There’s lots of buzz about Generative AI (GenAI), artificial intelligence that uses machine learning models to analyze large amounts of data and learn patterns to generate new content. In this article, we examine how GenAI might appropriately support investigators and what concerns we should have about using GenAI in workplace investigations.
GenAI’s primary benefit to investigators is efficiency. However, for investigators contemplating using GenAI, confidentiality, compliance, and avoiding the intrusion of bias and false information are key concerns. Further, because GenAI is challenged in understanding (and may misinterpret) nuance and context, investigators must take care to use GenAI judiciously only as a support tool, not a decision-maker.
This article does not attempt to provide comprehensive guidance for the appropriate use of GenAI in investigations or to suggest the use of specific GenAI products; rather it is intended to surface some of the key issues that should be top of mind as workplace investigators, and especially attorney workplace investigators, adjust to GenAI’s inevitable role in workplace investigations.
What concerns should attorney workplace investigators consider before using GenAI to streamline investigations?
As anyone who has experimented with GenAI knows, it is incredibly efficient. It can be effective at reasonably accurate transcription, synthesizing volumes of data and other information, and identifying relevant patterns at impressive speed, far faster than humans. Many investigators will be tempted to load reams of interview notes and other information into a GenAI product so it can spit out interview summaries and chronologies, identify inconsistencies and patterns, and suggest follow up questions. But before taking the leap of using GenAI to streamline investigations, there are important issues to consider.
Attorney workplace investigators owe a duty of confidentiality to their clients and cannot provide confidential information into a GenAI product without ensuring it has adequate confidentiality and security protections. The California State Bar’s practical guidance for the use of generative AI advises attorneys to anonymize any client information input into a GenAI solution that lacks adequate confidentiality and security protections. The State Bar further advises consulting with experts before inputting confidential client information into any GenAI product to ensure adherence to security, confidentiality, and data retention and protection protocols and regulations. Attorneys also should ensure that confidential information is not used to train GenAI products or shared with third parties.
Attorney workplace investigators also have duties of competence and diligence. Clients engage investigators to use our professional judgment and critical thinking in conducting thorough and impartial investigations. Because GenAI does not fully understand and may misinterpret nuance and context, we must take care not to inappropriately delegate reasoning and reflection to GenAI. For example, an investigator who relies on a GenAI product to synthesize information contained in interviews without employing their own critical analysis may find themselves unable to defend why they failed to follow up on evidence GenAI overlooked because GenAI did not appreciate the relevance of that evidence. In fact, an investigator’s over-reliance on GenAI products in service of efficiency will not serve the needs of clients who engage attorney workplace investigators with the expectation that an investigator’s process will withstand scrutiny at deposition or in court. For this reason, the State Bar advises using GenAI-generated outputs only as a starting point that attorneys should critically scrutinize and suggests attorneys disclose to clients their intent to use GenAI (along with the benefits and risks of such use). While delegating analysis of evidence to GenAI may be an appealing shortcut, it risks missed steps (and missteps) in the critical reasoning and professional judgment that results in a thorough and effective investigation process.
Further, if we do not have visibility into the data GenAI products are trained on, we may not know whether a GenAI product has baked-in biases that may impact its analysis, whether it is relying on false information, or whether it is engaging in hallucination (perceiving patterns that do not exist or providing incorrect information). For these reasons, the California State Bar advises attorneys to obtain a reasonable understanding of how GenAI technology works and its limitations before using any GenAI product.
While GenAI is a promising developing tool to streamline time-consuming data-intensive investigatory tasks when used appropriately, it is not a substitute for a skilled investigator’s careful gathering of evidence and thoughtful analytical reasoning.
Why shouldn’t investigators use GenAI as a decision-maker or to reach findings?
While there may be room for debate over the boundaries of GenAI use in the investigative process, there is little debate (at least currently) that humans should not rely on GenAI in decision-making or reaching findings. Just as an investigator must be able to explain why they did not follow up on potential evidence, they also must be able to explain how and why they reached their ultimate findings. Experienced investigators know that a deep and nuanced understanding of relevant evidence is necessary to reach well-reasoned findings. Because nuance and context typically are crucial components of well-reasoned and well-articulated factual findings, GenAI is not equipped to replace human decision-makers or fact finders. Especially for sensitive or complex investigations, clients engage skilled investigators to analyze the evidence gathered and provide sophisticated findings that deepen the client’s understanding of what more likely than not occurred, often to enable the client to determine appropriate next steps. Findings that do not demonstrate the investigator’s careful analysis and sound judgment will not earn the confidence of participants in the investigation, client stakeholders, or ultimate deciders in potential future court proceedings.
Generative AI undoubtedly has a growing role in investigations. However, investigators should employ GenAI products with care and healthy skepticism so as not to undermine confidence in the skill, impartiality, and reliability of the investigator’s process and findings. Attorney investigators also must comply with expectations and rules for the use of GenAI in the practice of law, such as the practical guidance articulated by the California State Bar.
Since 2014, our clients have entrusted Hulst & Handler LLP to conduct hundreds of thorough and impartial investigations in workplaces, schools, and universities.
