As businesses grapple with an ever-evolving landscape of data management and regulatory compliance, leveraging artificial intelligence has become not just an option, but a necessity. To shed light on this critical intersection, we spoke with Dr. Stefan Oglesby, a leading expert in AI and data strategy, about how organisations can harness the power of AI to drive innovation, enhance compliance, and gain deeper market insights.

This conversation comes at a pivotal time for businesses. As the pressures around Governance, Risk, and Compliance (GRC) and data management continue to mount, understanding how to effectively leverage AI is more important than ever. With data becoming a central part of every business decision and regulations getting stricter, companies need to find smarter ways to manage their information.

Introduction to AI and Data Classification:

Wimima4: What are some of the biggest misconceptions about AI in data classification that you encounter? How would you address them?

Stefan: The probably biggest misconception is overestimating the capabilities of AI. AI will not automatically deliver valid results without prior human input. AI models require proper training regarding their specific task. The output of AI models also needs continuous human oversight to ensure accuracy of results. A further potential misconception is that AI can replace human judgment. In general, AI can augment human decision-making, not replace it. Domain expertise is crucial. Consequently, the challenge is how to optimize efficiency with a maximum level of automation, whilst ensuring human-level quality control and input.

Wimima4: From your experience, what are the initial steps organisations should take when starting their AI-driven data classification journey?

Stefan: As any data related journey, you should start with strategy, as Bernard Marr puts it, i.e.  specify the business problem AI es expected to actually address. The output is as good as the input; therefore a proper data preparation process is crucial. The adequate infrastructure, i.e. tools, platforms, and teams, must be carefully selected, balancing costs and time. Finally, a “pilot and iterate” approach will avoid the risk of high costs and delays. Starting with a small, focused pilot project is highly recommended, followed by iteration based on the results.  

AI Algorithms and Contextual Analysis:

Wimima4: What are some of the most promising AI algorithms or techniques currently being used for contextual data analysis? How are they changing the landscape of data management?

Stefan: Classifying data with regard to PII (personal identified information), but also identifying e.g. confidential or proprietary topics of a document, can be addressed with a range of algorithms and techniques. Named Entity Recognition (NER) models based on Transformer architectures, such as BERT (Bidirectional Encoder Representations from Transformers) help to extract PII in large text data, which can be a first step towards classification. LLMs (large language models), such as GPT, are also method of choice for identifying critical topics and content through their capability of understanding context and semantics in unstructured data.

Reinforcement Learning from human feedback is probably needed to fine tune off-the-shelf models for the requirements of a given organisation, helping to optimize a continuous improvement of the classification processes.

Wimima4: What are the biggest challenges you see organisations facing when implementing AI for contextual analysis, and how can they overcome them?

Often, skill gaps and misjudging the required change management are significant barriers.  Cross-functional teams need to be established. Ongoing training must be provided, to achieve the necessary upskilling.

Classification has practical implications on how teams use and analyze data. An in transparent black box model may jeopardize the acceptance of the AI based classification approaches throughout an organisation. Investing into at least partially explainable and interpretable models will mitigate this risk.

Benefits of AI-Driven Classifications:

Wimima4: In your opinion, what are the most underutilized benefits of AI-driven data classification? Why do you think organisations are hesitant to fully leverage these benefits?

Stefan: I see the following underutilized benefits:

  • First, Enhanced Data Security: AI can automatically identify and protect sensitive data, reducing human error.
  • Second, Regulatory Compliance: AI streamlines compliance by continuously monitoring and classifying data according to evolving regulations.
  • Finally, organisations often hesitate due to concerns over initial costs, integration complexity, and a lack of in-house AI expertise.

Wimima4: How does AI-driven classification help in addressing ‘shadow data’ and ensuring more accurate data management?

Future Outlook and Innovations:

Wimima4: Looking ahead, what emerging trends or technologies do you think will have the most significant impact on AI-driven data classification and regulatory reporting?

Stefan: Parallel to the current gen AI hype, mistrust has also grown, which can go as far as rejection. The extensive and complex EU AI Act is an obvious symptom of this development. All stakeholders are therefore urged to enhances transparency and trust in AI-driven decisions, which are crucial for regulatory environments. I expect explainable AI (XAI) to be the most important emerging trend across the up-coming real-world applications of AI, including AI driven data classification.

Federated Learning is an approach allowing to collaboratively train a model while ensuring that their data remains decentralized. Avoiding sharing sensitive data while training models across decentralized data sources will improve data privacy, helping to comply with relevant regulations.

On a more technical side, Automated Machine Learning (AutoML) will enable faster, more efficient deployment of data classification models.

Wimima4: How do you envision the integration of AI with other technologies, like blockchain, transforming regulatory compliance in the future?

Stefan: The “immutable” characteristic of a blockchain ensures tamper proof data records. Blockchain technology can streamline audits, and it makes sure that any classified data – together with their classification – have not been manipulated or tampered.

Blockchain technology further can automatically execute compliance processes through self-executing smart contracts. With this scenario, the results of the AI driven data classification are stored on chain in a transparent and immutable way. Access to critical off-chain data can also be securely managed with blockchain technology. The combination of tamper-proof data, immutable classification, and decentralized data access control will result in a “trustless” approach to protecting critical data – making it immune against human error or manipulation.  However, such an approach may be an over-complex solution to a limited problem.

