Scaling insights

Editor's note: JD Deitch is founder and CEO of Passaggio. He can be reached at hello@jddeitch.com.

Artificial intelligence is causing a tectonic shift in the way research is conducted and research firms function. This is not merely a change in technology, as we witnessed with the internet and mass adoption of smartphones. It is far more profound. AI is redefining what researchers do and what they need to know, transforming them from hands-on operators to strategic overseers and product developers.

What does it mean for the industry when AI can take on every phase of the research process and perform the work of a massive research team? Let’s explore how work is done in traditional firms and compare it with the cutting-edge, equally competent approaches that are emerging from AI-enabled firms.

The new role of researchers

Since the dawn of market research, there has been one and only one way to build scale: through labor. Even in the digital age, the largest research firms have been those with the ability to muster battalions of researchers to design, run and interpret research. There was hope that insights platforms might change this. Yet while upstart firms have built competent, user-friendly, more labor-efficient platforms to execute research, they essentially transferred the labor problem to the client.

Advances in AI are now eliminating large labor pools as a necessary factor for scale in a way that even non-AI-based automation and DIY platforms could not, including those that previously required extensive human oversight. Today, there are already viable companies commercializing AI products that span the entire research process: interpreting clients’ business questions; creating a research brief; designing the research; fielding the research; processing the data; and reporting and interpreting both quantitative and qualitative findings

This means that AI can now effectively replace a full research and operations team, operating at a scale previously unattainable with human labor alone. This evolution from labor-intensive projects to AI-driven products marks a pivotal transformation in the industry. First-movers and disruptors who are not starting from scratch but are instead leveraging AI throughout the entire workflow have a significant advantage. These firms are positioned to take market share by integrating AI comprehensively, rather than using it for specific elements only, which ensures long-term success.

Critical elements of an AI product

Creating a proficient AI system involves a blend of technology, data and human expertise. Here’s a detailed look at the three critical elements, ranked in increasing order of importance:

Algorithms are the heart of AI’s decision-making. This is the set of mathematical instructions that dictates how the AI processes information and makes decisions. The quality of an algorithm is crucial, as it affects the system's predictive power and efficiency. Though foundational algorithms are widely available and relatively straightforward to implement, optimizing them for specific tasks can significantly enhance performance.

Data is the fuel for AI. AI systems require data to learn and refine their abilities. The volume, variety and veracity of this data plays a pivotal role in the training process. High-quality, well-annotated data enables the AI to develop more accurate and reliable outputs. Conversely, poor data quality can lead to flawed or biased decision-making.

Domain experience is the “soul” of the machine, guiding the development and optimization of AI. Domain knowledge is the most vital component, for it ensures that the AI not only performs its tasks competently but does so in a way that is contextually relevant and strategically aligned with specific industry standards and needs. Domain experts define what “good” looks like, help prioritize actions and ensure that the AI system adheres to relevant norms and values of the field within which it operates.

There are viable insight companies today – whose capabilities are improving rapidly – who are using researchers in new ways. In these companies, researchers are working not on projects but on products. They are channeling their domain expertise into the development of the intelligence in a way that will create heretofore unimaginable scale.

Think about it: What we see in large language models today is astonishing. Those who utilize ChatGPT for serious endeavors quickly realize its dual nature: exceptional in general applications yet limited in specialized fields due to its training on broad datasets not tailored for niche expertise. ChatGPT operates by predicting the next word that best fits the context of the prompt given, a process that showcases its robust general-knowledge capabilities. For use cases like ideation and copyediting, ChatGPT (especially the GPT-4 model) can produce some truly interesting and even unexpected insight. However, this approach also reveals its limitations, particularly its struggle with specialized knowledge domains, demonstrating that its expertise, while broad, lacks depth.

Of course, the other thing any serious user will know about ChatGPT is that it is far from perfect. It can be waylaid by any number of biases. Out of the box (that is, without additional structure and guidance), it won't necessarily perform well for highly specialized fields. It hallucinates – if machines can do such things – a term used in AI to describe when a model generates false or irrelevant output based on the gaps or biases in its training data. Occasionally, it will completely flake out. By way of example, older versions are incapable of counting or adding.

This is where domain knowledge becomes critical. Engaging skilled researchers in the development of the intelligence demonstrably improves the outcome – so much so that the quality of even moderately sophisticated research designed, run and interpreted by AI is now indistinguishable from that of an experienced researcher.

The real challenge will be understanding the nuanced way people communicate. Research firms wishing to leverage AI will need to make their own investments in advanced natural language processing techniques to allow for a deeper understanding of consumer sentiment. AI platforms will also need to accommodate multi-modal inputs, be they text, audio, photo or video, to capture communication in its many forms.

