An AI primer for propane retailers

July 16, 2026 By Christopher Caywood 0 Comments

I have written several columns about artificial intelligence (AI) over the past few years.

Those columns have asked whether AI and machine learning can help a propane operation today, weighed the pros and cons for propane marketers, looked at AI-supported customer communication, explored AI-assisted marketing and argued that human relationships still come first.

But I recently realized I may have started in the middle of the movie. Before asking how propane retailers should use AI, it is worth asking what AI is, how it works and why its usefulness depends less on magic than on the knowledge, judgment and creativity of the user.

So this column, unlike many of today’s movie thrillers, is going to step back and start at the beginning.

For me, starting at the beginning is rooted in more than just a few columns. It also comes from experience. During the last few years of owning a business, we implemented AI-supported or machine-learning solutions for accounts payable processing, customer communications and marketing.

Some of the basic, more self-contained tools provided real benefits. We saved hundreds of hours annually on bookkeeping with machine-learning tools and thousands of dollars by producing our own animated marketing videos and social media messages. We were even able to keep a more careful eye on customer communications.

But we never fully realized the benefits of the customer communication tools. The reason was simple: Our knowledge base, as good as we thought it was, just wasn’t quite good enough. We had started in the middle of the movie.

▶ Lessons learned

I still consider it a success, largely because I learned from the mistakes. And as Winston Churchill is often credited with saying, “Success is the ability to move from one failure to another with no loss of enthusiasm.” So with that in mind, let’s enthusiastically begin with a basic AI dictionary for propane retailers.

Artificial intelligence: Software that performs human-like tasks, including the three “r’s” – reading, writing and arithmetic – and then summarizing, classifying, predicting, recommending and sometimes taking action.

Large language model: A type of AI trained on enormous amounts of text and data to recognize patterns in language and generate responses.

Generative AI: AI that creates new material, including emails, policies, advertisements, call scripts, customer summaries, training outlines and operating procedures.

Machine learning: The process of finding patterns in data and improving those patterns as more data becomes available.

Agent: An AI tool given a specific job, instructions, access to approved information and sometimes authority to use other tools.

Agentic AI: An AI agent on steroids. It can work through a series of tasks with less handholding, such as reviewing a customer message, checking policy, drafting a response, routing the issue and creating a follow-up task.

Knowledge base: The reference library used by AI. For propane retailers, that library may include company policies, safety manuals, compliance rules, delivery protocols, customer service scripts, tank-setting guidelines, pricing rules, training materials, website content, FAQs, call center materials and, perhaps most importantly, the institutional knowledge in your head and your employees’ heads.

A basic understanding of these definitions will help you follow the AI conversation, understand what you read in the press and on social media and maybe even make you the hit of a cocktail party when the topic comes up.

The next several columns will walk through the architecture: how to build a usable knowledge base, how to protect data, where to begin with AI, and, most importantly, whether and how to move from simple off-the-shelf tools to deeper integrations.

Christopher Caywood is a board member of Guardian Propane Partners LLC. Contact him at ccaywood@guardianpropane.com.

Featured homepage image: Kulpreya Chaichatpornsuk/iStock/Getty Images Plus/Getty Images

Connecting the dots: From operating efficiency to enterprise value

Relationships first, technology second

Guidelines to keep your 2026 technology resolutions