Practical Advice and Strategies:

Wimima4: For organisations hesitant to adopt AI for data management, what practical advice would you offer to overcome their concerns and start the integration process?

Stefan: Start small with a pilot: Begin with a small, manageable AI project to demonstrate value and build confidence. This allows organisations to see quick wins without a huge upfront investment. Focus on data quality and security: Address concerns by ensuring robust data governance, privacy, and security measures are in place.

Invest in education and training: Upskill employees to understand AI’s benefits and how it complements, not replaces, human expertise.

Wimima4: What should organisations consider when choosing an AI technology partner, especially in the field of data classification and compliance?

Stefan: Obviously, it is important to look for a partner with domain expertise in data classification. AI is just a new instrument in the toolbox which needs to be adapted to the specific requirements of the tasks to be executed. Therefore, I would focus on a potential partner’s track record in data security and compliance.

Flexible solutions that can be tailored to your specific needs and ongoing support and training are a must have.

I also recommend to look for partners who have shown a clear commitment to ethical practices in the past. I would also ask about their approach to explainability of AI models, to maintain trust and meet regulatory standards.

The Role of AI in Market Research:

Wimima4: As you are building up Insight Lab, an AI-driven qualitative market research platform, could you share how AI is transforming market research and what unique challenges it presents?

Stefan: AI is transforming market research by automating the transcription, processing, and analysis of large volumes of qualitative data, such as customer feedback and interviews/ focus groups, with unprecedented speed and accuracy. This automation allows researchers to focus on higher-value tasks like strategic analysis and decision-making. Generative AI can also speed-up the collection of deeper, qualitative customer feedback, by automating neutral, motivating probes during online interviews.

However, the challenges include ensuring the AI correctly interprets nuances and sentiments in data, managing data privacy, and maintaining the quality and accuracy of the insights generated. Generative AI is the never sleeping assistant to the market researcher. AI cannot replace the in-depth market and strategic understanding of the experienced senior researcher. Human oversight is still pivotal to validate AI findings. Insight Lab addresses these challenges by offering customizable tools and collaborative features that enhance both the efficiency and reliability of market research outputs.

Wimima4: How can AI-driven market research tools provide deeper insights that were previously difficult to obtain?

Stefan: Costs and timeline have been major barriers for collecting qualitative, deeper insights about consumers’ behavioural patterns, needs and wants. Ai-driven market research tools can reduce the time resources needed for the analysis of large amounts of consumer conversation by up to 70%. As a result, qualitative, deep insights get affordable and viable for a much broader range of strategic and tactical marketing challenges. Further, optimized AI tools, based on advanced and specifically fine-tuned LLMs (large language models) can analyse unstructured data in a highly accurate way, often avoiding the bias brought in by less experienced junior researchers, who often do the heavy lifting of qual projects.

I strongly believe, that the benefits of AI go far beyond costs and efficacy. Things can be done with Ai that could not be done before! With a “chat your data” approach, the decision maker can use the original, anonymized consumer conversations to answer questions at any time, and to support ideation processes. For example, a product manager can use an AI tool to create innovative product and services concepts based on the authentic, unfiltered voice of customers.

Personal Insights and Reflections:

Wimima4: What inspired you to focus on AI and data strategy in your career? Are there any personal experiences or insights you can share that have shaped your views on the future of AI in data management?

Stefan: The inspiration to focus on AI and data strategy in my career stemmed from witnessing the transformative power of data-driven decision-making and the efficiency AI introduces in processing and analyzing large datasets. Personal experiences with the limitations of traditional data analysis methods and the rapid evolution of AI capabilities highlighted the potential for significant advancements in this field. These experiences shaped my belief that AI will play a crucial role in the future of data management, particularly in enhancing precision and speed.

Wimima4: If you could foresee a major breakthrough in AI for data classification or compliance within the next decade, what would it be, and why?

In Summary

Dr. Stefan Oglesby’s expertise offers a compelling vision for the future of AI in data management and market research. As we navigate an era increasingly dominated by data-driven strategies, the insights shared in this interview underscore the importance of embracing AI not just as a tool for automation but as a strategic partner in innovation and compliance. Organisations that invest in robust data governance frameworks, upskill their teams, and carefully select their AI technology partners will be well-positioned to reap the full benefits of AI. By doing so, they can turn potential challenges into opportunities, fostering a culture of continuous improvement and agility. As Dr. Oglesby suggests, the journey towards AI integration is not without its hurdles, but with the right approach and mindset, it can lead to transformative outcomes for businesses worldwide.

More about our Expert

Dr. Stefan Oglesby is a seasoned data and AI strategist with a passion for driving innovation in data analytics and consumer insight. With a strong background in leadership roles across various industries, he excels at turning complex data into actionable business strategies. Currently, Dr. Oglesby is focused on building Insight Lab, an AI-driven qualitative market research platform that leverages advanced technology to deliver deeper insights. His expertise spans across multiple sectors, including consumer health, finance, and media, making him a versatile leader in the field of data science and AI.