The right data strategy can help your AI plan succeed

Generative AI (gen AI) has ushered in a wave of significant transformations across a range of business functions, including corporate legal functions. This innovative technology has the potential to transform legal departments and enhance efficiency across a spectrum of tasks, such as analyzing data, researching legal issues, summarizing documents, and comparing information. Legal organizations are attempting to embrace these new tools. However, they frequently fail to consider a key component needed for success: a robust data strategy.

Before exploring its role in powering the effective use of gen AI tools, it is critical to understand what a “data strategy” entails in a corporate legal setting. In general, a data strategy is a comprehensive plan that outlines how an organization collects, stores, manages, shares, and uses data. It is a roadmap that aligns the organization’s data initiatives with its strategic business goals. A robust data strategy addresses issues including data governance, data quality, data architecture, and data literacy, helping to ensure that enterprise data are treated as assets that can drive decision-making and innovation. Finally, a foundational component of a data strategy is a blueprint that provides tactical insights into the systems where data resides, the interconnectivity of those systems, and the data stored there, as well as the business questions/considerations that each system (and its associated data) is intended to address. In the context of a corporate legal function, this will typically influence how department resources are capturing and storing information related to legal matters, contracting, and law department knowledge.

Generative AI, a subset of artificial intelligence (AI), uses machine learning models to create, review, and analyze content including text, images, and even software code based on user input and the data it can access. As an example, in the context of corporate legal functions, gen AI can be leveraged or trained to support the following use cases:

• Review contract templates to help identify deviations in third-party paper and generate responsive positions that the legal team has pre-approved.
• Organize and manage the vast amount of knowledge within a legal department. It can help in categorizing, searching, and retrieving information quickly and efficiently.
• Analyze historical case history to generate strategic insights, aiding in the creation of tailored legal content to enhance a company’s negotiations and litigation strategy and results.
• Streamline the process of reviewing legal invoices. It can analyze line items, compare them against agreed-upon rates and rules, identify billing errors or discrepancies, and help ensure compliance with the company’s billing guidelines.

However, the quality and accuracy of the output heavily relies on the quality and depth of the underlying data available to the AI tool. Without a well-defined data strategy, in-house legal teams may face numerous challenges in implementing gen AI, which can limit the effectiveness of the tool. These include:

• Inaccurate outputs: If the data used to train the gen AI models are not accurate or comprehensive, the large language models underlying gen AI may generate faulty, incomplete, or misleading outputs. Unless caught and corrected, such outputs may lead to suboptimal decisions and increased risk for the organization. In addition to significant reputational and  other implications for the legal department and individual attorneys, low-quality gen AI outputs can also create other major risks—from financial, regulatory, litigation, public relations, and other perspectives—for the business as a whole.
• Inefficient or slow models: Without a well-organized data architecture, it may be difficult for gen AI models to access and use data effectively. This could lead to inefficiencies in the use of gen AI and a decrease in the overall productivity of the legal function.
• Compliance Issues: If the data used to train the gen AI models are not compliant with relevant data protection regulations (including but not limited to HIPPA, GDPR COPPA), the organization could face legal and reputational risks. Similarly, using improperly obtained and unlicensed data to train large language models (LLMs) can create serious intellectual property implications.

A robust data strategy can serve as the foundation for successful Gen AI integration. This strategy, tailored to harmonize with the company’s overarching business objectives and goals of the enterprise and its corporate legal function, helps serves as the North Star guiding all initiatives. At its core, the development of a data strategy for Gen AI implementation, typically a collaborative effort between legal operations and technology resources, should empower corporate legal departments to evolve into more effective business partners. Developing a data strategy usually involves these key steps:

• Educate internal legal stakeholders on the use of gen AI and the significance of data as part of the organization’s broader gen AI strategy. Identify early adopters and program evangelists to help socialize the power of data and the role individuals play in ensuring that the right kind of data is being used. Many companies are organizing immersive workshops intended to orient internal stakeholders to gen AI broadly as well as the organization’s priorities in the context of gen AI, and the critical role of enterprise data. In-house teams should be viewed as crucial constituents.
• Envision gen AI use cases: Begin by identifying specific gen AI use cases within the legal function and envisioning how gen AI can enhance these processes. This initial step guides the data strategy by determining the types of data that can be most beneficial for the gen AI models. Some law departments are electing to start with contracting and/or self-service knowledge management.

• Data discovery and collection: Identify and gather the types of data that will be useful for the gen AI models. For corporate legal functions, this could include data from legal documents, contracts, case files, and internal databases. The data collected should be accurate, relevant, and comprehensive to help ensure that the gen AI models can generate reliable outputs. Most importantly, be realistic both with the legal team and with broader organizational stakeholders as to the overall quality reliability of the available data. To get started, organizations should work to define clears roles and responsibilities around data discovery alongside a methodical approach to collection.
• Data organization and architecture: Following data collection, it is crucial to organize the data in a way that makes it easily accessible and usable for the gen AI tools being accessed. This involves creating a data architecture that can manage large volumes of data and allows for easy retrieval and updating of data. To the extent your data requires any amount of “scrubbing” and/or clean-up, this is the time to invest in that effort.

• Data analysis and assessment: The collected and organized data should then be analyzed to identify patterns, trends, and insights that can be used to train the gen AI models. The analysis should be conducted using advanced analytical tools and techniques, helping to ensure that a consistent and formal methodology is used to drive meaningful insights across the following:
  
  • Data availability and location: Do you have the data that you think you have and, if so, where is it being stored?
  • Data quality: How complete is the data that you have available? Are key elements missing that could lead to gaps and/or inaccurate outputs from your gen AI tool?
  • Data classification, naming conventions, etc.: Do the naming conventions and/or classifications currently in place support the organization’s requirements as it relates to defined use cases?

• Data management: Implement quality assurance measures such as regular data audits and data “cleaning” techniques, along with effective data management practices, like frequent updates; the removal of outdated, irrelevant, or incorrect data; and helping to ensure easy data accessibility for gen AI models. These steps can help improve the results AI generates. Corporate legal departments frequently overlook this area of responsibility even though it is critically important. In many instances, organizations are considering new roles and increased standardization to help ensure data integrity, quality, and completeness to support gen AI initiatives.
• Data governance definition: Simultaneously, as the other steps are being executed, it is crucial to institute policies and procedures that govern the use of data for gen AI models effectively. This encompasses comprehending current data privacy regulations; managing issues of attorney-client confidentiality, privilege, and attorney work product protections; determining who has access to what data, when, and why; and specifying the methods of data collection, storage, and protection. The establishment of your law department’s data governance in relation to gen AI is a critical measure in helping to safeguard the organization from potential legal and reputational risks.

A mature data strategy is a fundamental prerequisite for the successful integration of gen AI into corporate legal functions. Crafting a robust data strategy can help ensure that gen AI models have access to high-quality, relevant data, enabling them to generate reliable and useful outputs. This can not only help improve the efficiency and effectiveness of the corporate legal function, but also drive strategic decision-making and foster innovation. Investing in gen AI technologies and regularly updating the data strategy are key steps in this process. Ultimately, the integration of a mature data strategy and gen AI into corporate legal functions can significantly contribute to an organization’s overall success by leveraging technology to help stay ahead of the curve.
The journey toward integrating gen AI into your corporate legal functions is a strategic investment towards your organization’s overall success.