ring binders and “death by PowerPoint” to Infotools Harmoni, a SaaS data design, investigation, and reporting platform.