Agentic AI: Hype or teh Next Industrial Revolution?
Table of Contents
The Rise of Agentic AI: A Transformative Force?
The question on many experts’ minds is whether agentic AI will become as indispensable to businesses as the personal computer was in the 1990s. This technology, designed to act autonomously by making decisions and performing tasks, stands in contrast to generative AI, which primarily focuses on content creation. But is it ready for prime time?
While generative AI excels at producing text, images, and audio, agentic AI aims to proactively execute tasks. Though, the very definition of agentic AI
is a point of contention. some view it as overhyped, while others see its transformative potential. Ryan Salva, a former VP at github and now a senior director at Google, expressed his frustration with the term, stating, I hate the word agent… the industry abuses it to the point of making him lose all his meaning.
Key Challenges Facing Agentic AI Adoption
Despite the potential, several significant hurdles must be overcome before agentic AI can be widely adopted. These challenges range from energy consumption to data quality and the sophistication of learning algorithms.
The Energy Consumption Conundrum
One of the most pressing concerns is the insatiable energy demand of AI. The large-scale deployment of agentic AI hinges on resolving the energy crisis that accompanies its growth. The industry faces colossal energy demands, with hyperscalers even considering nuclear energy to power their AI infrastructure. Amit Walia,CEO of Informatica,notes that massive investments in GPUs and IA infrastructure recall the major past industrial revolutions,where rupture technologies have redefined the economy. But beyond equipment, energy efficiency will be a key factor in the adoption of AI.
AI models, especially those requiring real-time decision-making, demand considerable computational power. Companies that fail to optimize their infrastructure could face unsustainable operational costs. The adoption of more energy-efficient AI models is crucial.As Walia suggests,More effective AI agents,consuming less energy and reducing operational costs while aligning on environmental objectives,will be much more attractive.
Smarter Learning Algorithms: The Key to Adaptability
Beyond energy consumption, agentic AI requires more advanced learning and adaptation capabilities than traditional models. Reinforcement learning (RL) emerges as a promising solution. RL allows AI agents to refine their behavior over time by using real and synthetic data to simulate various scenarios. Srinivas Njay, CEO of Interface, emphasizes the central role of RL for AI agents handling complex tasks, stating, For agency AI, which must perform start-to-end tasks, the RL allows you to navigate in decision-making trees, adapt to changing conditions and improve continuously thanks to learning through experience. Rather than just generating an answer, the agent learns to act to obtain concrete results.
However, RL is not a panacea. It has limitations, including high data and computational costs, a lack of interpretability in decision-making, and limited transferability to new environments. Consequently, advanced AI applications now combine RL with other approaches, such as supervised learning, unsupervised learning, and retrieval-augmented generation (RAG) techniques, to overcome these limitations.
As AI evolves, increasingly complex algorithms will be necessary to avoid redundancy. Currently, many AI models perform similar tasks, distinguished onyl by slight advantages in specific areas. continuous technological development is essential to move beyond this standardization.
the Data Challenge: Quality and Availability
Data is the bedrock of AI performance, but it also presents a significant obstacle to the growth of agentic AI. An AI agent’s effectiveness hinges on the quality of its training data. Without high-quality, domain-specific details, it struggles to operate effectively in sectors like healthcare, finance, or customer service. According to Amit Walia,Our latest report on data directors reveals that 43% of companies consider the quality,completeness and availability of data as their main obstacle to the deployment of AI. Without high quality data and adapted to a specific domain, even the most advanced AI models will reach their limits.
Srinivas Njay highlights this issue in the financial services sector, where agentic AI is being tested for online banking and fraud detection. Data silos, regulatory constraints and inconsistent formats complicate the task of AI, which then struggle to act with confidence.
He advocates for a deep modernization of data infrastructure, emphasizing that the key lies in the unification of silos and real-time access to reliable and quality information.
Current Limitations and Future Prospects
Despite the excitement surrounding agentic AI, most companies are not yet ready to entrust critical decision-making to these agents, especially in sensitive areas like customer relations, financial transactions, or strategic planning. Amit Walia explains, There is no doubt that the agents of AI will transform the business world, but today, they excel especially in structured and repetitive tasks. Where enthusiasm goes beyond reality, it is in decision-making at high stake.humans create things. And managing these people is essential. AI agents are not yet able to manage complex customer relationships or operate without human supervision.
Srinivas Njay concurs,noting that while AI can automate processes like litigation processing or loan requests,human supervision remains essential for more complex decisions. The fundamental principle is as follows: AI for tasks, humans for judgment. IA agents excellent in rules based on rules, but humans remain essential to lay limits, especially in situations where confidence and empathy are essential.
A Strategic Approach is Key
we are only at the beginning of a transformation that could unfold over time, similar to the evolution of corporate software and cloud computing. While AI agents will inevitably become more capable, companies must prioritize fundamentals: ensuring data quality and availability, fostering an AI-centric culture within teams, and integrating these technologies to generate tangible productivity gains.
Amit Walia emphasizes that It all starts to be implemented: progress in terms of GPU efficiency, strengthening and optimization of data specific to each sector accelerate adoption. Companies that will master the management of their data and collaborate effectively with AI agents will be best placed to open the way.
Predicting the precise timeline for agentic AI to fundamentally transform business operations remains challenging. However, those who adopt a strategic approach, rather than succumbing to opportunistic enthusiasm, are poised to reap the most significant benefits.
