Why I’ve Stopped Using AI Notetakers and AI Meeting Summaries
Artificial intelligence has found its way into nearly every business process. One of the most popular uses today is the AI meeting notetaker. These tools promise to record meetings, generate transcripts, summarize discussions, identify action items, and distribute notes automatically. On the surface, that sounds like a productivity win.
However, after seeing the results firsthand and considering the potential legal implications, I have made the decision to completely remove AI notetakers from my meetings and avoid AI-generated meeting summaries altogether.
The decision is not based on resistance to technology. I use artificial intelligence in many areas of my work. The issue is accuracy, accountability, and the legal risks associated with creating records that may not accurately reflect what actually occurred during a meeting.
AI Notetakers Are Not Always Accurate
The biggest problem with AI-generated meeting notes is that they often present themselves as authoritative even when they contain errors.
An AI system may misunderstand a statement, misidentify a speaker, incorrectly summarize a discussion, or draw conclusions that were never intended. In many cases, the resulting summary sounds convincing. Someone reading the summary later may assume it is an accurate representation of the meeting when it is not.
I have personally seen situations where AI-generated summaries completely mischaracterized discussions and decisions. In some cases, the summary suggested that a decision had been made when no decision had actually been reached. In others, comments were attributed to the wrong person or critical context was omitted.
The result was confusion, unnecessary follow-up discussions, and wasted time as participants attempted to correct the record.
Ironically, the technology intended to improve productivity ended up creating additional work.
The Risk Increases When Others Rely on the Summary
The real danger is not necessarily the initial error. The greater risk arises when others rely on the inaccurate information.
A manager may review the summary and believe a team agreed to move forward with a project. An executive may assume a budget decision was approved. A client may believe commitments were made that were never actually discussed.
Once inaccurate information enters the decision-making process, the consequences can multiply quickly.
Unlike traditional meeting notes taken by an individual who understands the context of the discussion, AI-generated summaries often create a false sense of certainty. Readers may place more trust in the summary simply because it was generated automatically.
Legal Discovery Creates Additional Concerns
For me, the most significant concern involves legal discovery.
As an expert witness involved in legal matters, I regularly think about how documents, communications, and records may be viewed during litigation. Businesses often focus on emails, text messages, and internal reports when considering discoverable information. AI-generated meeting notes and summaries should be viewed through the same lens.
If litigation occurs, opposing counsel may request documents related to discussions, decisions, and communications. Depending on the circumstances, AI-generated transcripts, notes, summaries, and related records may become subject to discovery.
That creates a serious problem when the information is inaccurate.
An attorney reviewing an AI-generated summary may not know that the summary contains errors. The summary may suggest a statement was made when it was not. It may indicate a decision was reached when it was not. It may omit qualifying language that significantly changes the meaning of a discussion.
Even if those inaccuracies can eventually be corrected, the organization may spend substantial time and resources explaining why the AI-generated record does not accurately reflect what happened.
More Data Is Not Always Better
Many organizations operate under the assumption that collecting more information is always beneficial. That is not necessarily true.
Every recording, transcript, summary, and generated document creates another piece of information that must be managed, secured, reviewed, and potentially produced during litigation.
Organizations should carefully evaluate whether the benefits outweigh the risks.
If meeting participants are already taking notes, documenting decisions, and distributing approved meeting minutes, adding AI-generated records may simply create duplicate information that introduces additional opportunities for error.
Human Review Does Not Eliminate the Problem
Some vendors argue that AI-generated notes are acceptable because humans can review and correct them.
While human review certainly helps, it does not eliminate the underlying issue.
Someone must still spend time reviewing the transcript, verifying the summary, correcting inaccuracies, and ensuring that important context was not lost. In many situations, that review process consumes enough time that the productivity benefit largely disappears.
If a summary requires significant editing before it can be trusted, the organization should question whether the automated process is actually providing value.
My Approach
Because of the work I perform, particularly in legal matters and expert witness engagements, I have chosen to remove AI notetakers from my meetings entirely.
I do not use AI-generated meeting summaries. I do not rely on automated meeting notes. When documentation is necessary, I prefer records that are created intentionally, reviewed carefully, and approved by the individuals responsible for the content.
That approach may not be appropriate for every organization. However, I believe businesses should take a hard look at the potential risks before automatically enabling AI notetaking features across their meetings.
Convenience should never outweigh accuracy. When inaccurate information can influence business decisions, client relationships, regulatory compliance, or legal proceedings, the true cost of an AI-generated mistake can be far greater than the time saved by automation.
AI notetakers and meeting summaries can be useful tools in certain situations. However, organizations should understand that these systems are not infallible. They can misunderstand discussions, misrepresent decisions, and create records that may later become discoverable in legal proceedings.
Before enabling an AI notetaker, ask a simple question: If this transcript or summary were presented in court several years from now, would you be comfortable defending its accuracy?
If the answer is no, then it may be worth reconsidering whether the convenience is worth the risk.