These systems are increasingly “good enough” for a wide array of real-world applications, challenging the notion that high-quality research can only be conducted by human experts. Moreover, the pace of improvement in AI technologies suggests that they will become even more competent and indispensable tools, surpassing human capabilities in many areas.

The traditional model of deploying large teams of researchers to manage projects will become obsolete. The future lies in researchers developing products – intelligent, self-operating AI systems that handle the bulk of data analysis and insight generation. This shift means that a smaller number of highly skilled researchers are now required, primarily to “teach” and refine AI systems. These experts will focus on ensuring the AI operates within the correct contexts and maintains high standards of data interpretation and insight generation.

It also foretells an even greater democratization of insights that makes high-quality research accessible to more organizations, enabling smaller research companies to compete on a more level playing field with larger firms.

Change dramatically

AI insights products spell the end of large labor pools as a prerequisite for scale. As a consequence, the competitive landscape of the market research industry is set to change dramatically. We can classify these changes across four factors:

Revenue

  • Any competent AI firm will be able to challenge incumbent research firms, even for large-scale studies. Automation and APIs will facilitate the operational tasks and fieldwork. AI will make the decisions, guided by a handful of experts whose methodological prowess will enable high-quality studies at scale.
  • Firms with recurring work like tracking studies, normed or proprietary measurement work already had a powerful moat: fear of change. With synthetic data, there will be no need to run months of costly parallel testing. This work is now firmly in play.
  • AI has the potential to finally make research accessible to the long and un(der)served tail – companies that would have had to use DIY (and largely didn't) because no research agency would get out of bed for such minimal revenue. This implies go-to-market (GTM) strategies that will be very different from the traditional approaches, requiring innovative and more client-focused methods.

Go-to-market

  • The fact that challenger firms can realistically compete with research industry incumbents will require them to demonstrate their competence to a world that does not know their brand.
  • GTM will become far more marketing- and case study-driven – certainly in the short term, and certainly for firms pursuing long-tail opportunities – as firms provide proof points to both demonstrate their capabilities and educate their clients about what AI research looks like.
  • Closing enterprise deals will still require a strong sales team which can not only convince gatekeeping client-side insights teams but can also navigate the legal and data protection issues associated with AI.

Margins

  • Gross margins should look much healthier in this environment. That said, investment in the development and improvement of the AI will be significant, though capitalizable.

Value creation

  • Consultative know-how: Having people on staff who can join the dots between the client's execution and the research remains important. The client’s goal has never been to run a research project.
  • Methodology know-how: Having seasoned researchers, technologists and other business experts will be even more important for building highly predictive, highly activatable insights.
  • Tech and data: This will be table stakes.
  • Sample: While most AI companies will probably not have proprietary sample assets, accessing real people at scale will be critical. The best will figure out how to solve the endemic challenges of a programmatic sampling ecosystem in which companies struggle to tell the real people from the bots.

One of the most likely outcomes we can expect from the pervasive use of AI is a weakening – if not the outright death – of the iron triangle of insights: better, faster, cheaper. AI will indisputably benefit clients through faster delivery of insights, cost reductions due to automation and the ability to address previously unreachable markets due to enhanced scalability – providing the same essential output as a human researcher without compromising quality.

The evolving workforce

As recently as 2022, the industry was still focused on storytelling as an essential skill. ChatGPT learned it like Neo learned kung fu in “The Matrix.” The future workforce has at least several types of people who exist today. Methodology and operational SMEs are still critical to teach the AI what “good” looks like and how research works. Yet rather than working on projects, they will work on products. For large clients, sales engineers will be critical for explaining how things work and engaging with the client’s developers on data integrations. Likewise, large enterprise clients will need salespeople and key account managers who can provide expertise and evangelize across the client’s organization. Finally, industry experts who understand the client’s business problems will always be welcome.

One looming question is what happens when the number of people the industry needs to do work begins to dwindle, and how quickly that happens. There is a potential for significant displacement of researchers and operations people, most especially the armies of offshore workers who, for years, have called the industry home. Firms that can quickly adapt to and adopt AI-driven models will gain substantial advantages. These firms will not only operate more efficiently but also offer faster, more accurate insights at a lower cost, effectively outpacing competitors that cling to traditional human-heavy research models.

At a crossroads

Market research firms stand at a crossroads. The swift advancement of AI presents both significant challenges and substantial opportunities. To compete, firms will need to embrace these changes, deeply integrating AI into their operating models and redefining the roles of their workforce. This transformation, while undoubtedly disruptive, offers unprecedented opportunities to enhance the efficiency and effectiveness of research and expand the market. AI will undoubtedly remain the dominant force for innovation in consumer insights, creating a competitive advantage, especially for those who have invested early. Firms that adapt will benefit from greater analytical precision, reduced operational costs and improved value creation for clients and their shareholders.