From Chatbots to Digital Teammates: How AI Agents Fit Into Agribusiness
Index
AI Agents: A Digital Teammate
OODA Loop
Vertical Agents
Levels of Agents
3 Agent Use Cases for Agriculture and a Video
Where Do Agents Live?
Control Points for Agribusiness
Control Points for Farmers
Caveats
Final Thoughts
AI Agents: A Digital Teammate
In the ever-evolving world of AI, LLM agents— autonomous applications powered by Large Language Models—have the potential to evolve how we think about performing day to day task as knowledge workers in agriculture.
Today, there are numerous companies offering an AI interface:
I wrote a primer on LLMs in agriculture in early 2023 called ChatGPT Implications for Agriculture— you can read more on the fundamentals of LLMs there.
Agents are an advanced evolution of LLMs.
An LLM agent is an artificial intelligence system that utilizes a large language model as its core engine to deliver capabilities beyond text generation, including conducting conversations, completing tasks, reasoning, and even demonstrating some degree of autonomous behavior.
LLM agents can be directed through prompts that encode personas, instructions, permissions, and context to shape the agent's responses and actions.
The key advantage of LLM agents is their ability to have varying degrees of autonomy. Based on the capabilities built during the design phase, agents can exhibit self-directed behaviors ranging from purely reactive to highly proactive.
LLM agents, with proper access to data and prompting, can work semi-autonomously to assist people in a range of applications, from the typical chatbots to goal-driven automation of workflows and tasks.
This enables possibilities for a customizable digital teammate integrated directly into software where work happens today and accomplish tasks.
OODA Loop
You might be familiar with The OODA Loop, a decision-making framework that consists of four stages:
observation (O)
orientation (O)
decision (D)
action (A)
Developed by military strategist John Boyd, this process is intended to help individuals and organizations make effective decisions in rapidly changing environments by continuously cycling through these stages.
The way to think about agents is that they will be applying this type of “loop” continuously, working towards the goal or objective they were prompted to execute.
Vertical Agents
There is a liklihood everyone uses a multitude of agents— one might be for optimizing your travel schedule, another might be for project management via Microsoft Office at your job, another might be a research assistant.
Some might be highly domain specific— trained on industry jargon, and understanding of the nuance inherent in the industry— like agriculture. A vertical agent.
There is opportunity in vertical LLM agents— an agent designed and fine-tuned for a particular industry or domain, such as agriculture.
Unlike general-purpose LLMs, vertical LLM agents focus on understanding and responding to the unique terminology, challenges, and data within a specific sector. This specialization enables them to provide more accurate, relevant, and actionable insights, making them highly effective for industry-specific tasks.
Levels of Agents
Currently, LLMs can be used for some fundamental decision support like these:
But recently, I read Levels of AI Agents: from Rules to Large Language Models that highlighted the escalating capabilities of LLMs agents and what each level of agency looks like and how that can evolve the use cases and autonomy of these systems: