Harnessing Large Language Models to Transform Utility Vegetation Management

Utility vegetation management (UVM) has evolved significantly over the past several decades. Today, foresters and vegetation managers are spending more time behind a screen than under the canopy. From writing work plans and responding to regulators, to preparing stakeholder communication and compiling contractor audits, the work has become increasingly administrative. And while technology often promises to help, it can just as easily complicate.

That’s where large language models (LLMs) come in. These tools aren’t here to replace the need for field work or entirely remove administrative tasks; they’re here to make the office work more manageable. LLMs are helping UVM professionals draft documents, summarize reports, improve internal communication and reduce time spent on repetitive writing tasks. This article will explore where they fit into daily operations, how they’re already being used by UVM professionals and how you can get started.

Cutting Through the AI Buzz

Artificial intelligence (AI) has become the banner under which seemingly every new tool is marketed across every industry. From health care to finance to vegetation management (VM), “AI-powered” solutions are everywhere. The result? A mix of hype, confusion and skepticism.

UVM is just as susceptible to this AI-infusion, and this overuse of the term has led many professionals to tune out. Large language models offer something different. These tools are designed to work with words, not just data points or maps. They help with the tasks that occupy much of a UVM manager’s time: administrative office work.

Unlike much of the AI noise, LLMs have immediate, practical applications. They don’t require custom integrations or enterprise-wide rollouts. They are accessible in a browser and ready to support the real work happening in UVM offices today. Best of all: they are free to try and use on a limited basis.

Many professionals are already familiar with tools like OpenAI’s ChatGPT or Microsoft’s Copilot but have yet to apply them in a work environment. This is where a significant opportunity lies. By using familiar platforms with industry-specific prompts and materials, UVM teams can begin discovering new efficiencies without additional investment or IT department overhauls. With responsible experimentation and thoughtful application, these tools can become another tool in your UVM toolbox.

What Are Large Language Models?

Large language models are a form of AI trained to understand and generate human language. They excel at reading text, summarizing content, drafting documents and answering questions based on the information you give them.

To put it in context: Apple’s Siri is an example of AI because Siri understands spoken language and answers basic questions. Netflix recommendations are powered by machine learning (ML) to personalize your recommendations by predicting your preferences based upon behavior and viewing history. Face ID technology uses deep learning to analyze and recognize your face, regardless of whether you grow a beard, wear glasses or otherwise change your image. Generative AI goes a step further and creates new content, like digital artwork or music. Large language models like ChatGPT and Microsoft Copilot are part of this last group, and they focus specifically on working with language.

The advantage of LLMs is their flexibility. They work directly in a web browser, respond to natural language input and require no special software to get started.

Most LLMs operate as general-purpose tools, but their strength in UVM comes from specificity. The more you tailor your prompts and inputs — using real-world documents, reports and terminology — the more relevant and valuable the output becomes. For example, asking a model to rewrite a public notification letter with safety-focused language and a fifth grade reading level can yield a community-ready draft in seconds.

Moreover, LLMs are becoming increasingly multimodal, meaning they are beginning to interpret not just text but other data formats such as charts, photos and PDFs. While this is still emerging, the future includes models that can cross-reference multiple data types in a single session, allowing for deeper insight into UVM reports, visual inspections and asset management records.

A Real-World Example: Liberty Utilities

Jason Grossman, manager of vegetation management at Liberty Utilities in Missouri, has been experimenting with LLMs for over the past few years. His approach is focused, results-driven and directly tied to the work his team does every day.

Here’s how his team is using LLMs:

• Summarizing Incident Reports: Grossman inputs vegetation-related incident reports into an LLM and receives a concise summary with recurring patterns and suggested preventive actions.
• Turning Regulations into Microlearning: Long, complex utility documents are summarized into bullet points or turned into simple internal training pieces that are easier for staff to absorb.
  Drafting and Reviewing RFPs: From scopes of work to evaluation criteria, he uses LLMs to draft and refine documents that would otherwise take hours.
• Supporting Public Communication: Stakeholder letters, social media posts and public education content all start with an LLM-generated draft that his team tailors before publishing.
• Data Analysis: Grossman uses LLMs to analyze historical regrowth data following tree growth regulator applications, helping him evaluate treatment success over time. This allows him to make data-informed decisions about future use and advocate for broader integration of TGRs within his IVM program.

