Streamlining HR tasks with Large Language Models

Helena is the Head of People Operations at the Effective Ventures Foundation.

AI is changing how we work across a wide range of sectors, and People Operations and Human Resources are no exception. For me, one game-changer has been to use a Prompt Engineering tool to automate and fine-tune several tasks.

Keep reading to find out how.

Introduction

Back in 2022, when I first started to use ChatGPT, after an initial phase of asking it to write poems about cauliflower in the style of Tupac, I, like many people, quickly saw its potential to help me become more productive. Since then, I've put it to use in several ways. It has been my thought partner, critiquing and expanding on my arguments, fixing my Excel formulas, debugging my Salesforce flows, and a huge helper in composing professional emails. Using AI has significantly helped me in refining my thought processes and with creative thinking.

ChatGPT proved to be a handy tool for one-off tasks, quick answers, or when I simply wanted to have a chat with it. However, there were times when it fell short: I wanted an easy way to categorize and organize topics, the ability to effortlessly find and reuse a particular prompt, and most importantly, the convenience to tweak specific parts of a prompt without the constant copy-pasting. Enter PROMPTMETHEUS, a prompt engineering tool, which has essentially become my Swiss Army knife for prompt editing, organization, and reusability.

I work in People Operations, also known as the fun HR (yes, I'm aware this is not the official definition), so my focus lies mostly on employee engagement and performance, and aligning workforce strategy with strategic goals. PROMPTMETHEUS has been key in assisting me to streamline operations, and I will share a few examples below.

Job description creation

When I first stepped into my People Operations role, one of my first tasks was to conduct a compensation review and to create a career progression framework. To do this effectively, we partnered with a specialist firm to help us benchmark our salaries with similar positions in the market. But we hit a roadblock - we didn't have job descriptions for many team members. These descriptions were essential to accurately represent their roles and compare them with industry standards. Plus, having a job description is a best practice as it serves as a clear guide for new and potential employees to understand their roles.

Writing a job description that is both clear and accurate can be a time-consuming process and I needed to write over thirty of them! My strategy was to draft these descriptions and then forward them to the respective team managers for review. By doing this, I was not only saving their time (as they would only need to review rather than write from scratch) but also speeding up the process in general so we could achieve the end goal (the compensation and progression framework) faster.

This is where PROMPTMETHEUS came to the rescue. Let me break down how it worked.

1. I started by creating a dataset with the titles of all the positions I needed job descriptions for.
2. Next, I wrote the prompt. This included defining the task (i.e., ‘create a job description’), defining my preferred structure and components for the description (e.g. introduction, summary, key responsibilities, required skills, ...), and any additional instructions. I made sure to incorporate the dataset with the different job titles into the prompt.
3. Then came the refining stage. I tweaked the language, added or removed details, and adjusted the tone until it matched my vision. Tip: if you're feeling stuck, you can start with a simple prompt asking for a suggestion for a job description structure. Remember, once you're happy with the prompt, you can save it for future use or modification.
4. After refining the prompt, I ran it for each of the positions in the dataset. This is really a breeze as you just select each of the dataset options directly in the prompt with a single click, or even execute them all in parallel with "batch mode".
5. Finally, I copied the output to a Google doc using the 'Copy to clipboard' icon and forwarded it to each manager. They of course needed to make some revisions and add any company-specific details to the description, but a big part of the work was already done for them.

When we need to create a new position in the organization, I can simply add a new job title to my datasets and run the prompt to generate the first draft. A 20 minute task became a 20-second task!

If you're creating a job description for a job ad (rather than for internal use), you can add additional blocks of text to the prompt with sections on company description and compensation/benefits information. Inside the prompt, there's a handy toggle option to deactivate a block if you don't need a section to appear.

The following is a quick example of a simple prompt you can use to generate a job description for a Finance Associate position.

Here’s the respective dataset:

Refining employee communications

In People Operations, communication with the employees is a constant. Be it messages about upcoming team retreats, announcements that impact the whole company, requests for engagement survey participation, updates on performance review cycles, or notes to managers, communication is a critical part of my role. The tone used in these forms of communication should ideally echo our company's culture or be an agent of cultural improvement if that is a goal. Team communications are also important for effective team coordination and to ensure everyone is aligned. But, creating the perfect message that resonates with a diverse workforce can often be daunting.

