Will AI Replace Researchers? Two Senior Researchers Answer

Dataspace has been rebuilt around AI agents. Building an AI that knows research well was never a purely technical challenge — it rested on the deep domain expertise Opensurvey's researchers have accumulated since 2011. Those experts were part of the process of making Dataspace AI give better answers, and since launch they've been using it heavily in their own work. The question "won't AI replace researchers?" is still going around the market, but Opensurvey's research experts, who already use AI actively in their work, see things differently. We sat down with Sanchez (Lim Sung-jun), a researcher with 19 years in the field, and June (Kim Hee-yeon), with 13 years, to hear what they had to say.

Part 1. How Researchers Work in the Age of AI

Now that Dataspace AI has launched and we're getting the word out, we've asked the two of you to join us. Could you each introduce yourselves briefly?

Sanchez: Hello. I'm Sanchez, a researcher with 19 years in the field. My job is to turn a client's business problem into research questions and to help them reach insights they can actually act on. When a bottleneck shows up somewhere in the research process within our business group, I help clear it, and I support the team in doing their work better.

June: Hello, I'm June. I've been doing research work for about 13 years. I've run research projects for a lot of client companies over that time, and these days I mostly work with the companies that have long-standing relationships with Opensurvey.

We're curious what a senior researcher's job actually looks like. Could you walk us through a typical day?

June: After I get in, I check email first, since that's where client communication happens, and then I work through whatever needs doing by priority — designing surveys myself, analyzing data, and so on. Beyond that, there are a lot of meetings, internal and external. Client meetings fall into two categories: sessions where we discuss a research project, and sessions where we help Dataspace subscription clients get more out of the platform. Internally, it's mostly sales and product meetings, where I take the client needs I've picked up and think through how Dataspace can address them, then discuss possible solutions.

Opensurvey has a culture of using AI heavily regardless of role. As researchers, how do you two actually use it in your work?

June: I use it across the board. To be more specific about the research process — I design surveys myself from the start, but I get AI's help with the detailed work like question wording and answer options. When I'm cleaning collected data, I lean on Dataspace AI's features. For analysis, I use AI's responses to get a read on the data overall, and I use it heavily for the labor-intensive parts like cross-tabulations and pulling reports together.

Sanchez: June picked up the technology relatively quickly, but honestly I didn't use it much at first. There were a lot of errors in what AI produced. Then model performance improved significantly, and now I use it to the point where it would be hard to work without it. For research projects, I use it in much the same way June does. On top of that, my role involves advising account managers (AMs) a lot, and it helps there too. When I'm reviewing whether a client's concerns have been properly reflected in a survey, I used to have to go through the questionnaire line by line — a complicated survey could take half a day or more. Now I have AI do a first pass to classify the question structure so I can see the shape of it, and then I start reviewing. It's much faster.

So AI is making research work considerably more efficient. Even so, is there anything you never hand off to AI?

June: At the stage of deciding what kind of study to propose to a client, I don't use AI at all. You're not proposing a project direction based on the research topic alone — you're weighing the client's business situation, how they make decisions internally, what role each team plays inside the company, budget, all of it. For example, the RFP might point to qualitative research, but the way the client's organization is set up may make qualitative work difficult to run. Explaining all that context to AI piece by piece is slower and less accurate than just judging it myself, from the background I already have. I think this stage is something only a person can do.

Given how much AI can do outside of those areas, you must also have thought about what it means for the researcher role.

June: I actually worked through my thinking about the job of researcher before I started using AI, and I reached a conclusion of sorts. It started around the time Feedback (feedback.io), the predecessor of Dataspace, was released. We'd built a product that let companies run research themselves, so I found myself wondering what a researcher outside the company like me was supposed to do going forward.

But after the product launched and I worked with it, I realized two things, and that settled it. The first is that even with a product in place, there are clearly areas where you need an expert. Companies that have handled data themselves end up using data more, and from there they start attempting more complex research. So a need for researchers appears again. The second is that for a product to mature and get more complete, it needs continuous expert input. AI has to keep learning from good data and knowledge too, right? My conclusion was that even as the barrier to research comes down, there's still a role for experts. AI felt like the same story to me, so I took it in stride without much worry.