Each of these use cases began with curiosity and a willingness to test. His team documents the prompts they use, evaluates the results and refines their process over time. This allows them to create internal templates that others on the team can reuse, thereby making LLMs more efficient with each project. Their approach demonstrates that adoption doesn’t need to be formal or complicated. A practical, trial-based mindset can lead to measurable gains in quality and turnaround time.

Common LLMs in Use Today

Before diving into hands-on applications, it’s helpful to know which tools are leading the way. Several LLMs are readily available for use in UVM-related workflows, each offering unique strengths.

ChatGPT (OpenAI): Perhaps the most widely recognized, ChatGPT is available through a browser-based interface and offers a conversational approach to language tasks. It’s capable of drafting content, answering questions and summarizing documents. Paid versions include access to advanced models and features such as custom instructions and memory.
Copilot (Microsoft): Integrated into Microsoft 365 applications like Word and Excel, Copilot brings LLM functionality directly into familiar office tools. It’s especially useful for teams already using Microsoft systems and can support task-specific writing, formula generation and report drafting within your existing workflow.

• Claude (Anthropic): Claude is known for its helpful tone and strong performance in long-document summarization. It’s designed to be more cautious in its output, which some users prefer for sensitive or regulatory language.
• Notebook LM (Google): Designed specifically for knowledge management, Notebook LM allows users to upload internal documents and ask questions based only on that library. It’s a good choice for organizations building searchable archives of institutional knowledge.
• Gemini (Google): A flexible, Google-connected model with strong search and reference abilities. Gemini is good at pulling in external context and responding in a straightforward tone.
• DeepSeek (DeepSeek AI): An emerging LLM with strong multilingual capabilities and fast performance. However, it is owned by a Chinese company, and some utilities may choose to restrict its use due to data security and jurisdictional concerns.

Regardless of the platform you choose, it is essential to follow best practices around data privacy. Under no circumstances should sensitive, proprietary or customer-specific information be shared with a public LLM. Treat each interaction as if it were a public conversation and always review and edit content before sharing it externally.

Five Ways You Can Start Using LLMs

You don’t need to overhaul your systems or change your workflows to try an LLM. These tools are available via web browser and can support you in a few common UVM tasks:

• Draft Homeowner Notices: Input a few details, and the LLM will draft professional, respectful notices that can be reviewed and mailed out.
• Summarize Regulatory Documents: Turn long reports into key bullet points or even first-draft SOPs.
• Build Templates: Create consistent audit forms, QA checklists or weekly summary reports.
• Capture Field Insights: Convert notes or debriefs into training documents or job briefings.
• Educate the Public: Generate plain-language content explaining pruning practices, herbicide use or storm preparation measures.

Each of these use cases can be expanded as your team gains confidence. A single success often opens the door to more creative applications, such as onboarding support, annual report writing or grant proposal development. In many cases, teams find that LLMs are most helpful not in generating perfect results, but in giving them a strong first draft to improve upon.

Time savings is a major benefit, but so is consistency. Once you’ve created a prompt and reviewed the output, the same task can be repeated month after month with minimal edits. This makes LLMs especially useful for recurring reports, social media updates and seasonal communications.

Bridging the Gap Between Office and Field

Clear communication is the backbone of a successful vegetation programs. From utility staff and supervisors to foremen and field crews, miscommunication can lead to delays, safety issues or unnecessary conflict.

LLMs can help. By converting technical language into field-ready instructions or customizing summaries for different roles, they ensure that everyone is working from the same page. For example, a vegetation manager’s notes can become:

• A concise job brief for the general foreman.
• A one-pager for the crew.
• A summary for the contractor’s admin staff.

Grossman has used LLMs in this exact way to improve clarity while reducing the need for follow-up calls or last-minute corrections.

This is about much more than saving time; it’s about improving alignment across the entire UVM program. When everyone from planners, tree crews, herbicide crews and utility staff are working with the same information, formatted in a way that speaks to their role, the result is better coordination and fewer errors. And when unexpected changes arise — as they often do in field work — the ability to quickly update and redistribute clear, revised instructions becomes even more valuable. This alone is a prime reason to try LLMs.

UVM work involves unique constraints: terrain, weather, local policy and landowner expectations all shape what can be done. LLMs can’t replace judgment, but they can help document and share that judgment more effectively.

