The history of artificial intelligence (AI) dates back to the 1950s and has evolved significantly. Today, AI plays a prominent role for both individuals and organizations. In 2024, corporate investment in AI reached $252.3 billion, a 13-fold increase from 2014. Demand is high, and everyone is hungry for revenue growth and cost reduction. $252.3 billion $252.3 billion Most recently, Generative AI has created a tidal wave across industries. Based on a survey by Google Cloud, 74% of respondents reported ROI on GenAI investments. Yet, the evolution didn’t stop there, with the next leap being the rise of AI Agents. Still in its infancy, the capabilities of AI Agents have yet to be fully explored. So, is this technology as promising as it seems? a survey by Google Cloud, a survey by Google Cloud, Generative AI vs. Agentic AI: What’s the Difference? Generative AI vs. Agentic AI: What’s the Difference? Generative AI operates by receiving data and following instructions to produce outputs. On the other hand, AI Agents are designed to learn and act autonomously to achieve predefined goals. For instance, a banker might provide a GenAI with data and ask it to generate a report. The AI produces the report accordingly, but can’t act independently and still requires a human to review it. AI Agents go a step further; rather than just following rules, they can interpret each case and make suggestions or even make decisions without human intervention within a defined regulatory and risk-control framework. Accordingly, it is evident that technology has advanced from passive content generation to autonomous agentic solutions, creating new opportunities for investment returns. How can this next-generation intelligence impact industries? How can this next-generation intelligence impact industries? Here are some tasks AI Agents can handle when properly trained: Customer Service: examples include managing accounts, processing loan applications, and resolving customer issues. Back-end operations: handling complex tasks that require human intervention, such as optimizing trading strategies and managing payments. Financial guidance: tailoring portfolios by providing personalized financial advice. Interestingly, according to a survey by Forrester, 70% of respondents anticipate using Agentic AI to deliver tailored financial advice that was solely available to high-net-worth individuals, which would ultimately disrupt traditional models of exclusivity. Fraud detection and prevention: learning and identifying different types of fraud patterns, flagging anomalies, freezing suspicious transactions. Regulatory monitoring: scanning for policy updates and adjusting processes accordingly. Credit scoring: speeding up the rate of creditworthiness assessments and providing opportunities for borrowers with “thin files” by looking at other data sources beyond traditional credit reports. Customer Service: examples include managing accounts, processing loan applications, and resolving customer issues. Customer Service Back-end operations: handling complex tasks that require human intervention, such as optimizing trading strategies and managing payments. Back-end operations Financial guidance: tailoring portfolios by providing personalized financial advice. Interestingly, according to a survey by Forrester, 70% of respondents anticipate using Agentic AI to deliver tailored financial advice that was solely available to high-net-worth individuals, which would ultimately disrupt traditional models of exclusivity. Financial guidance survey by Forrester survey by Forrester Fraud detection and prevention: learning and identifying different types of fraud patterns, flagging anomalies, freezing suspicious transactions. Fraud detection and prevention Regulatory monitoring: scanning for policy updates and adjusting processes accordingly. Regulatory monitoring Credit scoring: speeding up the rate of creditworthiness assessments and providing opportunities for borrowers with “thin files” by looking at other data sources beyond traditional credit reports. Credit scoring Real World Applications: Early Signs of Adoption Real World Applications: Early Signs of Adoption Some companies are leading in this technological movement through early adoption. For example, BNY has its own enterprise AI platform named Eliza, which offers multiple AI models from leading providers for BNY employees. “Digital workers” at BNY find new business leads, write code, handle payment processes and client onboarding, and handle reconciliations. Currently, BNY reports having over 100 digital employees. Moreover, JPMorgan Chase’s agentic deployment demonstrates its capabilities empirically. They introduced LAW, which consists of multiple specialized agents in the legal domain that respond to complex legal queries. The study’s empirical benchmark consists of a dataset of 720 queries. Accordingly, LAW excelled in complex tasks compared to the baseline, which is GPT-3.5-turbo (GenAI). For instance, in calculating contract termination dates, LAW performed 92.9% better than the baseline. (GenAI) Investor Checklist: Key Considerations Investor Checklist: Key Considerations Indeed, we anticipate investment growth for companies that successfully implement AI Agents. However, there are several challenges that investors need to consider when assessing companies’ AI approaches: The proper deployment of AI Agents In general, across different AI initiatives, extracting value from these models remains a challenging process. An IBM survey of 2,000 CEOs found that only 25% of AI initiatives delivered the expected ROI over the past three years, and just 16% scaled at the enterprise level. So, for companies, having a budget isn’t enough. Moreover, history shows that technological potential alone isn’t enough and that it requires proper deployment. According to an MIT study, this was evident with GenAI implementations, which were often fragmented or poorly launched. Based on executive interviews, this study found that 95% of organizations with GenAI models are getting zero return, while a small portion is “extracting millions in