Sanchez: It was a bit different for me. When I was using Open Analytics, which supported the analysis process (it's since been absorbed into Dataspace's analysis features), or Feedback, my reaction was just "this is good, this is innovative." But watching up close as we rebuilt Dataspace around AI agents and embedded our research know-how into it, I started thinking much more seriously about my role as a senior researcher.

Honestly, I haven't finished working through that yet. What I can say is that as working with AI became routine, I got clear about what to hand to AI and what I, as a senior researcher, should do — and I've come to feel that however well AI does the work, the judgment still has to come from a person. The eye for recognizing "is this a good output?" matters more than ever, and that comes from experience. I think that's one of the places June was talking about where a researcher is needed.

Part 2. What It's Like to Use an AI Your Own Know-How Went Into

Dataspace AI has Opensurvey's accumulated tacit research knowledge and know-how trained into it as a knowledge base. What was it like to be part of the process of your own expertise going into a product?

Sanchez: Even back when I was in other organizations, I was always trying to pass on what I'd learned. The difference is that I used to pass it to people, and now the recipient is AI. Running internal seminars to share best practices, sharing manuals — all of that is knowledge transfer. But that kind of effort felt one-off, like it evaporated. Unless you're teaching one-on-one or working a project together, getting a large group to internalize something is close to impossible.

Before we had Dataspace AI, Rim (Song Kyung-lim, COO) built "Sanchez Bot," trained on the research know-how that internal experts including me had left behind in documents and Slack, and we used it inside Opensurvey. Junior researchers could take simple questions to Sanchez Bot first, which let me concentrate on the more fundamental ones. Dataspace AI works the same way — it takes clients' basic inquiries and the simple but labor-intensive tasks, so Opensurvey's internal experts can spend their time on deeper questions and business problems. From a researcher's point of view, I think that's largely a good thing.

From the client's side, the fact that Opensurvey connected 15-plus years of know-how in the research domain to Dataspace AI is a real advantage. You don't have to go looking for that knowledge deliberately — you pick it up naturally in the course of using Dataspace AI. I think knowledge of research and data use will end up more solid than it's ever been.

At the same time, with individual know-how moving into Dataspace AI, I felt I needed to build up other capabilities as a researcher. It made me think harder about what role I should be playing, and it pushed me to stay sharper on research methodology and trends rather than settling where I am. Rather than stopping at knowing statistics and writing good questionnaires and reports, I've come to think more deeply about how to solve a client's business problem through research — and in that process, using AI more effectively and efficiently feels like the capability that will matter more for me as a researcher.

After a lot of trial and error and internal validation, Dataspace AI launched in April 2026. As working researchers, what was it like using it in your own work? We're also curious about the moments where it's most useful in practice.

June: The biggest change is the workflow for designing and analyzing surveys. Now, when I build a survey, I don't bother with Excel — I work in the editor with the Dataspace AI sidebar open beside it. The best part is that if there are surveys already in the space, I can build a new one by referencing them. I don't have to remember which survey a past question was in; I just ask Dataspace AI and it finds it right away.

Analysis is going well with it too. This was early after launch — a project I was handling had so many questions that it was awkward to analyze directly on the results page. Using Dataspace AI to compare multiple sets of data, I could run the comparison immediately without any separate merging work, and that felt genuinely useful. I gave Dataspace AI a huge amount of work, and by the time it was done I'd used six months' worth of credits from a paid subscription account. It got me through the presentation in good shape.

For clients, I think people who aren't used to research will get a lot of help just getting started, and people who are used to it will feel the benefit most in analysis. Analysis that used to require going through your assigned AM can now be done directly with Dataspace AI as long as you have the data. For both survey writing and analysis, the more data accumulates, the more useful it gets and the smarter you can be with it.

You could use a general-purpose LLM for research work too. Is there a particular reason you reach for Dataspace AI?

June: More than anything, it's that it sits inside the workflow. Once LLM performance leveled up across the board, accessibility became the deciding factor. For example, when I'm writing something in Google Docs, I naturally end up using Gemini — research is the same. When I'm building a survey in the Dataspace editor or analyzing data on the results page, using the Dataspace AI that's attached right there feels like the most efficient thing. I do use general-purpose LLMs like Claude, but only when I need an analysis method Dataspace AI doesn't support.

