From the smallest charities to the largest multinational corporations, organizations have traditionally viewed keeping minutes as a necessary evil, endorsing them as a critical tool for documenting corporate decision-making, but often bemoaning the time, expense and drudgery associated with their preparation.
Now, AI tools that record, transcribe or summarize meetings are touting their ability to produce high quality minutes at a lower cost. It is worth closely examining, however, what these tools actually produce. Traditionally prepared minutes are a carefully curated record: a concise, official account that a board reviews and deliberately adopts. AI note taking tools instead generate (or at least attempt to generate) a verbatim transcript or an automated summary of everything that may have been said. Labeling an AI output “minutes” no more makes it so than calling a rough draft a “contract” makes it binding. Raw transcripts and automated summaries become minutes only after careful human curation, and the gap between AI-generated outputs and curated records drives most of the benefits (and most of the risks) discussed below.
In general, the growing use of AI note-taking tools is a promising development. But pitfalls lurk. This article (i) outlines general advantages and disadvantages of traditionally prepared minutes and “minutes” prepared with AI tools, (ii) describes the best and worst case scenarios for using AI note-taking tools in such cases, and (iii) ends with a few concluding thoughts.
When the decision making of an organization’s directors or other fiduciaries is challenged by litigants or regulators, minutes are typically among the first documents requested. If well-prepared, minutes can provide important evidence of the directors’ reasoned decision-making. Conversely, poorly prepared minutes can create legal exposure, including by being weaponized by litigants to extract monetary damages or by regulators to impose fines.
A. Advantages and Disadvantages of Traditionally Prepared Minutes. When properly managed by corporate secretaries, traditionally prepared minutes can serve as the official record of how directors deliberated and reached their decisions. Careful document-management practices can minimize the creation or retention of digital recordings, directors’ notes, draft minutes or other materials that might present an incomplete or competing account. This reduces the opportunity for litigants or regulators to “cherry pick” isolated statements to question the directors’ motives or actions. Plus, traditionally prepared minutes are typically reviewed by company directors and counsel, which reduces the risk that any deliberations or commentary will be misstated or misinterpreted. Careful preparation and review of minutes can also help to ensure that confidential data and privileged communications are properly safeguarded. Well-prepared minutes also typically avoid naming which director raised particular concerns or provided specific input, which maximizes free flowing debate and minimizes the ability of litigants or regulators to pit directors against one another. Finally, skilled corporate secretaries are adept at summarizing discursive debates in a manner that fairly and concisely reflects the ultimate consensus decisions reached by the directors.
Weighing against these benefits, however, is the time, expense and “headache” of traditional minute preparation. For meetings involving decision making likely to be second guessed, companies typically conclude that they have no choice but to spend the time and effort to produce minutes designed to withstand scrutiny. Conversely, many organizations determine that routine meetings can be safely documented in a more cursory manner, or not documented at all. While this approach does not often create material risk, it can be perilous if an unforeseen legal challenge later arises and the organization has little or no documentation of the basis for the relevant decision making.
B. Advantages and Disadvantages of AI-Assisted Minutes. In many respects, the advantages and disadvantages of AI-assisted minutes are the converse of traditionally prepared minutes. The ease of recording meetings with AI notetaking tools is certainly saving organizations substantial time and expense. Plus, these tools are expanding the number and type of meetings being adequately documented. Meetings of volunteer-led non-profit organizations that previously were often under- or undocumented can now be adequately and efficiently reflected using AI tools. In addition, AI tools readily enable informal fact finding meetings to be documented by professionals who in the past were not properly trained or too busy to document their internal decision making processes. For example, an internal project team conducting routine operational planning can use an approved tool to create a preliminary record of action items and decisions that previously may have gone undocumented.
But lax use of AI note taking tools offered by third parties can have significant negative consequences, several of which stem from potential third party access to data recorded or transmitted on the third-party platforms. A company’s claim of attorney-client privilege could be challenged if the confidentiality of its communications is undermined by using a note-taking tool that makes data accessible to the vendor. Federal courts have begun grappling with how privilege and work product doctrines apply to AI tools, with outcomes turning heavily on who used the tool, the platform’s confidentiality terms, whether the tool was used at counsel’s direction and whether the use materially increased the risk of adversary access. Attorney-client privilege is especially confidentially dependent. Work product protection is generally more durable, but it is not immune from challenge where AI use is unmanaged, undertaken independently of counsel or conducted on consumer-facing platforms without adequate safeguards.
