As artificial intelligence (AI) continues to surge beyond the boundaries of known possibilities, the irreplaceable value of human input becomes even more apparent. While technology can craft machines that mimic human thought, the unique touch of human intuition remains unparalleled. When merged, the result is a formidable synergy that drives innovation to new heights.
“It’s not just about teaching machines to think; it’s about integrating human expertise to achieve superior accuracy and reliability,” says Aravindh Manickavasagam, a multifaceted technology specialist who spearheaded groundbreaking AI initiatives at Ernst & Young LLP (EY). As industries evolve, the fusion of human intelligence with machine learning—known as Human-in-the-Loop (HITL)—is becoming a cornerstone of advanced AI systems, offering insights that are redefining the standards of document processing.
The Evolution of AI with “Human in the loop”
Achieving the desired precision in AI systems often demands more than just data labels; it requires a nuanced understanding only humans can provide. The HITL approach embodies this symbiosis, where human input continuously refines and optimizes machine learning models. This way, machine output can be corrected “real time” and models eventually start performing more accurately in real-world applications.
At EY, the implementation of HITL has had a particularly profound impact on document processing. “We recognized early on that our expert consultants possessed deep domain-specific knowledge. By bringing this expertise into our AI systems, we created a robust feedback loop that continuously improved our models,” explains Manickavasagam.
This methodology not only enhanced the accuracy of the models but also ensured their relevance and applicability to specific use cases.
Welcome to the Real World: Applications of HITL at EY
One of the most notable applications of “Human in the loop” at EY is in document intelligence. Traditionally, translating investment management agreements into actionable rules required extensive manual effort. EY developed a document intelligence platform with machine learning models capable of automating this translation process by integrating HITL. These models, continually refined by expert feedback, could check the compliance of trades against investment agreements with unprecedented accuracy.
Another significant application is within the ESG (Environmental, Social, and Governance) and wealth management space. Manickavasagam pioneered the use of human in the loop powered machine learning models to extract and map ESG-related language from offering documents providing clients with valuable insights.
“The ability to systematically analyze and compare ESG disclosures has been a game-changer for our clients,” notes Manickavasagam. “It enables them to make more informed decisions and aligns with the growing emphasis on sustainable and responsible investing.”
The “Human in the Loop” technique was also instrumental in the audit revenue recognition standard. EY’s machine learning models could translate contractual documents into terms and languages that conform to audit standards, significantly streamlining the audit process and reducing potential errors. This capability elevated efficiency while reinforcing the accuracy and reliability of audit reports, which are vital for maintaining trust in financial markets.
The Balancing Game: The Human Touch in AI
Manickavasagam’s role at EY was essential in the end-to-end management of the document intelligence platform. His vision of using Human In The Loop to take advantage of the proficiency of EY’s subject matter experts and consultants has been transformative. “The iterative feedback loop we established allowed our models to evolve continuously,” he explains. “Consultants could train the models while performing their usual work, seeing to it that our AI systems were always up-to-date and highly accurate.”
Despite the success, the integration of HITL has not been without its challenges. Balancing the need for human intervention with the efficiency of automated systems requires careful coordination and management.
“It’s about finding the right equilibrium,” says Manickavasagam. “Too much reliance on human input can slow down processes, while too little can lead to inaccuracies. We aim to achieve a seamless blend where human and machine contributions are optimized.”
Staying in the Loop: The Future of Document Processing
The future of AI and HITL appears promising, with industries increasingly recognizing the value of integrating human know-how “real time” into machine learning processes. As Manickavasagam reflects on his path, he remains optimistic about the potential of these technologies.
“The AI revolution is here, and HITL is at the forefront of this transformation. We can develop AI systems that are highly accurate, ethically sound, and widely applicable by combining the best of human intelligence and machine learning.”
Overall, the innovations led by Aravindh Manickavasagam at EY highlight the progressive potential of HITL in document processing and beyond. As AI continues to strengthen its presence, integrating human skillfulness will remain crucial in making certain that these systems achieve the highest standards of accuracy and reliability.
With leaders like Manickavasagam at the helm, the future of AI shows no signs of dimming. It promises advancements that will benefit industries and society as a whole.
