Name: AI Prompts for HR
Uploaded: 2026-09-23T17:00:44.894Z
Duration: 31 min 56 s
Description: AI Prompts for HR

Transcript for "AI Prompts for HR": Hello. Happy Wednesday. How's everybody doing? Good morning. Good morning. I'm calling in from, beautiful New York City. It is a stunning fall day. I think it's our first fall day of the season. And so excited to see so many people join today's lunch and learn. Maybe kicking off with a little bit of an introduction, on myself. Hi, everybody. My name is Gabriela Garcia. I'm our chief operating officer as well as our chief people officer, here at WARP. A little background on myself. I'm actually originally Colombian. I'm from VOTA, and then later moved and grew up in The United States. I studied mechanical engineering and computer science at MIT. I've built hardware at Apple. I've built software products at software at Facebook, and was even a product manager at Google before falling in love, with people operations and operations in general. I've always personally thought of, myself as HR and operations are really engineering problems themselves. So systems, workflows, and edge cases to be to design around, and I think that's a lot of the lens that I bring out to AI too. Thank you everyone, you know, for for attending, today's chat. I think a lot of the original ideas around creating this event was really looking at OpenAI's latest research that that released that about 10% of all HR professionals are actually using AI for their work. And the majority of that is actually not, you know, for HR workflows. And so we really want to change that, here at WARP. And so the next thirty minutes are really just to sit down today and to think about how do we develop great frameworks for writing your own prompts? How do we use a, AI in the HR and as well as like people operations, workflows and how you can get started quickly today. And so maybe some quick housekeeping items for everyone today. This session is recorded, and so we'll be sending this to you afterwards along with every prompt that we run today. In addition, we'll be also, drop you feel we'll be also sending over, you know, some additional materials so that, in with other prompts that you guys can run. And as always, you can drop questions in the q and a at at any time. I'll be answering them, more towards the end of this, this meeting. So maybe, to kick it off, I'll send it over to our moderator to kick off, our first poll. So if everybody looks up on the platform, I'd love for you to respond to that question. What do you mostly use AI for at work right now? Great. I'm seeing some responses. Seeing oh, amazing. Looks like a lot of folks are using it, for writing emails. Oh, there we go. Love to see the votes increasing. A lot of them for HR specific work. A few folks not really using AI yet. No worries. You know, that's why we're here today is is to talk through some of these use cases, and then, of course, a lot of non HR work. Great. So maybe I'll send it over, to, our team to start screen sharing some of our slides for today. Fantastic. As I mentioned, today's focus is really on how should we be designing, prompts for HR? What does a good prompt look like? And then even what are some examples, of prompts. So let's go to the next slide. I think a lot of how we designed today's presentation, was actually from asking all of you guys, what are the types of problems, what are the types of use cases, and what types of prompts are you guys really excited about that you're seeing nowadays? So as a result, today's topic and we'll be having many of these, so don't worry. Today's, meeting is gonna cover these three topics. Number one, onboarding checklists and interactive onboarding tools. Number two, how to use, AI tools, especially with sensitive, data, whether that's PII or PHI. And then number three is how can we use AI and pro and writing prompts in general to investigate, and deal with, escalations around, employer relations? Fantastic. Well, maybe to start off, and, actually, I'll go back, to the previous slide. Let me just see. Is, I personally believe, like, before you write any prompt, whether that's in Claude, if that's in Gemini, or if that's even in chat GBT, it's very important to understand the right framework of writing said prompt. Writing your prompt in the right way will end up dramatically changing the output of these, like, AI systems. And so there's four things that we should be thinking about when we write our prompts. In general, I like to think of this as an example is like write your prompt in the way that you've you'd brief a sharp hire, write a new hire on day one. They're smart, but they don't know anything about your company, your policies, or even who you're talking to. So, the four things that you need to be including on each of your prompts, and maybe I'll skip over, to one of these slides to give you guys the example is number one, context. When you're inputting a prompt, identify