AI in Workplace Investigations: Cautiously Optimistic
This is the first in a two-part series discussing our Firm’s experience with generative artificial intelligence (GenAI). This two-part series will not cover how GenAI works, what tools are available on the market, or what regulating bodies, like the State Bar of California, or state legislatures are doing with respect to GenAI. Rather, the focus is our philosophical approach to exploring and effectively implementing GenAI. We know that all of you will do the research to use GenAI responsibly and with caution.
Workplace investigators have a complicated relationship with report writing. It is often the most difficult to master, the most time-consuming, and the most delay-creating factor in completing an impartial, prompt, and thorough investigation. But for many, it was our skill at writing that attracted us to this work in the first place. So, when generative artificial intelligence (GenAI) came on the scene, promising to alleviate one of our hardest—yet beloved—tasks, we had mixed feelings.
GenAI is shaping up to be a once-in-a-generation technological revolution with broad implications across our society. Every industry is grappling with its potential, its risks, its pitfalls, and how it will enhance, eliminate, or change day-to-day duties.
We must do the same thing in our niche practice of workplace investigations. As investigators, we bear a tremendous responsibility with every finding we make. Jobs, livelihoods, relationships, reputations, business missions, public agency services, and more are impacted by our work. While we operate under the forgiving standard of “preponderance of evidence,” we know our impact is significant. And, often, our work is misunderstood by those to whom it should matter the most: complainants, respondents, witnesses, decision-makers, judges, juries, and even the public when scandals hit the news. Therefore, before we jump on a new technology wagon promising to revolutionize and simplify our work, we must tread carefully.
At VM, we have been quick to explore and invest in the potential of GenAI, but slow to implement it. We experimented with a multitude of commercial offerings, from widely available chatbots to specialized solutions for legal services. Ultimately, we have invested in building our own tool, built by an expert in machine learning, from the ground up with the investigation process, and industry guiding principles, integrated into the system. We believe that, while every investigator should explore its potential, it is even more important to understand how GenAI works, its risks, and, most of all, its limitations, based on our deep experience as investigators. Only when we understand these elements can we make use of the technology in a way that upholds our ethical responsibilities and standards of excellence in the execution of our critical work as fact finders.
It has always been VM’s philosophy that our investigative industry is better the more we share information and collaborate with colleagues. We believe this extends to leveraging GenAI technology. The workplace investigation community is a rich brain trust and is accountable to clients and decision-makers for the responsible implementation of this technology, keeping industry guidance like the Guiding Principles published by the Association of Workplace Investigations (AWI) in mind. With this objective, following are some initial observations we have made about the potential uses, risks, and pitfalls of integrating GenAI in our work. We encourage others to share their experiences as well. Please note that our work to date has been exploratory and we have not fully adopted GenAI into our client work flows yet. We anticipate doing so in 2025.
We began by evaluating how GenAI could apply to the different stages of the investigation process. The following is a list of practical use cases that we believe are within the current capability of GenAI. However, how the tool is built and trained will significantly impact the quality of its output.
Complaint Intake:
Transcribe hotline complaints
Review complaint information and identify potential allegations
Identify relevant policies and procedures
Craft responses to complainants acknowledging receipt through a hotline system
Categorize complaints based on the nature of the allegations (e.g. harassment, fraud)
Investigation Planning:
Develop an evolving investigative plan for conducting the investigation
Determine initial scope (and continue to reevaluate based upon evidence)
Identify potential witnesses and related matters, including interview format, witness sequencing, and notifications
Identify potential documents and evidence to obtain
Prepare for interviews, including identifying specific statements from other witnesses about prompted topics, e.g. in preparation for the respondent’s interview
Consider interim action based on organizational past practices and prioritize investigative steps based on user-identified risks and urgency
Evidence Gathering:
Accurately transcribe witness interviews in real time
Provide a running list of topics covered during interviews and recommend other issues to address or questions to ask
Real-time verification of whether all planned topics from an interview outline have been covered during the interview
Identify potential evidence to gather based on analysis of interview responses
Analyze and highlight contradictions between witness statements in real-time
Documenting the Investigation:
Assist with drafting initial elements of the report, including witness summaries and synthesizing data
Offer revisions or suggestions on draft language, including ensuring language is neutral in tone
Identify information from witnesses that supports the investigator’s finding
Proofread written work product for errors and inconsistencies
Part Two will explore some of our key learning outcomes in implementing GenAI, some of the key risks in adopting GenAI in your practice, and advice on how to evaluate potential tools.
Van Dermyden Makus benefits from the kind of expertise that can only be gained through the collective experience of several thousand interviews. Our team approach means every investigation we conduct benefits from that experience. Chances are, whatever difficult interview (or investigation) issue you are facing, we have dealt with something similar before; if not, we welcome the new challenge.