by Avi Gesser and Michael Pizzi
From left to proper: Avi Gesser and Michael Pizzi (pictures courtesy of Debevoise & Plimpton LLP)
In Half 1 of this collection, we “vibe coded” a recreation designed to assist prepare legal professionals {and professional} employees on the agency’s new AI coverage utilizing Generative AI. In Half 2, we show how legal professionals can use AI instruments to quickly transcribe and summarize audio recordings from congressional hearings, oral arguments, or depositions, and rework them into quite a lot of easy-to-digest codecs. Particularly, we used NotebookLM and Gamma to transcribe a current 30-minute Second Circuit oral argument after which create a two-page written abstract, a “podcast”-style audio abstract, and a PowerPoint presentation—all in below 9 minutes.
Suppose you characterize a consumer in a false promoting lawsuit. The case raises a number of authorized points which might be at the moment being litigated in comparable circumstances in courts across the nation. To advise your consumer on technique, that you must hold observe of the arguments being made in these different circumstances, assess how they’re touchdown with judges, and anticipate questions you may get requested at an upcoming oral argument in your individual case. Lots of the courts publish audio recordings of the arguments on their web sites, however no transcripts.
For this demo, we used a current oral argument in a false promoting case that’s at the moment earlier than the Second Circuit (Zesty Paws LLC v. Nutramax Laboratories, Inc.). Taking it step-by-step:
- We downloaded the audio recording as an .mp3.
- We created a brand new NotebookLM pocket book after which dragged and dropped the .mp3 into the “Sources” window.
- Within the “Chat” window, we instructed NotebookLM to create a full transcript of the audio file with speaker tags and time stamps.
Instance Immediate: “Please create a transcript of this recording. Put the speaker names in daring and add time stamps each one minute.”
Instance Immediate: “I’m an affiliate at a legislation agency and I would like to offer a consumer with an correct two-page abstract of the argument. My agency has no involvement on this case. Are you able to put together that for me?”
- We additionally created an audio abstract of the oral argument to take heed to on our morning commute. To create the podcast, we began by producing an extended abstract utilizing the “Chat” window to offer NotebookLM extra content material to work with.
Instance Immediate: “Are you able to do a five-page abstract as a substitute?”
- We saved this five-page abstract as a Supply by deciding on “Add” within the “Sources” menu, copying the textual content into “Paste textual content”, and clicking “Insert”. Again in the primary pocket book view, we unselected the .mp3 file in order that NotebookLM would solely summarize the five-page abstract when producing the podcast. We then chosen “Audio Overview” within the “Studio” menu and, in about 5 minutes, had a podcast-style, 17-minute abstract of the oral argument that we might obtain or stream on the NotebookLM app.
- Whereas the podcast generated, we turned this abstract right into a client-ready presentation. We opened Gamma, chosen “Paste in textual content”, then copied within the five-page abstract from Step 5. We selected “Generate” from the choices under and, utilizing the fields in “Textual content Content material”, instructed the device on our desired degree of element, format, language, tone and design. We made a pair changes to the default settings, together with the next:
Quantity of Textual content: Minimal
Write For: “Consumer, mixture of enterprise and authorized professionals with restricted information of trademark/promoting litigation.”
- Lastly, utilizing the “Agent” function, we instructed Gamma so as to add a further slide, proper after the title slide, that gives an summary of the case.
We haven’t completely QC’ed the accuracy of those summaries, and we did discover a number of typographical errors within the transcript—e.g., “a stoppple” as a substitute of the authorized time period “estoppel”. However a fast comparability of those supplies in opposition to the recording suggests they captured the final thrust of the authorized arguments on the listening to.
After all, as legal professionals, we have to fastidiously contemplate the dangers inherent in utilizing these or comparable instruments for consumer work and take a number of steps to mitigate these dangers. Along with fastidiously QC’ing any outputs for accuracy, you may also:
- contemplate including a disclaimer stating that these summaries are preliminary, had been generated utilizing AI instruments, haven’t been checked for accuracy, shouldn’t be relied on as an correct abstract of the argument, and many others.;
- keep away from importing any confidential consumer data into these instruments, until your agency has an enterprise / licensed model, and such use will not be opposite to any consumer directions; and
- follow trusted supply materials, resembling an official unedited audio recording from the Second Circuit’s web site
Avi Gesser is a associate and Michael Pizzi is an affiliate at Debevoise & Plimpton LLP. This publish was initially revealed on the Debevoise Information Weblog.
