TLDR: The success of Generative AI (GenAI) initiatives is fundamentally dependent on the quality of the data used for training, a critical insight highlighted by cybersecurity leader Trellix. High-quality, well-structured, and accurately labeled data is essential to prevent inaccurate outputs, ensure reliable AI-supported decision-making, and drive tangible business value. Companies like Trellix have demonstrated that investing in data integrity and human-AI collaboration is key to unlocking GenAI’s full potential, leading to improved efficiency, cost savings, and enhanced employee roles.
In the rapidly evolving landscape of artificial intelligence, a consensus is emerging among industry experts: the true power and success of Generative AI (GenAI) applications are inextricably linked to the quality of their underlying data. This critical truth is underscored by the transformative experiences of companies integrating GenAI into their operations.
Cybersecurity giant Trellix, serving over 40,000 clients globally, stands as a prime example of this principle. Michael Alicea, Chief Human Resources Officer at Trellix, emphasized that the company’s journey with AI clearly demonstrates how the efficacy of generative AI hinges on the quality of the data it learns from. Trellix has strategically deployed GenAI to streamline internal processes and enhance client-facing solutions.
Internally, GenAI has revolutionized HR functions, starting with impactful implementations like chatbots. These bots efficiently answer employee queries regarding benefits and holidays, adapting responses based on country-specific calendars and employment details. This shift has not led to job displacement but rather has freed HR professionals from routine tasks, enabling them to focus on higher-value, strategic activities. Alicea noted that three individuals from the People Services team have already been promoted into strategic HR business partner roles, illustrating how AI can catalyze career growth.
Beyond HR, Trellix leverages AI in software development, automating repetitive coding tasks to accelerate the creation of cybersecurity solutions. This approach does not replace human developers but redeploys them into roles focused on design and innovation, fostering a renewed sense of purpose among staff. As Alicea highlighted, it is the synergistic combination of AI and human effort that truly drives results.
Building trust and confidence in GenAI outputs has been a crucial part of Trellix’s strategy. Their approach involved hands-on experimentation coupled with rigorous attention to data quality. Initial efforts yielded an accuracy rate of approximately 50%, but through continuous feedback and iterative refinement of data sets, accuracy has now climbed to over 90%. This demonstrates that success with GenAI is a dynamic process requiring ongoing recalibration and validation of AI performance by regularly returning to the data.
Experts concur that high-quality, well-structured, and accurately labeled data is paramount to prevent inaccurate outputs and ensure reliable AI-supported decision-making. Poor data quality can cascade through machine learning pipelines, leading to flawed business decisions and missed opportunities. The challenges often include unlabelled or poorly labeled data, inconsistencies, and inaccuracies, which compromise the reliability and trust in data, ultimately impeding the training and performance of GenAI models.
Charlie Farah, Qlik’s Chief Technology Officer for Analytics and AI, stated, ‘Trust and data quality will define the success of AI in 2025.’ He further noted that solutions allowing users to query datasets using natural language will be favored for meeting growing demands for usability and reliability. The true value of AI, according to Farah, lies in its ability to help businesses operate data responsibly, balancing innovation with control, security, and compliance. Forecasts indicate that by 2025, proprietary business data will be a core element driving advanced AI outcomes, as basic model performance alone may no longer suffice.
To ensure high-quality data for GenAI, enterprises must prioritize contextualization, comprehensiveness (ensuring data encompasses all relevant groups and scenarios), and bias mitigation through iterative dataset reviews. Data freshness is also vital, as outdated data can mislead models. Furthermore, human supervision and ‘human guardrails’ are essential to train GenAI models and mitigate biases and inaccuracies present in training data.
Also Read:
- Federal Agencies Prioritize Workforce Training to Bridge Generative AI Skills Gap
- Generative AI: Capgemini Reports Rapid Transition to Core Business Transformation Enabler
As the cost savings and efficiency gains of AI become undeniable, particularly in areas like threat detection where AI can sift through vast volumes of telemetry data faster and more accurately than humans, the case for AI becomes increasingly compelling. Trellix’s experience offers a vital lesson for all organizational leaders: GenAI’s power is only as strong as the data that fuels it. Organizations that prioritize data integrity, foster human-AI collaboration, and maintain ethical vigilance will be the ones to thrive in the next era of digital transformation.