Preserving Institutional Knowledge

Much of what makes a vegetation program run smoothly is stored in people’s heads. As senior staff retire or change roles, there’s a risk that institutional knowledge disappears with them.

LLMs can help turn informal knowledge into reusable resources. By inputting field notes, audio transcripts or rough bullet points, you can generate formal documentation like safety briefings, onboarding materials or process guides.

Notebook LM is one tool designed for this kind of task. Unlike general LLMs, it only uses the material you upload. That makes it ideal for creating a secure, internal reference library based entirely on your own documentation.

Think of this not just as backup but as a foundation for training and continuous improvement. A structured knowledge library ensures that lessons learned on one circuit, in one season, or with one contractor don’t get lost. Instead, they become part of your team’s long-term memory, accessible and easy to update.

And when paired with prompt-based workflows, the benefit multiplies. You can ask the model specific questions such as, “What did we do differently on the Fairfield substation after the 2022 storm season?” and receive a clear, concise response drawn directly from your archived documents.

Running a Pilot: Start Small, Learn Fast

If you’re curious about LLMs, the best way to begin is by trying one. Choose a task that’s repetitive but important, like drafting customer notices or summarizing outage logs. Assign a team member to test the tool, review the results and share what they learn.

Create a shared folder for successful prompts so your team can build on each other’s experience. You’ll quickly identify where LLMs help and where human input is still essential.

One common concern is accuracy. LLMs should never be used to make safety decisions or replace subject matter expertise. Instead, treat them as a digital assistant that’s fast, adaptable and capable of helping with language-based tasks. The most effective pilots often begin with non-public-facing content such as internal training materials, policy drafts or QA checklists.

Be intentional about feedback. Keep track of what works and what doesn’t. Over time, you’ll develop a list of best practices that reflect your unique workflows, tone and communication style. That internal playbook will become a valuable asset.

Where This is Headed

While today’s LLMs operate mostly as stand-alone tools, future versions will likely be integrated with GIS, inspection software and contractor management systems. This could allow LLMs to generate reports based on real-time data, assist in multilingual communication or even flag high-risk areas by referencing historical patterns.
Utilities that begin experimenting now will be better prepared for these advancements.

Developing comfort and fluency with LLMs today makes it easier to evaluate and adopt more advanced integrations later.

The convergence of LLMs with other AI tools — such as image recognition, LiDAR and drone inspection software — could lead to systems capable of producing comprehensive vegetation risk reports with minimal manual intervention. Even before those systems arrive, LLMs can assist in synthesizing outputs from existing platforms, offering narrative summaries and recommendations alongside maps and tables.

It’s not hard to imagine a future where LLMs are embedded directly into vegetation dashboards, helping managers interpret contractor progress, identify missed work and draft weekly updates — all from the same interface.

Writing Better Prompts for Better Results

Getting the most out of an LLM requires one key skill: asking better questions. A well-crafted prompt leads to a better, more useful response.

Instead of saying, “Write a letter,” say, “Write a letter to homeowners letting them know that [insert contractor name] will be pruning trees in two weeks as part of our regular safety maintenance. Emphasize professionalism and safety. Include [insert utility] contact information.”

Specify your audience, your tone and your goal. Save prompts that work well so others on your team can use them too. You’ll soon have a small internal playbook for everything from stakeholder notices to QA summaries.

Prompt crafting is not about using perfect language. It’s about being specific, thoughtful and willing to revise. Don’t be afraid to ask for rewording, alternative formats or summaries with varying levels of detail. LLMs are tools, and they perform best when guided by someone who knows the context.

As your team gets better at working with prompts, you’ll begin to see patterns — certain wording works well, certain structures yield better outputs. Treat this knowledge as a living resource, one that can be shared and improved upon over time.

Supporting the People Who Keep the Lights On

UVM is a people-first profession. It’s about safety, reliability and community. It’s also about making complex decisions, often with limited time and resources. Large language models offer a way to support our UVM teams doing this work not by replacing them, but by giving them tools to work more efficiently.

Whether you’re a program manager, forester or general foreman buried in spreadsheets and reports, LLMs can help. They won’t take you out of the office, but they might help you spend less time there. After all, most of us entered this profession because we appreciate trees. Wouldn’t it be nice to spend more time with them in the field?

Editor’s Note: To learn more about AI, ML and LLMs in UVM, listen to the Line Life Podcast episode featuring Phil Swart and Jason Grossman at linelife.podbean.com.