who you are, what's going on, any materials that you have, the more context, the better. Number two, be very specific with your ask. Don't just say help me with onboarding, clarify a very specific deliverable. Number three, if you have an example of what good looks like, give it. This can be the right format. If you don't give details on what the output is going to be, for example, a two page memo, it will write a 15 page memo. So write what's the format, what's the length, what's the audience, and what's the tone. And then finally, this is, also very important is to ensure that you have guardrails in your prompt. Right? What should the AI not do? For example, don't invent a policy. Ask me questions if you're unsure. And so as we look on the slide, I'm outputting a an example of a bad prompt and then a good prompt. Right? So prompt on the left hand side, very generic, make an onboarding checklist for new customer success manager. The AI has no idea. Right? Like, what to do from here and it's going to create a very generic response. Option two, which is on the right hand side, is really adding in these principles. Right? The context, what you need, what great looks like, and finally, what it should do and not do. Perfect. If all of that makes sense, to you guys, I think we can kick off, with one of our first use cases, here for today's talk, which is about onboarding. Onboarding is critical in the HR and the people space. I always tell, you know, like my team, a great onboarding can make or break, a hire. How do you ramp them up on day one, on what their role is, what their team is? And then of course, what is the company and what does it believe in? One thing that makes onboarding even more tricky is onboarding touches at least three different people, right? The new hire, the manager, and then the people ops team. The reason why most onboarding workflows fail, with AI prompts in general is they always ask for very generic checklists instead of very specific rule specific ones. And so today, I'm gonna walk through a couple of examples, and then we're going to explore how do we develop, how do we develop a checklist? How do we then, you know, build a tracker, for our teams? And then finally, if we have some extra time, I'm gonna show you guys, cloud design. Great. So now I'm gonna be moving over to the live demo. Great. Can everybody see my screen? Thumbs up in the chat? Great. Thanks, Alex. Thanks, Sydney, as well. So I'm going to be walking us through Claude. Claude is the AI system that we use here at Warp. The actual company is called Anthropic. So let's go back to starting with the onboarding workflow. Number one, let's think about that initial prompt that I included. Right? So as you can see, I add in the context. I'm the people ops lead at a 120 person outdoor retail company. We're onboarding a customer success manager who starts in two weeks. Right? I'm inputting their job description, but what I really want is a thirty, sixty day onboarding plan. And then specifically, as I said, let's not just get one, let's get the onboarding plans across the team. Now here I have, what does great look like? Once again, you have to be highly specific because these systems will be very generic and they'll give you way more than you need. So I want a checklist, eight to 12 items, plain language. I don't want weird terms. And then if there's anything that I have to do in those initial days, for example, getting that I nine verification, I wanna make sure that that is highlighted and bold upfront. Finally is the guardrails. Right? So don't invent anything. If you have questions, just ask me. Finally, this is the final thing, on the context, which is the job description. So one other thing that's important to note here, with these AI systems, which is really cool, is you get to change your models that you are using. You can think of these as essentially a spectrum of dumb and fast to very smart, but also is going to take a long time. In general, we're gonna be having another session on how to choose your AI model depending on your task. But for today, I'm going to choose one of the higher tier ones because this is a more complicated task. And so in this case, Opus 5.5, which actually just released last night. So I'm excited to see how it does. Great. So as we see, it's reading all of my, inputs. It's thinking. It's mapping out the due dates. It's building a new hire checklist. Great. Follow the instructions. And great. Right at the top, it identifies what are some of these assumptions. Right? Number two, it's flagged the I nine timing, and it's created this checklist. Great. If I wanted to start stop from here, what I would typically do, as someone who runs the people team is I can copy and paste this. I can also ask Claude to generate this into a Notion doc if that's what I use. If I want to input it into my HRRS system, I can ask it for the right format to then, to then upload into this. In this