Looking Back - AI One Year on

In November 2022, ChatGPT was released to the general public, marking a significant milestone in the public perception of AI. Since then, Generative AI has seen explosive growth, growing at a faster rate than Facebook, X (Twitter) or TikTok.

Over the past year, the expectations around AI have also grown rapidly with changes expected in how people live, work and connect with others. Many new tools have been launched, promising to optimize supply chains, personalize customer experiences, streamline job applications and so on…

Meanwhile, in the Market Research industry, AI adoption has also been a hot topic. New products have emerged and existing products have been quick to incorporate generative AI into their offering. In this blog, we're looking back and reflecting on what changed during 2023. Has AI met the expectations of the Market Research industry? What tools have emerged, how are they being used and how has the industry changed as a result?

Spoiler: People are still in charge

Throughout 2023, there were notable technical advancements in Generative AI. For example, multi-modal large language models such as GPT4 are now able to handle video and images in addition to text. Retrieval Augmented Generation (RAG) can integrate large amounts of proprietary data to provide more pertinent, subject-specific responses. However, human oversight and supervision still remain essential across most use cases to ensure precise and accurate outcomes.

Take Microsoft’s Copilot, for instance. Whilst it is a helpful tool for, amongst other things, writing and improving emails, a human still needs to run a sense check for an email before hitting send.

Or, similarly, NotionAI is great for generating initial drafts of blog content, but these drafts are far from polished and require a substantial amount of editing before they're fit for publication.

Even meeting notes transcribed by platforms like Supernormal demand validation to ensure accuracy and completeness.

The ESOMAR 20 Questions to help buyers of AI-based services focuses on critical aspects of AI use in Market Research, including the need for human oversight. A key takeaway is that while technology can advance us significantly, humans must contribute their expertise to achieve optimal results.

But, AI is making us more productive

In Market Research, AI-powered tools are mostly employed to boost the productivity of existing processes. These tools can assist in research design, questionnaire design, data analysis and in streamlining the research process.

For example, SurveyMind speeds up focus group transcription and summarization, identifying differences and similarities.

Listen Labs uses AI to conduct conversational interviews and create digestible summaries with key highlights and automatic tagging of recurring themes

Our own product, codeit combines AI with human oversight to ensure an optimal balance between automation and human curation. Utilizing a 3-step Extract/Refine/Apply model for coding verbatims, codeit enables coding to be significantly more productive without compromising quality and accuracy.

And, giving us new tools

In addition to boosting productivity, several innovative concepts emerged over the past year, promising to change how research is conducted. While it is still too early to draw firm conclusions, these use cases have significant potential in the coming years.

Qualitative Research

One of the most promising disruptive applications of AI in Market Research is its use in analyzing qualitative data. These tools significantly reduce the time needed for the analysis of qualitative data, which has been historically one of the most labor-intensive tasks in Market Research

For example, coLoop is an AI tool that assists with qualitative analysis by uncovering themes, summarizing data, answering questions, and creating slides.

Synthetic Data

A more controversial use case is the use of Synthetic data.

Fairgen is an AI-based solution for boosting niche research samples with synthetic data trained exclusively on data provided by customers.

If successful, tools like this could have a significant impact on the way that research is conducted (or not!) in the years ahead.

Conversational Surveys

Finally, while chatbots and conversational surveys have been employed in Market Research for some time, new concepts like autonomous agents using large language models have significantly improved the performance and usability of these tools.

Yasna uses an AI assistant for conversational research that collects facts, opinions, and ideas from any audience in any country.

Other companies like Tellet offer conversational AI tools that generate survey questions and chat assistants that probe respondents for deeper insights and perspectives that go beyond traditional approaches.

While these tools are promising, we still need to define and cement their role in the research process. It's crucial to integrate these tools into the everyday toolkit of market researchers to ensure real adoption and uptake of AI (yes, blockchain I’m looking at you!).

The future for LLMs

2023 has seen an arms race between AI companies each trying to outperform one another: Anthropic, Google, AWS, OpenAI (Azure), and Mistral all have produced large language models (LLMs) that routinely break the latest performance benchmark, for example the simulated Exams from GPT4.

These large “foundational” models are at the base of most AI technology and advances in Market Research in the past year.

For the end user, the emergence of these large models is a win-win situation as they gain greater flexibility, improved AI products, and more effective solutions that enhance the research process.

But, while generic commercial large language models are convenient, their performance is not yet transformative when it comes to more complicated processes and tasks.

In fact, the paradigm of larger and larger generic models that will handle a multitude of tasks might be at a tipping point. Instead, the ability to easily customize and specialize these models for each task and dataset will become key in the next few years.

Similarly, codeit's platform uses advanced custom Machine Learning models to enhance data coding capabilities. The ability to interchange, improve, and personalize these models represents a likely future direction for AI coding.

A final note

So, in the end, has AI lived up to the hype?

It’s certainly making some things that were not possible before, possible. We firmly believe that human-assisted AI can take us further, help productivity, and save time.

AI's transformative impact is evident in how it continues to disrupt the way humans work together with technology to create enhanced solutions.

However, the complete picture is not clear. AI has not yet replaced existing processes but rather has been making us more productive. It is likely that more transformative changes will happen as new technologies are more widely available and understood.

To excel in an AI-driven world, users must do more than simply input information into a generator. Instead, the key is to learn how to integrate AI tools with human thinking, practices, and techniques.

Book a Demo with codeit AI to learn what AI tools the platform offers users to automatically extract, refine, and apply unstructured data to meet essential requirements for an automated verbatim coding system.