Embrace an Iterative Approach to the Development of AI Solutions


Baringa’s Paul Jones shares insights into data analytics and AI strategies that unlock the full potential of data. 

Baringa prioritises the development of strong client relationships and tailors its approach to meet unique challenges and objectives. This strategy ensures the delivery of impactful solutions that meet the specific needs of clients. Paul Jones, the Director of Data Analytics & AI Practice at Baringa, is a highly experienced professional in the field and the author of The Data Garden. He provides valuable insights into data analytics and AI, shedding light on strategies and techniques to unlock the full potential of data.

Jones is responsible for developing data strategies to drive value, designing and building data teams, and leading enterprise-wide data transformations. He emphasises Baringa’s people-first approach, which prioritises cross-functional teams with diverse skills and experiences. According to Jones, successful data and AI projects require technical expertise and business change skills to support cultural shifts and train end-users. Building and working with cross-functional teams is essential to achieving clients’ business goals. Effective communication, collaboration, and the ability to bring together different perspectives and areas of expertise are crucial to success.

He addresses the biggest hurdles for enterprises embarking on their AI journey, the skills data science professionals should focus on, the innovative use of AI, and how to detect and address model decay. Jones and Baringa’s approach to data analytics and AI is focused on delivering tailored, impactful solutions that meet the specific needs of their clients.  

Excerpts from the interview;

With generative AI taking on more cognitive labour in coding, what skills should data science professionals focus on?

In my view, it is crucial for data science professionals to have a solid foundation in programming, data analysis, and machine learning. Even if AI takes on more cognitive labour in coding, understanding how things work will help data scientists use generative AI and other tools more effectively to support their coding.

As AI continues to evolve and automate coding, it is increasingly important for data scientists to focus on developing skills that cannot be easily automated. Domain expertise, for instance, is a crucial capability that allows data scientists to provide valuable insights and context that AI systems alone cannot.

Another important capability for data scientists is communication and collaboration skills, as AI systems must be integrated and used with human expertise. This includes explaining complex concepts to non-technical audiences and working effectively with cross-functional teams.

Furthermore, staying up to date with the latest developments in AI and related technologies is always important, as the field is constantly evolving.

What are the biggest barriers for enterprises beginning their AI journey, and what advice would you give them?

One of the most significant hurdles enterprises face when embarking on their AI journey is a lack of understanding of how AI can be applied to solve business problems.

Before tackling the technical challenges of implementing AI solutions, it’s essential to clearly understand the problems you are trying to solve or the value you seek to deliver. While AI provides a variety of tools and techniques to solve business problems, implementing AI without a clear purpose is unlikely to yield a positive return on investment.

I would advise starting small and focusing on one specific use case or problem that AI can address. Choose the right use case that is significant enough to deliver tangible value but small enough to be completed within a reasonable timeframe. You are more likely to achieve success in this case.

Take the time to learn from each experience and embrace an iterative approach to the development of AI solutions. Experimentation, iteration, and continuous improvement are critical to delivering value quickly and sustainably.

What can readers learn from your book The Data Garden?

One of the most common challenges organisations face when attempting to drive more value through their management and use of data is a lack of understanding of what data management is and why it is important.

That’s why I wrote The Data Garden, which consists of six fictional stories that bring key data management topics to life for non-technical readers.

The book offers an entertaining and thought-provoking way to learn about key concepts, including:

  • Data Management and the Career of a Data Manager (The Data Garden)
  • Data Governance (The Data Governance Country)
  • Data Quality Management (The Data Quality Hospital)
  • Data Architecture and building Data Transformation teams (The Data Architecture Construction Project)
  • Metadata Management (The Metadata Mess)
  • Data Literacy (The Data Literacy Driving School)

The stories aim to make data management accessible to people without technical backgrounds by using analogies and easy-to-understand language that anyone can relate to.

How can data analytics and AI be used to inculcate a culture of innovation in enterprises?

Data analytics and AI are not only powerful tools for driving innovation, but their implementation can also directly support the creation of a more innovative culture.

AI solutions can be employed to identify areas for innovation by analysing customer behaviour and market trends. Based on this analysis, opportunities for innovation can be suggested, making AI a key component in a range of innovative solutions. Predictive analytics, automation, prescriptive analytics, and other techniques can be used conventionally or innovatively to deliver new and exciting outcomes.

Furthermore, the iterative and experimental nature of data analytics and AI solutions can create space for a variety of possible solutions. This way of working can lead to more innovation, adding to the already impressive potential of AI technologies.

While fostering a culture of innovation involves more than just implementing AI technologies, data analytics and AI play an important role in creating such a culture. 

Considering the dynamic market and changing customer trends, how can data leaders better track model decay and ensure they are relevant for new circumstances?

As the world changes, the ability of a model to consistently deliver the level of performance it did when it was first trained reduces. For instance, if a model is trained based on one set of market conditions, it’s possible that the model may not be able to perform for the new situation when those market conditions change. This degradation of performance over time, due to changes in the problem space that the model is designed for, is what we refer to as model decay.

Model decay can result in poor business decisions or missed opportunities if it’s not detected and addressed promptly. Therefore, it’s crucial for data leaders to monitor and assess their models’ performance continuously and make updates or improvements as necessary to ensure they remain relevant and useful as circumstances change.

To track and mitigate model decay, various approaches can be used. At a high level, these include:

  • Continuously monitoring model performance: By continuously evaluating the performance of models enables model decay to be identified early so that it can be addressed.

  • Conducting regular model retraining: Machine learning models will require retraining to adapt to new data and changing business requirements, so this should be planned into the way the model is maintained.

  • Staying informed about market trends and customer behaviour: Beyond the technical side of a model, understanding the problem space that the model has been designed for and keeping up to speed with changes that might impact a model’s effectiveness is a key way to be proactive about addressing model decay

  • Implementing an Agile development approach: agile development approaches enable a quick response to changing business requirements and market conditions. This involves continuous testing and iteration, with a focus on delivering value quickly and efficiently.

For more information and registration, visit Velocity UAE.