I have used PROMPTMETHEUS to refine my employee communications when the stakes are high. Whether it's an email or a Slack message, PROMPTMETHEUS helps me create communications that are clear, concise, and in sync with our team's communication style.

One of the major benefits I've experienced is its ability to use past messages I've crafted before in the tool as a template to develop new communications. It's like having a personal assistant who knows your style and your audience and creates messages that make sense for the situation.

Another feature that I like to use is the "Temperature" adjustment. This control lets you tune the randomness or predictability of the AI's outputs. A lower temperature means more predictable, conservative results, so the AI adheres closely to patterns it's learned from its training data and the prompt. On the other hand, a higher temperature boosts creativity and randomness, leading the AI to produce more varied and unexpected suggestions.

Below is an example of asking the application to generate a message to the team based on a previous one. On the completions section (right side), you can see on top the output using a temperature of 0.2. This is a lower temperature so the AI is sticking more closely to the structure and style of my provided example, prioritizing accuracy and relevance. The completion below uses a temperature of 1.1, which is high. Here, the AI is diverging a bit more from the exact pattern of the example as it's encouraged to be more creative and to explore less probable options. While in this case the more creative option is not exactly what I’d be after, for other communication needs, this creativity could come in handy.

As with the job descriptions, I've found its ability to save prompts for future use very useful. This means that I can create a library of messages that I can easily access and modify as needed. This has saved me the time that I would have otherwise spent writing messages from scratch or rebuilding prompts on ChatGPT.

Idea generator for retreat team building activities

Planning a team retreat can be a huge opportunity for People Operations to make an impact. A successful retreat strengthens team bonds (especially for remote teams like mine), aligns everyone on strategy and goals, sparks laughter, and creates memories that foster a more productive and engaging work environment. Yet, brainstorming new and engaging activities for each retreat can be a challenge, so it’s very useful to have the tool helping me generate a variety of engaging team-building activities.

With PROMPTMETHEUS, I can ask the LLM for team-building activities that, for example, can be conducted outdoors, involve elements of problem-solving, or cater to a diverse range of interests. The tool then returns a list of unique ideas. I select some of these ideas, place them in a dataset, and generate a new prompt that provides more detail on how each of these activities could be executed.

What I find truly valuable about PROMPTMETHEUS is the ease with which I can modify the generated ideas. If an idea seems interesting but needs a bit of tweaking, I can quickly edit parts of the prompt and re-run it to get a refined result. This iterative process helps me to come up with activities that are fun, meaningful, and customized to our team's specific needs.

PROMPTMETHEUS also proves to be an asset when it comes to designing the retreat schedule. I feed the tool with broad descriptions of the activities, meals, breaks, and discussion sessions, and it delivers an organized, logical, and balanced schedule. It takes into account factors like the duration of activities, rest periods, and the best times for in-depth discussions. This way, I can ensure that our retreat is enjoyable, well-structured and beneficial for all.

Drafting procedures

In the world of HR, clarity is essential - especially when it comes to procedures. From hiring to performance reviews, the steps we outline for these processes can impact their effectiveness. A well-drafted procedure can minimize confusion, reduce errors, and ensure consistency across the board. However, creating these procedures can be a painful task, often involving numerous revisions, restructuring, and the constant need for updates.

Here the beauty of using PROMPTMETHEUS lies in its ability to assist in drafting clear, concise, and comprehensive procedures. It's not a magic wand that will conjure up perfect procedures out of thin air, but it certainly makes the process less tedious and more efficient.

Let's take the example of a procedure for running a performance review cycle. When working on this, I had a basic structure in mind. I knew the key stages – preparation, self-evaluation, manager evaluation, calibration, and feedback sessions. But the challenge was to develop each stage clearly and concisely, and ensure the procedure was easy to follow and left no room for ambiguity. This is where LLMs shine.

