ing” to train a generative AI tool to use published analyst estimates and past company announcements to create an earnings script. “Troughout the training process, we ensured that the gen-
“Even the tone of your presentation is customized because when the AI program looks at our scripts, it’s taking into account the preferred tone of our CEO, CFO, and IRO.”
- Amit Sanghvi, Q4
erative AI output would remain true to each speaker’s tone and style,” he says.
Te team also trained a tool to forecast the types of questions
that might be asked on the next earnings call by reviewing previ- ous questions from analysts, as well as peer events. Other source information that can support AI analysis of IR activities includes 13-F filings, shareholder information, web traffic data, email read notifications, and event attendance statistics. “What we’re doing is taking the natural language generation
power of the AI program without having to rely on the full corpus of data that it has. You can train it in a safe enclosed environment with data that you want to provide. We started to get some amaz- ing results. “We prompted the program with instructions that said, ‘Don’t focus on the specific contents of this script; focus on how the scripts are constructed. Start to decipher how we present, what we present, when we present, and what order we present in. Te next prompt said, “Use [this] data set of publicly available
information that we have released since the end of the last quar- ter to make notes on information that could be used in our next earnings presentation.” A further exercise by the Q4 team provided the following prompts
to an AI program to help it tailor earnings scripts depending on different scenarios: Prompt: Armed with the earnings template, analyst estimates, and notes for the new scripts, populate the script with an assumption that Q4 will: • Beat forecast • Meet forecast • Miss forecast Finally, they manually pulled appropriate language and further tweaked the script for accuracy and relevance. “Even the tone of your presentation is customized because when the AI program looks at our scripts, it’s taking into account the preferred tone of our CEO, CFO, and IRO,” Sanghvi says. Using these tools, he notes, “I’ve never seen our management
team more relaxed while preparing the presentation leading up to an earnings call. A lot of the legwork was taken care of through this sort of an approach.” IR
Al Rickard, CAE, is Editor-in-Chief of IR Update and President of Association Vision, the company that produces IR Update for NIRI;
arickard@associationvision.com.
niri.org/ irupdate IR UPDAT E ■ FA L L 2 0 2 3 1 3
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36
