Is your data strategy ready for generative AI?

Already, the machines have learned so much. Artificial intelligence (AI) systems can scan X-rays and biopsy slides to alert pathologists to anomalies that require focus, create realistic training simulations and market reports, monitor financial transactions to flag potential fraud, personalize online content, and answer customer service questions.

None of this is possible without data—high-quality, trustworthy, and accessible unstructured data to train the models and for models to process. Advanced applications, especially those employing generative AI, rely increasingly on unstructured data from images, texts, video, and audio.

IT and data professionals know all too well the allure of generative AI. It is capturing leaders' attention in business, healthcare, education, finance, law, energy, and entertainment, leaving organizations scrambling to leverage and control it. These same professionals are also well-versed in their organizations' data challenges. According to a survey of 700 IT and data decision-makers sponsored by Iron Mountain, key data-related challenges include crafting and gaining agreement to a generative AI strategy; sourcing, protecting, and preparing data for generative AI training; and protecting and managing data and other assets created by generative AI.

Crafting and gaining agreement on an AI strategy

Thirty-four percent of IT and data decision-makers say they have challenges crafting and gaining agreement on a generative AI strategy that aligns with organization goals. For generative AI applications to live up to their potential, business leaders must work together with IT and data leaders on an AI strategy that is technically feasible and grounded in business goals. Many IT and data decision-makers think focused AI leadership can facilitate this collaboration.

Because AI strategies and data strategies are intertwined and interdependent, a data strategy must be an integral part of the AI strategy. A data strategy provides the necessary framework and resources for AI, while an AI strategy leverages data to generate insights and drive decisions. To develop a successful AI strategy, teams need to build a shared understanding of generative AI capabilities and constraints through cross-functional workshops, training programs, and continuous communication.

A data strategy provides the framework for protecting and managing physical and digital data across their lifecycle. It ensures high-quality, trusted, accessible data while addressing ethical considerations and regulatory compliance. As data privacy laws and intellectual property regulations evolve, they must be integrated into the data strategy.

Protecting and managing data from physical and digital assets

As one McKinsey Digital article said, “If your data isn’t ready for generative AI, your business isn’t ready for generative AI.” Establishing a solid data foundation is critical for generative AI, which requires large volumes of high-quality unstructured data for training and validation. In the survey, 36% of IT and data decision-makers cited this as a challenge. Protecting and managing data requires a robust governance, security, and compliance focus. The processes support the availability of relevant and trustworthy data for unbiased AI systems. Meanwhile, effective data architectures support the scalable and efficient storage and processing of growing volumes of unstructured data and generative AI systems.

Once identified, relevant data should be consolidated, cataloged, and prepared for use with AI. Any physical assets must be digitized—by scanning documents, capturing high-resolution images, or converting audio recordings into digital formats. Digital assets must be collected from databases, social media platforms, and online repositories.

To meet the quality and accessibility requirements of AI models, organizations must clean, normalize, validate, and annotate the data with metadata and labels. They also must index the data and ensure it is properly organized. Techniques like AI/ML (machine learning), natural language processing, computer vision, and audio analysis are often applied to extract meaningful information from unstructured data.

Throughout the data lifecycle, organizations must enforce rigorous data protection measures and comply with evolving data privacy laws and intellectual property regulations.

Managing data and other assets created by generative AI

Similar and distinct considerations apply when managing data and other assets produced by generative AI compared to traditionally created data. While traditional data management processes generally apply, organizations must also address risks to uphold ethical use, maintain security and integrity, and manage intellectual property concerns related to AI-generated assets.

Ensure ethical use

Promoting the ethical use of generative AI reduces potential negative impacts and helps maintain trust in AI systems and the data that trains them. This approach involves addressing biases, managing content responsibly, and navigating ethical dilemmas.

• Generative AI can perpetuate or amplify existing biases present in training data. Regular audits and updates of training datasets are necessary to minimize bias and promote fairness in AI models.
• Generative AI can produce misleading, harmful, or inappropriate content. Organizations should develop and enforce strict content guidelines, implement content filtering systems, and continuously monitor AI outputs.

Maintain the security and integrity of generated content

Maintaining the security and integrity of AI-generated content is vital to protecting personal information, ensuring trust in digital outputs, and complying with evolving regulations. Achieving this objective requires robust security measures, transparent verification processes, and ongoing regulatory awareness.

• Generative AI can inadvertently expose or misuse personal information. Implementing data anonymization techniques and complying with data privacy regulations are crucial.
• AI systems and data are targets for cyberattacks. Encryption and regular security audits help protect AI systems and data.
• AI-generated content can blur the lines between what is real and what is fake, undermining trust in digital content. Organizations must verify and authenticate AI-generated data and content and establish guidelines for appropriate use.
• Evolving regulations do not fully address AI-generated data concerns. Organizations can anticipate regulatory developments by staying informed and working proactively with regulators.

Protect intellectual property

Creating and using AI-generated content can raise complex intellectual property questions about ownership and originality. Establishing clear policies on intellectual property (IP) rights for AI-generated content and consulting legal experts can help navigate IP laws.

An agile data strategy that is AI-ready

Crafting and implementing a data strategy for generative AI is not a one-time effort. It is a continuous process that evolves as technology advances and business needs change.

Organizations with a comprehensive and nimble data strategy will be better poised to leverage the transformative powers of AI. Those who keep close collaboration among IT, data, and business experts—following the essentials of data strategy and working with trusted technology partners—will build a strong foundation for all that is to come.

Read the research report, Capitalizing on generative artificial intelligence, to see what 700 IT and data leaders reveal about top generative AI uses and barriers to success. Learn how other organizations accelerate innovation while minimizing risk with a unified asset strategy.

Explore solutions such as Iron Mountain Insight® DXP that can help measure productivity gains and demonstrate tangible returns on your data.