The 80% Solution: Where AI Actually Adds Value in Market Research

By Forsta

• article

• Innovation Research
• Dashboards
• Data Visualisation
• Survey Analysis
• AI
• Reporting
• Automated Reporting

Tobi Andersson is General Manager for Market Research at Forsta. He founded Dapresy – now Forsta Visualizations – in 2003 after recognising the need for better data visualisation whilst working at a fieldwork house in 1999. Over 15 years, he scaled Dapresy from a local Swedish company to a global business with over 100 employees.

Learn more by watching or listening to Tobi on the Founders and Leaders Series podcast here:

Over the past 25 years working in market research technology, I have seen several significant trend shifts. The move from telephone to web interviews around 1999-2000. The integration of multiple data sources. And now, artificial intelligence.

What makes this current shift different is the speed at which AI capabilities have matured. In the last six to twelve months alone, we have moved from interesting pilot projects to concrete applications that are delivering measurable value in production environments. But we need to be clear-eyed about where that value actually lies.

The challenge with any technological hype cycle is that isolated solutions often look impressive in demonstration videos, yet struggle to fit into the day-to-day reality of how research organisations actually work. I learned this the hard way building Dapresy. Having a brilliant standalone tool means nothing if it requires people to completely change their established processes. The friction is simply too high.

So where is AI actually adding value today? In my view, there are three specific areas where we are seeing tangible benefits: automation of repetitive tasks, data preparation, and first-draft reporting.

Automating Survey Scripting

For more than 20 years, we have manually scripted surveys – dragging and dropping, copy-pasting question text, and setting routing logic line by line. This manual approach will not disappear overnight, but over the next 12 to 24 months we will see it increasingly complemented by automated processes.

The technology can now take a questionnaire specification and generate a functional survey script. It understands question types, routing patterns, and quota structures. More importantly, it can learn the specific conventions that your organisation uses. Every research department has its own style and standards for how questions should be formatted or how data should be coded. AI can observe these patterns and replicate them.

Does this eliminate the need for human oversight? Absolutely not. You still need experienced researchers to review the output, refine the logic, and ensure the questionnaire will collect the data you actually need. But you reach the review stage much faster, and your team can focus on strategic aspects rather than mechanical typing.

Preparing Data More Efficiently

Anyone who has worked with market research data knows there is a substantial preparation phase before analysis can begin. You decide whether to include or exclude certain response options, create top-box and bottom-box measures, combine multiple variables into summary questions, and build derived variables based on your analytical framework.

These tasks require judgment, but they are also highly repetitive. Once you have decided how to prepare data for a tracking study, you will likely apply the same logic to every subsequent wave. Once your organisation has established its standard approach to brand health metrics, that approach should be consistent across projects.

This is precisely where AI can learn and automate. The system observes the transformations you apply to raw data, understands the sequence of steps you take, recognises the business rules you follow, and can then apply those same transformations automatically to new datasets.

The result is not just speed, but also consistency. When data preparation is manual, there is always the risk that someone applies a slightly different rule, introduces a small inconsistency, or simply makes an error at the end of a long day. Automated preparation, once properly configured, eliminates that variability.

Again, this does not remove the human element. You still need someone to define the rules, validate the output, and make judgment calls when unusual data patterns appear. But you reach the point where you can begin analysis much faster, and with greater confidence in the consistency of your data.

Generating First-Draft Reports

Perhaps the most visible application of AI in research today is in report generation. The technology can now analyse your data, identify the most significant findings, and create a first draft of a PowerPoint presentation.

To be clear about what I mean by “first draft”: the AI can understand what background variables you have, how your questions are structured, and what your typical reporting format looks like. It can select appropriate charts, write initial commentary on the findings, and organise slides in a logical sequence.

What you get is perhaps 80% of the way to where you need to be. Then, as the researcher, you handle the final 20%. You refine the narrative, add context that only someone with deep knowledge of the client can provide, identify the implications that require strategic judgment, and ensure the story makes sense and drives actionable recommendations.

This 80-20 split is important to understand. If someone promises that AI will produce a finished report with no human involvement, they are either overselling their technology or they have a very different definition of what constitutes a quality research deliverable.

Nevertheless, getting to 80% automatically is genuinely valuable. It means your senior researchers can spend more time on interpretation and strategy, and less time on the mechanical task of creating charts and formatting slides. It also means faster turnaround times for clients and the ability to handle larger volumes of work without proportionally increasing headcount.

The Question-and-Answer Interface

There is one more area worth mentioning, though it is still evolving. Many of us now have a new behaviour in our daily lives. When we want to know something, we ask ChatGPT or Gemini. We type a question and get an answer back. Then we ask a follow-up question and go deeper.

This same interaction pattern is starting to appear in how people consume research data. Instead of navigating through tabs and filters in a traditional dashboard, users simply ask: “What do millennials think about this product feature?” or “Where do I see the biggest gap between stated preference and actual behaviour?”.

The system retrieves the relevant data, performs the appropriate analysis, and provides an answer. If that answer raises new questions, the user can continue the conversation. If they want to see the underlying data tables or validate the finding with a different cut, they can request that.

This does not replace traditional reporting and visualisation. Those remain essential, particularly for communicating findings to stakeholders who need to see the full picture. But the conversational interface adds a new entry point to the data. It makes insights accessible to people who might not have the time or expertise to navigate a complex dashboard.

Staying Grounded

I have genuine enthusiasm about what AI can do for our industry, but I also want to be realistic. These applications work because they fit into existing research processes. They make current tasks faster and more consistent, and do not require organisations to fundamentally change how they operate.

That is the test I would encourage everyone to apply when evaluating AI solutions. Does this tool make something you already do more efficient? Does it fit into your current workflow with minimal friction? Does it get you 80% of the way there, whilst still recognising that your expertise and judgment deliver the final 20%?

If the answer to those questions is yes, then you are looking at a genuine application of AI that will add value. If instead the promise is complete automation, radical transformation, or the elimination of the need for experienced researchers, then I would suggest a healthy dose of scepticism.

We are in an exciting period of technological development, and the organisations that will benefit most are those that stay focused on solving real problems in practical ways, rather than chasing the hype.

Learn more by watching or listening to Tobi on the Founders and Leaders Series podcast here:

Author

Learn more about