Harnessing AI to Revolutionize IT Service Management: Insights from ManageEngine's Director of AI Research

As artificial intelligence (AI) continues to transform various aspects of business operations, its potential to revolutionize IT service management (ITSM) is becoming increasingly evident. Ram Ramamoorthy, Director of AI Research at ManageEngine, shared valuable insights on this topic during his keynote and then in a one-on-one interview at the recent ManageEngine User Conference.

The Future of ITSM Applications

When asked about the future of ITSM applications, Ramamoorthy emphasized that AI will become deeply interlaced with enterprise software, much like search functionality. "I see AI as not a standalone product. I see it deeply interlaced with enterprise, consumer, and software tools that you use. More so in the enterprise, because you have a ton of observation and tracking data that you are able to analyze and make predictions with it," he explained. 

Ramamoorthy predicts AI will enable a shift from process automation to decision automation, where AI augments and accelerates decision-making. However, he believes that human-augmented decisions will remain the norm for the next few years before fully autonomous decision-making becomes prevalent.

Measuring and Ensuring Digital Maturity

Digital maturity is a moving target, and organizations need to track key performance indicators (KPIs) to ensure they are making progress. According to Ramamoorthy, better digitization leads to greater AI impact. He recommends moving under a single system, leveraging machine learning to sift through data, and focusing on the consumer experience.

Ramamoorthy emphasizes the importance of process streamlining, data streamlining, and automation in achieving digital maturity. Process streamlining involves capturing and digitizing processes across the organization, enabling seamless integration between departments and systems. This is particularly crucial for IT teams, as they need to adapt to, and leverage, the increasing agility of infrastructure and application teams.

Data streamlining is essential to make data accessible and usable for AI applications. Ramamoorthy suggests creating an analytics platform and a data lake to centralize data from various sources. This enables organizations to break down data silos and gain a holistic view of their operations. Additionally, leveraging low-code extensibility allows for easier integration of data from different systems, even when native integrations are not available.

Automation is another key aspect of digital maturity. By automating repetitive tasks and processes, organizations can free up human resources to focus on higher-value activities. Ramamoorthy recommends investing in AI-powered automation tools that can learn from historical data and improve decision-making over time.

Overcoming Data Silos and Selecting the Right Models

Data silos are a common challenge hindering AI initiatives. To overcome this, Ramamoorthy advises organizations to use low-code extensibility to integrate and aggregate more data from more sources. This approach enables organizations to connect disparate systems and create a unified data platform to get the most value from their data.

When it comes to selecting AI models, organizations must strike a balance between pre-built and custom models. Pre-built models can provide a quick start and are suitable for general use cases, such as natural language processing (NLP) and search functionality. However, for more specific and complex use cases, custom models may be necessary to address unique business requirements.

Ramamoorthy emphasizes that customization can be expensive, and organizations should carefully evaluate the costs and benefits before embarking on custom model development. He also stresses the importance of integrating AI with existing IT service providers, rather than treating it as a standalone solution. This approach ensures that AI is seamlessly embedded into the organization's IT ecosystem, enabling better decision-making and improved service delivery.

To further streamline AI adoption, Ramamoorthy recommends collaborating closely with IT service providers who have deep expertise in AI and can guide organizations through the process of model selection, deployment, and integration. By leveraging the knowledge and experience of these providers, organizations can accelerate their AI initiatives and avoid common pitfalls.

Avoiding Pitfalls and Capturing Incident Learnings

Implementing AI in IT operations comes with its own set of challenges. Ramamoorthy cautions against setting moonshot expectations and viewing AI as a silver bullet. Begin with small projects that can be successful. Doing so enables companies to train employees on how to use AI effectively, and keep decisions in check while gaining practical experience.

Capturing and sharing learnings from incidents is crucial. This promotes collaboration which is critical for the success of AI initiatives.  Ramamoorthy suggests using service desk tools, annotation, and internal social platforms like Zoho Connect to facilitate knowledge sharing across the organization.

Evolving Roles and Quick Win Use Cases

As AI becomes more prevalent in IT operations, the role of IT teams will evolve. Ramamoorthy advises professionals to keep realistic expectations and develop skills in human-augmented decision-making. He also highlights the importance of understanding confidence intervals and leveraging autonomous decision-making for high-confidence scenarios.

For organizations starting their AI journey, Ramamoorthy recommends focusing on natural language processing (NLP) and human-generated language as “quick win” use cases. Replacing traditional search functionality with AI-powered search can deliver immediate benefits without extensive training. This delivers huge benefits to employees and customers alike. 