I started by inputting my basic structure into PROMPTMETHEUS and prompted it to expand on each stage and consider aspects I might have missed. For example, when explaining the 'self-evaluation' stage, it suggested including guidelines on what employees should concentrate on during this process, and how they can use this exercise for their professional development.

As I continued to work on the procedure, I could refine the prompts, making them more specific and detailed. It also saved the data, enabling me to easily modify and tweak the procedure in the future.

Once the procedure was drafted, I used the tool to format it, making it easy to read and understand, and making the process of drafting, refining, and formatting the procedure much more manageable.

I would like to stress that while PROMPTMETHEUS is a powerful tool that takes advantage of the capabilities of generative AI, it doesn't replace the need for human intellect. It doesn't replace the need to understand the process, to think through the steps, or to consider the needs and perspectives of the people who will be using the procedure. However, it does make the task of drafting procedures less intimidating and far more efficient.

Other use cases I haven’t explored yet

While I've been making good use of PROMPTMETHEUS for a lot of HR tasks, like generating job descriptions, refining communications, and even ideating team-building activities, there's an exciting bunch of other potential applications I haven't yet explored.

Take the onboarding process, for example. It's a super important step for any new hire, but it can be a bit of a drag with all the paperwork and sometimes not-so-exciting introductions to the team and company. I imagine PROMPTMETHEUS could really shake things up here, helping HR teams to craft onboarding experiences that are not just efficient, but also engaging. I can picture PROMPTMETHEUS helping to produce a template that strikes the perfect balance between ticking off necessary tasks and creating opportunities for new hires to connect with their team and to get comfortable in their new workplace. The plus here is the ability to keep data organized, as we could keep tweaking these onboarding procedures based on feedback and changing needs.

Another idea is to create original training materials. This can be a time-consuming process, especially when you are trying to cover a variety of roles and skills within an organization. With the help of PROMPTMETHEUS, I believe that we can significantly speed up the process. As an initial thought, I could give PROMPTMETHEUS a brief overview of the role (even using the job descriptions that I've already created with the other prompt!) and the broad skill areas needed, and it could generate a comprehensive training module tailored to that specific role, which could include case studies, interactive quizzes, and more.

Moreover, as the workforce strategy evolves, we'll likely need to continually develop and improve upon our current HR and People practices. PROMPTMETHEUS could be our assistant, for instance in brainstorming new employee engagement strategies, coming up with performance improvement initiatives, or finding creative solutions to improve work-life balance.

While I've already found some ways to use PROMPTMETHEUS in simplifying and automating many of my tasks, there's still a lot more to explore. I’m excited to continue exploring the capabilities of the tool and I’m pretty sure it's going to become an even more important part of my HR toolkit.

A word on sensitive and confidential data

In the midst of our excitement about the transformative potential of LLMs and tools like PROMPTMETHEUS, it's important to pause and consider the implications for data security. HR tasks often involve sensitive and confidential information, so I want to emphasize the importance of being careful when handling such data.

When it comes to LLMs, it's important to remember that the data you feed into the system might be used to improve their performance. So, you need to use it carefully: don't send sensitive or confidential data to LLMs. This includes personal details about employees, proprietary company data, or any other information that you wouldn't want to be shared or analyzed outside of your organization.

AI tools are here to assist us, but do not (at least yet) replace the human judgment and discretion that are so crucial in HR and People Operations. So, while PROMPTMETHEUS can be a fantastic tool making HR tasks more efficient, it's up to us to use AI responsibly and ethically.

Conclusion

As we look at the fast-paced changes happening in HR, it's clear that tools like PROMPTMETHEUS are a glimpse into the future of our work. They challenge us to re-think our roles as facilitators of human capital. I’d encourage people working in People areas to dive in, explore the possibilities, and share their findings. Hopefully we can learn from each other, and evolve together to make the most of AI in our careers.

PROMPTMETHEUS isn't just a tool, it's a partner that helps us streamline our processes, refine our communications, and ultimately, create a more productive and iterative work environment. So, let's embrace this change, because the future of HR is already here, being driven by AI.

Thanks for reading!

If you want to explore how you can use AI in your business for HR and People Operations, don't hesitate to get in touch with Helena on LinkedIn.