value.” Although this study has its limitations, as it only measured ROI six months post-pilot, it highlights the issue: misapplication rather than technological failure. “Agentic washing” This occurs when companies or vendors claim their AI systems have agentic qualities, but in reality, they do not. After analyzing thousands of “supposedly” Agentic AI vendors, Gartner analysts report that only around 130 products exhibit agentic traits. Accordingly, Gartner projects that over 40% of agentic projects will be canceled by the end of 2027 due to factors such as “unclear business value, inadequate risk controls, or escalating costs.” Therefore, as investors, it is important to look beyond marketing claims. First movers gaining an edge IBM research highlights a clear performance gap between AI-first organizations and those with gradual implementations (see Figure 1). The proper deployment of AI Agents In general, across different AI initiatives, extracting value from these models remains a challenging process. An IBM survey of 2,000 CEOs found that only 25% of AI initiatives delivered the expected ROI over the past three years, and just 16% scaled at the enterprise level. So, for companies, having a budget isn’t enough. Moreover, history shows that technological potential alone isn’t enough and that it requires proper deployment. According to an MIT study, this was evident with GenAI implementations, which were often fragmented or poorly launched. Based on executive interviews, this study found that 95% of organizations with GenAI models are getting zero return, while a small portion is “extracting millions in value.” Although this study has its limitations, as it only measured ROI six months post-pilot, it highlights the issue: misapplication rather than technological failure. The proper deployment of AI Agents The proper deployment of AI Agents The proper deployment of AI Agents In general, across different AI initiatives, extracting value from these models remains a challenging process. An IBM survey of 2,000 CEOs found that only 25% of AI initiatives delivered the expected ROI over the past three years, and just 16% scaled at the enterprise level. So, for companies, having a budget isn’t enough. IBM survey So, for companies, having a budget isn’t enough. Moreover, history shows that technological potential alone isn’t enough and that it requires proper deployment. According to an MIT study, this was evident with GenAI implementations, which were often fragmented or poorly launched. Based on executive interviews, this study found that 95% of organizations with GenAI models are getting zero return, while a small portion is “extracting millions in value.” Although this study has its limitations, as it only measured ROI six months post-pilot, it highlights the issue: misapplication rather than technological failure. MIT study “Agentic washing” This occurs when companies or vendors claim their AI systems have agentic qualities, but in reality, they do not. After analyzing thousands of “supposedly” Agentic AI vendors, Gartner analysts report that only around 130 products exhibit agentic traits. Accordingly, Gartner projects that over 40% of agentic projects will be canceled by the end of 2027 due to factors such as “unclear business value, inadequate risk controls, or escalating costs.” Therefore, as investors, it is important to look beyond marketing claims. “Agentic washing” “Agentic washing” “Agentic washing” This occurs when companies or vendors claim their AI systems have agentic qualities, but in reality, they do not. After analyzing thousands of “supposedly” Agentic AI vendors, Gartner analysts report that only around 130 products exhibit agentic traits. Accordingly, Gartner projects that over 40% of agentic projects will be canceled by the end of 2027 due to factors such as “unclear business value, inadequate risk controls, or escalating costs.” Therefore, as investors, it is important to look beyond marketing claims. Gartner analysts report that only around 130 products exhibit agentic traits. Gartner analysts report that only around 130 products exhibit agentic traits. First movers gaining an edge IBM research highlights a clear performance gap between AI-first organizations and those with gradual implementations (see Figure 1). First movers gaining an edge First movers gaining an edge First movers gaining an edge IBM research highlights a clear performance gap between AI-first organizations and those with gradual implementations (see Figure 1). IBM research So, AI-first organizations demonstrate improvements in revenue and operating profits compared to their other AI initiatives. They are more likely to realize measurable ROI. Google Cloud reports similar findings, reinforcing the link between early strategic commitment and realized ROI. Google Cloud Google Cloud The hype is real, but proper evaluation is real-er The hype is real, but proper evaluation is real-er The hype surrounding Agentic AI is real. Based on surveys and interviews with over 2,000 respondents, an MIT Sloan report states that 35% of companies are already using Agentic AI, and another 44% plan to adopt it soon. For us, investors, enthusiasm alone doesn’t create shareholder value. Proper evaluation builds confidence when investing in a company: Check companies’ implementation plans and their progress. Review their budget to ensure that it is sufficient. Evaluate the authenticity of the agentic capabilities being pitched. Search for empirical evidence of agentic outcomes rather than pilot-stage promises. Track the speed of adaptation to new agentic developments and innovations. Check companies’ implementation plans and their progress. Check Review their budget to ensure that it is sufficient. Review Evaluate the authenticity of the agentic capabilities being pitched. Evaluate Search for empirical evidence of agentic outcomes rather than pilot-stage promises. Search Track the speed of adaptation to new agentic developments and innovations. Track When all the criteria above are met, Agentic AI has genuine potential to deliver realized ROI and improve stock performance.