Sanchez: For me, the split probably tilts a bit toward general-purpose LLMs. A lot of my work is listening to a client's problem and finding a suitable research methodology, or doing early exploration, and for that purpose I use general-purpose LLMs, where I can look across a wide range of information. Everywhere else, I go to Dataspace AI. With clients who have a research history with Opensurvey, there are cases where you need to refer to earlier research, or run the next study on the basis of past insights, and Dataspace AI is a real help there. It answers based on the questionnaires, result data, insight reports, and so on. When I'm scanning result data, or wrestling with the angle to take while writing a report — in other words, whenever I need to do something grounded in data — that's when I use Dataspace AI.

A lot of clients like Dataspace's analysis features. Did you notice anything that sets Dataspace AI apart from general-purpose LLMs when analyzing data?

Sanchez: Dataspace AI is very honest. It doesn't over-read the data, and it's conservative in that sense. General-purpose LLMs, for instance, will confidently build a story even when there's no statistically significant difference. For a practitioner who doesn't know research well and finds data analysis difficult, that's a dangerous answer, because they can't verify it. You don't get that when you analyze with Dataspace AI. Because of that, I think it does a good job as a compass for setting the starting point of an analysis.

June: Agreed. Dataspace AI doesn't exaggerate or force a narrative arc. I once analyzed open-ended responses for a client project. A general-purpose LLM gave me an analysis saying that reasons for satisfaction and dissatisfaction differed by product model. That wasn't something Dataspace AI had produced from the same data. So I went back and cross-checked with Dataspace AI, and it told me that only three responses actually mentioned a specific model in their evaluation, so you couldn't interpret the reasons for satisfaction and dissatisfaction as differing by model.

General-purpose LLMs give you a rich answer to anything you ask. At a glance it all looks plausible. But research is about making decisions based on data, so you can't build a story that goes beyond the range of the consumer data. From one angle, Dataspace AI's answers might feel thin compared with a general-purpose LLM. But it holds the line at the data — it never crosses it. For me, that's the reason I can trust it.

Part 3. AI That Extends Your Capability Rather Than Replacing It

For two people with long research careers, what has Dataspace AI become to you?

June: An assistant I can rely on. As I just said, general-purpose LLMs often get things wrong when handling numbers, so they always need careful verification. With Dataspace AI, I don't worry about a figure being wrong. Early after launch I verified every analysis I asked for, and it was never wrong. So for me it's an assistant I trust. I think that's a major strength for a research tool, especially at the analysis stage.

Sanchez: It feels like a companion in research. For people with a lot of research experience and people with little, it substantially lowers the barrier of knowledge and procedure that consumer research requires. It'll be a big help to anyone trying to work through a business problem with data. Just as the AMs in our business group have improved their basic research skills and raised the level of their thinking by using AI, I think Dataspace AI will raise the level at which people understand consumers through data.

Finally, if you had one thing to say to researchers who'll be working alongside AI, and to people who are just starting out in research, what would it be?

Sanchez: If you're curious about the data in your own research, I’d encourage you to take that small extra step and try Dataspace AI. Practitioners find working with data hard, but Dataspace AI is conversational, so it's as simple as using a general-purpose LLM. Throw even a short question at it and a new world opens up.

June: First, for people who aren't researchers but have to run research regularly and find it a struggle — I think the benefit they'll feel from Dataspace AI is significant. You don't need specialist knowledge, because the AI guides you. Most people are more comfortable reading data as text or charts than as numbers, and Dataspace AI reads and organizes the complicated numbers for you, which lowers the hurdle to using data considerably. And for researchers like me, I think the synthetic consumer feature we recently launched will be the wow moment. Building a segment profile and exploring by talking directly with that persona is a different kind of experience from analyzing data. It answers more of your questions and opens up new lines of thinking. Whatever your role, if you want to understand consumers through data, I'd recommend meeting Dataspace first.

People talk about AI replacing researchers. But these two say AI isn't taking the researcher's seat — it's bringing the areas that need real expert judgment into sharper focus. Thanks to an AI that knows research well, researchers can concentrate on what matters, and non-specialists who need research can make more active use of consumer data as the field opens up. Meet Dataspace AI, at the center of that change.