#15 AI Agents
This week, I discuss AI agents—an exciting, fast-emerging area of AI.
AI will change the world, but how will it change M&A? I want to focus on AI’s impact on M&A in this newsletter. I am not an expert on either M&A or AI, but I want to learn about both topics and how they intersect. I thought there might be others in my situation (or people who are experts in one field or the other) who would find information on M&A and AI helpful in their careers, so I created this newsletter to track and share what I learn.
AI Agents
AI agents are chatbots that can accomplish complex tasks that require multiple steps. I recently dove into some articles about AI agents, and here is what I learned.
What is an AI Agent?
According to a Nvidia Blog post, an AI agent “is a system that can use an LLM to reason through a problem, create a plan to solve the problem, and execute the plan with the help of a set of tools.”
Today, when you ask ChatGPT to write an essay, it will write an essay “without using the ‘delete’” key. AI Agents, on the other hand, can plan their work on the essay just like a human. For example, when asked to write an essay, an AI agent will research, create an outline, write a first draft, and edit the draft. As a bonus, each step in the plan can be performed by an agent (or specialized AI model) that specializes in the task. The idea is to combine specialized AI models into one—improving overall performance.
Here is a great video explaining AI agent technology:
The video is a commentary on a talk that Andrew Ng, an AI expert, gave to Sequoia Capital, the famous VC firm. The commentator gives great background on agents and makes Andrew’s talk make sense to people who are not experts on AI.
Use cases
The Nvidia blog points out a few use cases for AI agents. The following are the ones I think would be the most helpful in M&A.
First, the “talk to your data” agent, or as I like to call it, enhanced “ctrl-f.” AI’s enhanced “ctrl-f” features will be one of the most impactful for many professionals. AI agents will take “ctrl-f” a step further by adding the ability to search multiple documents and accurately answer multi-step questions. The article gives an example of a prompt that requires an AI agent: “What was the difference between Company X’s Q3 earnings and Q4 earnings?” This prompt requires three steps: (1) What were Company X’s Q3 earnings? (2) What were Company X’s Q4 earnings? (3) What was the difference between the two? Unless there is an article published somewhere on the internet that contains the answer to this question (which in this simple example is very possible), normal chatbots would be unable to answer it because it would require finding and searching multiple documents to answer one question. AI agents can answer the question by sending an agent to each document to retrieve the information and then analyzing it together to get the correct answer.
Second, multi-authored AI outputs. Different tones are needed for different writing tasks (I write this blog in a different tone than my law school exams, for example). Normal AI chatbots, according to the article, struggle to pick up on necessary changes in tone. An AI agent can take multiple steps when prompted to create written work. Continuing my previous example, when asked to write a blog post, it could “research” my previous posts and match my tone in its response. It is possible to fine-tune today’s AI models to pick up on tone (by providing examples in the bot’s training or in a prompt), but they are likely not as good as using AI agent-enabled models.
Last, multi-modal agents would help M&A professionals analyze documents more quickly. For those unfamiliar, “multi-modal AI” refers to an AI model’s ability to analyze and produce different types of outputs, like text, photos, or audio.1 A multi-modal tool could analyze a slide deck’s graphs and texts, giving the user a more complete and accurate view of the slide decks. The “agentic” piece to this would be that only one AI agent chatbot could perform these tasks, instead of the multiple chatbots that would be required today.
Conclusion
Many in the AI community consider AI agents as the future of AI, and the tools are developing quickly. As far as M&A use cases, I think the above examples give a good baseline for what AI agents could do in an M&A setting. I think the overall effect of AI agents will be more specialized models that are really good at distinct tasks. Contrast this with today’s chatbots like ChatGPT, which are generally good but struggle with highly specific tasks.
About Me
My name is Parker Lawter, and I am a law student pursuing a career as an M&A lawyer. I am in my last semester of law school, and with some extra time on my hands, I decided to create this newsletter. I hope it is informative and helpful to anyone who reads it! I am not an expert at either M&A or AI, but I am actively pursuing knowledge in both areas, and this newsletter is a part of that pursuit. I hope you’ll join me!
Follow me on LinkedIn: www.linkedin.com/in/parker-w-lawter-58a6a41b
All views expressed are my own!
https://ai.meta.com/tools/system-cards/multimodal-generative-ai-systems/