The views, opinions and positions expressed inside all posts are these of the authors alone and don’t characterize these of the Program on Summer time 2025 Compliance & Enforcement Weblog Company Compliance and Enforcement (PCCE) or of the New York College Faculty of Legislation. PCCE makes no representations as to the accuracy, completeness and validity or any statements made on this web site and won’t be liable any errors, omissions or representations. The copyright of this content material belongs to the authors and any legal responsibility on the subject of infringement of mental property rights stays with the authors.
by Avi Gesser and Michael Pizzi
From left to proper: Avi Gesser and Michael Pizzi (pictures courtesy of Debevoise & Plimpton LLP)
In Half 1 of this collection, we “vibe coded” a recreation designed to assist prepare legal professionals {and professional} employees on the agency’s new AI coverage utilizing Generative AI. In Half 2, we show how legal professionals can use AI instruments to quickly transcribe and summarize audio recordings from congressional hearings, oral arguments, or depositions, and rework them into quite a lot of easy-to-digest codecs. Particularly, we used NotebookLM and Gamma to transcribe a current 30-minute Second Circuit oral argument after which create a two-page written abstract, a “podcast”-style audio abstract, and a PowerPoint presentation—all in below 9 minutes.
Suppose you characterize a consumer in a false promoting lawsuit. The case raises a number of authorized points which might be at the moment being litigated in comparable circumstances in courts across the nation. To advise your consumer on technique, that you must hold observe of the arguments being made in these different circumstances, assess how they’re touchdown with judges, and anticipate questions you may get requested at an upcoming oral argument in your individual case. Lots of the courts publish audio recordings of the arguments on their web sites, however no transcripts.
For this demo, we used a current oral argument in a false promoting case that’s at the moment earlier than the Second Circuit (Zesty Paws LLC v. Nutramax Laboratories, Inc.). Taking it step-by-step:
- We downloaded the audio recording as an .mp3.
- We created a brand new NotebookLM pocket book after which dragged and dropped the .mp3 into the “Sources” window.
- Within the “Chat” window, we instructed NotebookLM to create a full transcript of the audio file with speaker tags and time stamps.
Instance Immediate: “Please create a transcript of this recording. Put the speaker names in daring and add time stamps each one minute.”
Instance Immediate: “I’m an affiliate at a legislation agency and I would like to offer a consumer with an correct two-page abstract of the argument. My agency has no involvement on this case. Are you able to put together that for me?”
- We additionally created an audio abstract of the oral argument to take heed to on our morning commute. To create the podcast, we began by producing an extended abstract utilizing the “Chat” window to offer NotebookLM extra content material to work with.
Instance Immediate: “Are you able to do a five-page abstract as a substitute?”
- We saved this five-page abstract as a Supply by deciding on “Add” within the “Sources” menu, copying the textual content into “Paste textual content”, and clicking “Insert”. Again in the primary pocket book view, we unselected the .mp3 file in order that NotebookLM would solely summarize the five-page abstract when producing the podcast. We then chosen “Audio Overview” within the “Studio” menu and, in about 5 minutes, had a podcast-style, 17-minute abstract of the oral argument that we might obtain or stream on the NotebookLM app.
- Whereas the podcast generated, we turned this abstract right into a client-ready presentation. We opened Gamma, chosen “Paste in textual content”, then copied within the five-page abstract from Step 5. We selected “Generate” from the choices under and, utilizing the fields in “Textual content Content material”, instructed the device on our desired degree of element, format, language, tone and design. We made a pair changes to the default settings, together with the next:
Quantity of Textual content: Minimal
Write For: “Consumer, mixture of enterprise and authorized professionals with restricted information of trademark/promoting litigation.”
- Lastly, utilizing the “Agent” function, we instructed Gamma so as to add a further slide, proper after the title slide, that gives an summary of the case.
We haven’t completely QC’ed the accuracy of those summaries, and we did discover a number of typographical errors within the transcript—e.g., “a stoppple” as a substitute of the authorized time period “estoppel”. However a fast comparability of those supplies in opposition to the recording suggests they captured the final thrust of the authorized arguments on the listening to.
After all, as legal professionals, we have to fastidiously contemplate the dangers inherent in utilizing these or comparable instruments for consumer work and take a number of steps to mitigate these dangers. Along with fastidiously QC’ing any outputs for accuracy, you may also:
- contemplate including a disclaimer stating that these summaries are preliminary, had been generated utilizing AI instruments, haven’t been checked for accuracy, shouldn’t be relied on as an correct abstract of the argument, and many others.;
- keep away from importing any confidential consumer data into these instruments, until your agency has an enterprise / licensed model, and such use will not be opposite to any consumer directions; and
- follow trusted supply materials, resembling an official unedited audio recording from the Second Circuit’s web site
Avi Gesser is a associate and Michael Pizzi is an affiliate at Debevoise & Plimpton LLP. This publish was initially revealed on the Debevoise Information Weblog.
The views, opinions and positions expressed inside all posts are these of the authors alone and don’t characterize these of the Program on Summer time 2025 Compliance & Enforcement Weblog Company Compliance and Enforcement (PCCE) or of the New York College Faculty of Legislation. PCCE makes no representations as to the accuracy, completeness and validity or any statements made on this web site and won’t be liable any errors, omissions or representations. The copyright of this content material belongs to the authors and any legal responsibility on the subject of infringement of mental property rights stays with the authors.
