How a Fire Department’s HR Leader Used AI to Rebuild Recruiting

The adoption and implementation of artificial intelligence in HR is no longer a hypothetical. SHRM’s The State of AI in HR 2026 report revealed that 39% of HR professionals are using AI in their HR functions, and in the process, have seen slight or significant improvements in their efficiency (87%) and work quality (75%).  

Across the 16 HRX dimensions, from recruitment and compliance to talent management and leadership development, AI is being used to augment the HR function. To demonstrate how everyday HR professionals are doing this, we reached out to members to share their experiences. This is the first in a new article series of real-world AI use cases in HR, starting with recruitment.  

This article is an example of the Job Description Drafting and Refinement Job Descriptions use case. Discover 137 other AI use cases by visiting SHRM’s AI Field Manual for Employers: 138 Use Cases for AI at Work.

Early Adoption in a Virginia County’s Fire and EMS Department

Chesterfield County in Virginia’s fire and EMS department once fielded 2,000 to 4,000 applications for open positions, according to Christina Smith, an HR administrator who has worked in fire service administration for 26 years, the last 10 with Chesterfield County. Over the course of the pandemic and the years following, however, the applicant pool shrank to as few as 300. Today, thanks in part to AI-assisted recruiting, the department is climbing back toward pre-pandemic numbers. 

Smith was an early, informal adopter of AI tools — so early that she and a coworker gave their AI assistant, Microsoft Copilot, a name: Aiden.

"We always give credit to Aiden for the work that he does,” Smith said, but the solution to revamp recruiting started with her initiative and curiosity. Here’s how she dove headfirst into the world of AI to achieve that goal. 

The AI Use Case: Drafting and Refining Job Descriptions and Interview Tools

Smith's recruiting work wasn't purchased as a packaged AI solution. Instead, it grew from a tool already available to her, Microsoft Copilot, the platform her department uses to help optimize their work. There was no formal rollout or committee; Smith simply started experimenting and expanded her use over time as county guidance around AI caught up with practice. 

The first application of AI was for drafting and refining job descriptions. Chesterfield County shifted its hiring philosophy to a "skills-first" approach, prioritizing traits including teamwork, adventure, and mechanical aptitude over degree requirements. As such, it was important to update the existing job descriptions. To do this with AI, Smith feeds Copilot the department's mission, vision, and values, along with a job's core purpose, and asks it to check for ADA compliance and cultural alignment. She builds job descriptions in small sections rather than all at once, a habit she says cuts down on rework at the end. 

From there, Smith extended AI into the candidate evaluation process. Working from existing interview questions, rating forms, and values statements, she used Copilot to build a full scoring matrix and rubric that ties every interview question back to organizational values. Human review remains constant: Smith or her boss double-checks every AI-generated document before it's used. 

What Worked

The clearest win was consistency in the applicant review process. Chesterfield County runs roughly 300 interviews over eight weeks across three shifts, with rotating panelists. The rubric Smith built removed the guesswork from scoring applicants.  

Applicant volume also rebounded from roughly 300 candidates a few years ago to around 500 today, Smith said, with numbers leveling off. She credits the organization’s use of more personalized, persona-driven messaging over generic advertising to promote their jobs. 

"We got rid of big advertising ... I think AI is helping with our messaging, because we're taking a more personalized approach," Smith said.  

The department also used ChatGPT to script a public-safety recruitment video simulating a missing-child search, including coordinating fire, police, and 911 dispatch. Smith’s approach was to ground the project in as much detail as possible by providing a specific scenario and exact script lengths and timing to hit which, she said, made the process "flawless."

What Didn't Work

Not every attempt succeeded. Smith tried building a candidate-screening tool using AI-generated code and hit a wall. “[The tools told me] you're asking for too much ... this code is too long. Python scripting is not working for you," she said. The experience taught her to scope requests more narrowly rather than push a tool past its limits. 

Another slight roadblock has been employee resistance to AI, making adoption move slower across Smith’s department in some cases. However, Smith’s approach isn’t to heavily intervene. Some longtime staff simply opt out of using AI tools in their own work, and Smith lets them. "People who want to do it, do it. Those who don't, don't,” she said. She doesn't mandate adoption; instead, she builds trust gradually by having hesitant colleagues start with low-stakes tasks, like editing a job description, before taking on more.

Human Oversight and Governance Remain Paramount 

Chesterfield County's guardrails are practical and straightforward. Copilot is the county-sanctioned tool for work use; Smith can't run ChatGPT on her county computer, so any ChatGPT work happens on her personal machine. Staff are required to strip personally identifiable information before feeding documents into any tool. Outputs that involve AI assistance carry a disclosure that AI was used as a supporting tool.

Every AI-generated document — job descriptions, policies, interview materials — gets human review before use, a lesson reinforced when Smith's own boss ran a job description through Copilot and found it "not 100% right." For Smith, review means checking not just for factual accuracy, but for whether the output truly reflects the department's culture and values — something she says AI can miss on its own. 

Smith is direct about why oversight matters. "AI is going to use whatever information it can find, and it may or may not even align with your organizational culture ... you have to have somebody there to weigh the two."

Advice for HR Teams 

Smith's guidance for HR professionals just getting started in using AI for recruitment: 

• Start small. Begin with repetitive tasks like routine letters, minor job description updates, or a single email rather than complex, multi-step projects. 

• Feed it your culture, explicitly and repeatedly. Generic outputs happen when AI doesn't know your mission, vision, and values. 

• Be specific about what you want. "If you don't know what you want, it's like garbage in, garbage out,” she said. 

• Always verify before you send. AI-assisted work still requires a human final check, every time. 

• Disclose AI involvement. Transparency with employees and applicants builds trust rather than eroding it. 

For Smith, the underlying principle hasn't changed even as her AI use has expanded. "It is a tool. It is not a replacement for anybody,” she said.  

The relationships her team builds with candidates remain what actually drives Chesterfield County's recruiting success, Smith said. AI has just cleared the path to get there.