Like other similar cloud-based services, AI note-taking tools can create risks of data leakage, unauthorized access or theft. AI-recorded discussions of sensitive matters or strategic plans could be accessed by third parties, potentially creating serious competitive, legal or reputational damage. Unauthorized access to protected data may also trigger breach notification reporting and other obligations under regimes such as HIPAA, GLBA or state statutes. In addition, litigants or regulators may attempt to obtain AI-generated transcripts or summaries of company proceedings, thereby undercutting the company’s goal of treating its board-approved minutes as the official record of those proceedings. These risks can be reduced by using enterprise-grade products, coupled with appropriate contractual protections, access controls and defensible retention policies. Organizations should establish how long AI-generated summaries, transcripts and recordings will be retained and deleted in accordance with applicable document retention policies, provided that routine deletion is suspended once litigation or a regulatory inquiry is reasonably anticipated. But many companies lack the financial capacity or technical expertise to properly implement and operate enterprise-grade products in this manner, and even those that do cannot be completely assured that their efforts will succeed. If any such third-party data request is successful, the organization’s attempt to curate minutes that concisely summarize its deliberations into a single narrative could be thwarted if recorded conversations enable a challenger to conjure alternative narratives.
Additionally, many AI note-taking tools ascribe particular views to particular directors, which is contrary to the prevailing traditional view that the specific views of individual directors should generally not be recorded. The mere act of recording or transcribing meeting proceedings could therefore inhibit free-flowing and vigorous debate, which is often essential to optimal decision-making. AI transcription tools also make mistakes, sometimes attributing remarks to the wrong speaker or mis-transcribing a contraction (such as recording “doesn’t matter” as “does matter”). Perhaps even more troubling, statements made rhetorically or sarcastically might be transcribed literally by AI note-taking tools. Similarly, preliminary impressions, off-hand comments, jokes, or colorful exaggerations could be transcribed or summarized literally by an AI tool, even if they were never intended to represent final views or corporate positions. Summarization tools can also “hallucinate,” generating content that was never part of the meeting. While many of these deficiencies are correctable (and may diminish with improvements in AI technology), heedless use of current AI tools can lead to unintended results.
Finally, the unsupervised use of AI products by directors or senior executives can create other substantial risks. One particularly significant governance risk arises when absentee directors use unauthorized AI note-taking tools to summarize meetings they cannot attend. Unless the meeting administrator is aware of these unapproved parallel recordings, an organization could end up with two competing summaries of the same meeting, which could create substantial legal disadvantages if the summaries are materially inconsistent. An absentee director may also be unable to correct transcription or summarization errors because the director did not participate in the meeting. For these reasons, many organizations are either carefully screening virtual meeting participants or are flatly prohibiting any director’s use of AI note-taking tools. In addition, recent decisions addressing generative AI and privilege underscore that sensitive exchanges with public or consumer-facing AI chatbots may not be protected from discovery, particularly where the tool is used without counsel’s direction or adequate confidentiality safeguards. Other courts have been more protective where AI is used as a litigation-preparation tool, especially when the use is tied to counsel’s work or does not materially increase adversary-access risk. For that reason, organizations should caution directors and executives against using AI chatbots to discuss privileged, confidential or strategically sensitive corporate matters unless the use is authorized, controlled and appropriately documented.
A. Best Cases. Assuming that an organization has approved the particular tool and established appropriate access, security and retention controls, there are many scenarios where the benefits of using AI note-taking tools can exceed the disadvantages, including:
B. Worst Cases. The worst case scenarios for using AI note-taking tools include meetings involving any of the following:
In many circumstances, matters presented to a board may be too sensitive or confidential to warrant use of AI tools, subject to the limited exceptions noted herein. For this reason, many current AI tools are ill suited for routine use by large complex organizations that routinely discuss confidential or sensitive matters at most or all of their directors’ meetings. Conversely, in a limited number of cases, matters may be so routine that AI tools can be trusted to produce summaries with relatively modest human supervision. But, based on current technologies, the wisest approach in a wide variety of cases will likely be increased adoption of AI tools under the careful supervision of skilled corporate secretaries who understand their advantages and limitations. If properly managed, this balanced approach could successfully combine the cost saving benefits of AI with the time-tested risk reducing talents of corporate secretaries. The practical takeaway is straightforward: AI tools operate best as drafting aids, not at creating official records.
Visit Jones Walker’s AI Law and Policy Navigator at www.ailawblog.com.
This article was originally published in Corporate Counsel.