With the Impending Rise of Agentic AI Systems, AGI Does Not Seem Too Far

• Artificial Intelligence

2023 can be described as one of the most eventful years for AI in recent times. It was full of action with the advent of large language models, generative capabilities, and the ability to engage seamlessly with users and offer unique, engaging, and exciting interactions. These developments continue to democratize AI and hold immense promise to make AI work for the benefit of humanity.

The year culminated with the arrival of multi-modal Generative AI models like GPT-4 Turbo/Vision and Gemini, with advanced capabilities to process text, image, audio, and video simultaneously and provide unified insights similar to the way humans perceive, understand, and process such information. This development has further inspired confidence to embark on research and development of Agentic AI systems.

What are Agentic AI systems?

Agentic AI systems are advanced artificial intelligence systems capable of pursuing complex goals with limited direct supervision by human beings. They are either an individual AI agent or a collection of AI agents working together in an autonomous mode in a clearly carved-out environment with definite scope to achieve set goals and objectives. They differ from traditional AI systems in multiple ways and following are some of their unique capabilities:

• Purposeful, goal-driven, initiative-taking qualities
  Agentic AI systems are not tactical and do not just react to input or instructions. They are strategic and have their own goals, take initiative, and actively pursue them to achieve desired outcomes.
• Learning and Adaptability
  Within a constrained environment and defined parameters, these systems can adapt to changing circumstances and make decisions on their own. They do not require constant human intervention and can learn from experience and adapt to situations to improve their performance.
• Complex reasoning, decisioning, and planning
  Agentic AI uses various algorithms that employ advanced reasoning capabilities, consider potential options/outcomes, and choose the best course of action to achieve specific goals.

How would an Agentic AI system manifest in the real-world?

The following is an example of how an agentic AI system would work in a customer service business function for a manufacturer of personal computing devices.

Perception and Understanding

• Process internal data
  The system would analyze chat, email, and phone interactions with customers regarding PCs, laptops, tablets, monitors, servers, and other products. It could access customer purchase history, warranty information, technical specifications, and other structured and unstructured knowledge repositories.
• Augment with external data
  The system would also monitor social media channels and online forums to understand customer sentiment, identify common issues, and track trends in technology preferences.
• Personalize
  Based on customer profile and their interactions, it would identify the customer’s needs, preferences, and expectations.

Goal Setting and Planning

The system can work with business goals/KPIs, monitor them, and constantly endeavor to improve. The following are some of the KPIs that can be tracked for agentic AI systems as they collaborate with human agents to improve overall performance:

• First Contact Resolution (FCR) rate
  Measure the percentage of customer issues resolved during the first interaction with the agentic AI, indicating its ability to handle routine and complex inquiries effectively.
• Average Handle Time (AHT)
  Track the average time it takes to resolve customer issues, highlighting the efficiency gains achieved through AI automation.
• Customer Effort Score (CES)
  Measure the effort customers must exert to get their issues resolved and the effectiveness of the AI system’s suggestions to make a task easy for customers.
• Agent Productivity
  Evaluate the impact of the agentic AI on agent workload, measuring the increase in time available for handling complex cases or proactive customer outreach.
• Customer Satisfaction (CSAT) score
  Monitor customer feedback and surveys to understand how satisfied customers are with their interactions with the agentic AI.
• Net Promoter Score (NPS)
  Track customer loyalty and willingness to recommend a brand, potentially boosted by positive experiences with the agentic AI’s helpful and efficient service.

Such a system enables:

Automated Technical Support & Troubleshooting

• Resolving technical issues quickly
  This could involve diagnosing hardware or software problems, guiding customers through troubleshooting steps, recommending drivers or firmware updates, and assisting with warranty claims.
• Providing accurate and up-to-date technical information
  The system could draw from the organization’s extensive knowledge base and technical documentation to answer customer questions about specifications, features, troubleshooting procedures, and warranty policies.
• Resolving issues autonomously
  The system could handle tasks like resetting passwords, updating billing information, and tracking order status without requiring human intervention. It could also engage multiple agents when the issues need multi-system interactions to understand the problem and resolve it.
• Reducing agent workload
  By handling routine inquiries, scheduling service appointments, and providing self-service resources, the system could free up human agents to focus on solving complex technical problems and building customer relationships.

Decision Making & Action Selection

• Guiding customers through complex processes
  When purchasing new products, setting up software, or configuring systems, the system could offer step-by-step instructions, video tutorials, and interactive co-pilots.
• Identifying and escalating complex cases
  Based on customer queries and system analysis, the system could seamlessly transfer customers to human agents for issues requiring deeper technical expertise or personalized support.

Proactive Service & Engagement

• Proactive support and recommendations
  The system could analyze customers’ usage patterns and predict potential issues, proactively notifying customers and suggesting preventive measures. It could also recommend compatible accessories, upgrades, or new products based on individual needs.
• Personalized and empathetic service
  The system could understand the emotional state of customers based on their current or recent interactions, review their purchase history, tailor communication style to their preferences, and reach out to them through their preferred communication channels at the right time. It could also recognize frustration and offer words of encouragement or apologies when appropriate.

Learning and Adaptation

• Refine responses and recommendations
  The system could learn from customer feedback, past interactions, and successful resolutions to continuously improve its accuracy and relevance.
• Adapt to changing technologies and products
  The system could regularly update its knowledge base with new product specifications, software releases, and technical documentation.
• Personalize interactions and communication over time
  The system could analyze customer preferences and communication patterns over time to adapt its tone, language, and style for a more natural and engaging experience.

Human Empowerment

• Human-AI collaboration and shared goals
  The agentic AI would not replace human agents but rather work alongside and augment them, acting as a powerful tool to enhance their efficiency, improve customer satisfaction, and ultimately build lasting customer loyalty.
• Human experts remain crucial
  The organization’s engineers and support specialists would still play a pivotal role in overseeing the system’s performance, providing training and feedback, and handling complex technical issues or customer escalations – consistently ensuring positive outcomes.

By implementing such an agentic AI system, the company could revolutionize its customer service landscape, offering faster resolutions, personalized support, and a seamless experience across all touchpoints. This would not only enhance customer satisfaction but also allow the company to focus its human capital on higher-value interactions and strategic initiatives, thereby driving overall business growth and success.

Will Agentic AI be the stepping stone to Artificial General Intelligence (AGI)?

Agentic AI, with its purposeful, goal-driven, and adaptable nature, is seen by many experts as a stepping stone towards achieving AGI, the ultimate form of artificial intelligence. While Agentic AI represents a significant step towards more advanced and autonomous AI systems, AGI remains a future possibility with potentially transformative implications. With more research and development and experiences from real-world use cases, such systems can gradually metamorphose to develop human-equivalent AGI capabilities like faster domain adaptability, advanced reasoning, problem-solving, creativity, innovation, self-awareness, and reflection.

As agentic AI becomes more prevalent, it will inevitably impact the way we work, live, and interact with technology. While such systems offer a lot of promise, they raise ethical concerns around autonomy, inherent biases, and transparency. They have the potential to revolutionize various aspects of our lives, but it is crucial to approach them with a thoughtful and balanced perspective, ensuring their advancement benefits humanity. Careful development and oversight are crucial to ensure these systems act responsibly and ethically. Guardrails need to be set so that they work under set parameters and constraints, their decisions are auditable, and they process human feedback and make needed adjustments/alignments as required.

Agentic AI is an ambitious endeavor to create next-generation autonomous systems and its success is likely to pave the way for AGI soon!