06.07.2023

“Just like with the internet and social media revolution, as an agency we’ve always embraced tech and innovation, investing in discovery to accelerate success. However, to safely and creatively experiment with AI, we first needed to overcome the data privacy obstacle.”, says Verhaeghe.

OpenAI systems are designed to continually collect data to learn and improve, which obviously raises concerns about data ownership and the impact on confidential research data and PI information. Therefore, earlier this year, we developed a proprietary AI research assistant, using the ChatGPT algorithm to conduct a series of experiments with real research data, putting generative AI to the test.

AI needs human data

“With our proprietary AI tech at hand, we defined a series of A/B experiments to gain a better understanding of the potential and implications of generative AI systems for our business and the industry at large. Our experiments were designed to explore well-known challenges related to AI data quality, data sampling and even data hallucination.”, says Verhaeghe.

Generative AI systems are only as smart as the data on which they are trained, typically relying on ‘older’ data. Our experiments demonstrate that applying generative AI as an analysis tool to thick data collected through primary research leads to significantly more actionable research results, compared to relying solely on open-source AI technologies.

AI needs human eyes

Verhaeghe also issues a word of caution: “AI in the hands of a person who doesn’t know how to use it, is not truly artificial intelligence but rather artificial stupidity. Effectively using AI requires mastering the art of asking the right questions, fully understanding clients’ needs and context, and prompting the system appropriately. This is where trained humans come into play.”

To address this, we has launched an AI prompt school for its consultants, enabling our global workforce to successfully collaborate with artificial intelligence while ensuring data quality and diversity.

Enhancing the tech suite with a personal AI research assistant

Following our research-on-research journey, we have now implemented its personal AI research assistant across its global business to bolster project efficiency. Harnessing the power of AI automation, we envision more efficient analysis of unstructured data, allowing more room for human interpretation and consultancy.

Verhaeghe concludes: “We firmly believe that relying solely on AI algorithms trained on public internet data will not provide the necessary insights for companies to gain a competitive advantage. By engaging the right people in more versatile ways, we can gather diverse deep data. And when this is combined with consultants who thoroughly understand the clients’ business and know how to prompt AI systems effectively, fresh and actionable insights are generated. Just as humans need machines, machines also need humans.”

Want to learn more on Human8’s AI research-on-research journey?
Tune in for our upcoming virtual event on Wed Aug 2!