Ensuring Trustworthy and Unbiased AI

As AI becomes increasingly integrated into IT service management and decision-making processes, ensuring its trustworthiness, explainability, and fairness is paramount. Ramamoorthy emphasizes the importance of treating data like code to maintain the integrity and reliability of AI systems.

Treating data like code involves implementing robust access controls and security measures to protect sensitive information. This includes enforcing strict authentication protocols, such as two-factor authentication, to prevent unauthorized access to data. By treating data with the same level of security as code, organizations can minimize the risk of data breaches and ensure that only authorized personnel can access and manipulate the data used to train AI models.

Access controls also play a crucial role in maintaining data privacy and compliance with regulations such as GDPR and HIPAA. By implementing granular access controls, organizations can ensure that sensitive data is only accessible to those who need it for specific purposes, reducing the risk of data misuse or leakage.

In addition to access controls, Ramamoorthy stresses the importance of conducting bias checks to ensure that AI models are fair and unbiased. Bias can creep into AI systems through various means, such as biased training data, flawed algorithms, or human prejudices. If left unchecked, biased AI can lead to discriminatory decisions and perpetuate societal inequalities.

To mitigate bias, organizations should regularly audit their AI models and training data for potential biases. This involves analyzing the data for underrepresentation or overrepresentation of certain groups, as well as testing the model's outputs for fairness and consistency across different demographics. Bias checks should be conducted at various stages of the AI lifecycle, from data collection and preprocessing to model training and deployment.

Ramamoorthy specifically highlights the importance of bias checks in sensitive areas like service desk ticket prioritization. In this context, biased AI could lead to unfair treatment of certain users or customers based on factors such as their name, location, or communication style. By conducting regular bias checks and ensuring that ticket prioritization algorithms are fair and objective, organizations can provide equitable service to all users and maintain trust in their IT service management processes.

Explainability is another crucial aspect of trustworthy AI. Explainable AI refers to the ability to understand and interpret how an AI model arrives at its decisions. This is particularly important in high-stakes scenarios where AI-driven decisions can have significant consequences, such as in IT security or incident response.

To ensure explainability, organizations should prioritize the use of interpretable AI models, such as decision trees or rule-based systems, over black-box models like deep neural networks. Interpretable models allow IT teams to trace the logic behind AI-driven decisions and identify potential errors or biases. In cases where complex models are necessary, organizations should invest in tools and techniques for explainable AI, such as feature importance analysis or counterfactual explanations.

By prioritizing trustworthiness, fairness, and explainability in their AI initiatives, organizations can build confidence in their AI-driven IT service management processes. This trust is essential for driving adoption and realizing the full potential of AI in improving service delivery, reducing costs, and enhancing user experiences.

Moreover, trustworthy and unbiased AI is not only an ethical imperative but also a legal and regulatory requirement in many jurisdictions. As AI governance frameworks and regulations evolve, organizations that prioritize responsible AI practices will be better positioned to comply with these requirements and maintain public trust in their services.

Treating data like code, implementing robust access controls, conducting regular bias checks, and ensuring explainability are essential components of building trustworthy and unbiased AI systems. By prioritizing these practices, organizations can harness the power of AI to transform their IT service management processes while ensuring fairness, accountability, and transparency.

The Future of AI in IT Operations

Looking ahead, Ramamoorthy is excited about the potential of combining different AI models to solve complex problems. For example, using computer vision to read PDFs, understanding user roles and requirements, and automating decision-making based on that context.

To future-proof AI investments, Ramamoorthy advises avoiding standalone AI providers and partnering with vendors who have survived previous technology waves. Building a strong foundation and being resilient to change is key to adapting as new technologies emerge.

Empowering Through Education: The Zoho Schools Program

Beyond AI, Ramamoorthy also shared insights on Zoho's inspiring initiative to empower individuals through education. The Zoho Schools program, running for 15 years, provides high school students with a tailored two-year training program, including hands-on experience with product teams. Remarkably, 15% of Zoho's workforce consists of graduates from this program.

By focusing on skill development, job readiness, and empowering local talent in smaller towns and villages, Zoho is making a significant impact on the lives of individuals who may not have had access to traditional educational opportunities.

As organizations navigate the AI landscape and strive for digital maturity, the insights shared by Ram Ramamoorthy provide valuable guidance. By leveraging AI to enhance ITSM, streamlining data and processes, selecting the right models, and fostering a culture of continuous learning, businesses can harness the power of AI to drive innovation and deliver exceptional IT services.