case, I'm gonna do something a little bit fun. So as I said, we've got the three checklists that I wanted for my new hire, for their manager. As you can see, a bunch of tooling. I identified the owners, who's on what, what's that timeline, and then finally, my checklist. Right? Confirming these, sending the welcome emails, etcetera. And then finally, it wraps up. Do you want a doc? Oops. Sorry. I don't wanna dock, personally. I want to do something a little bit more creative. And so what I'm going to start here is I'm going to then use this plan to create an interactive app for my team. And so I'm inputting in this prompt. I want an interact interactive tracker, show a progress bar at the top, include what's next, include deep include, the sections with the check boxes. And then finally, these are my decisions on the design. Let's see what it does. Great. So it's making its own web page. And so what's happening in the background and, yes, shout out to Emeka. I'd also love Claude Cowork. In another session, I'll talk about the difference between Claude chat versus Claude CoWork. In general, Claude CoWork is really great for use cases that need to you have to upload a lot of data or they need to interact with other products. We'll talk about that a little bit later, in the PII section. So for everybody today, like, what is happening in the background right now? What is happening is Claude and my AI, system is essentially building a mini app for me, a mini website, right on the spot. So I I don't need a developer. I don't need an engineer. I don't need a data analyst or even another HR tool to be creating this. And so while it's growing, it's doing some of the final things. But what's really exciting oh, no. I'm still talking, Katie. Must be your, must be your Wi Fi. So as it's working, we can also be, adding an additional prompts. Right? Like, let's say, well, this is working. I'm like, actually, I want this to be, super colorful and include my logo. Then I can go in, I can press this, upload files, photos. Maybe I wanna take a screenshot, to really make this custom to my company. I think in general, it's always important as, like, people leaders to, ensure create this, like, universe around their products, and around the onboarding. So this is taking a little bit longer than I expected, but I'm sharing I'm gonna be showing you right now, what's happening right here. So it's building the tool the web page. It's reading what are the, what's my guidance on how how to build a great app. It's actually creating the file. Great. So maybe question, for the audience while this is loading. I know I know historically, I was creating, welcome packets in Google Slides or even Google Docs. What were you guys historically using before? Perfect. As you guys enter, some of your responses in the channel, look at this output. Pretty snazzy, if I say so myself. So here's the first thirty days. It's a really fun UI. If I'm a new hire, right, I can click these things, look at the progress bar. I'm filling out this form. Let's say I just finally, finish doing all of the logins for all of my tech tools, Salesforce. I click that up. As you can see, it's now then uploading and updating, the central web page. Cool. You can play around. You can customize this. I think one thing that's really powerful with this is also showing how you can share these tools. Right? And so this is me, Gabrielle at Warp. Let's say I want to then send this over to my new hire. Let's imagine her name is Rachel. I can add her to Rachel. She can then be the owner of this doc, and we can then collaborate together and have visibility into this. Yeah. It looks like a lot of people are Google Docs, Notion users. Fabulous. Maybe going back, to this. Oops. I'm actually gonna pop up the chat, this way. Great. Here's new hire tracker. Check boxes are being saved. Let me close this a little bit on the side. Fabulous. Now I'm gonna show you guys one thing that I love to do because sometimes you want an interactive dashboard or an interactive page, and sometimes you actually just want slides, right? A lot of the people, in the channel are talking about Google Docs. I'm also a former Zugler, love Google Docs. I also love exporting this into different formats. So now I'm gonna ask the, now I'm gonna ask the AI. Actually, I changed my mind. Put this into a doc so that I can upload it to Claude Design and make something really pretty. Great. So now we've seen, the interactive dashboard. If you guys want to, like, really go hard into this, you can then publish this, and create real apps and dashboards. You can, you know, use tools like Supabase on the back end or even deploy it over to Vercel. You can have websites that say, for example, in this case, onboarding.harbor, harbor, harbor.com, and then every new hire can go into that. Cool. So now we've got the actual tracker. This is what's also really powerful with Claude that I personally love, because there's so many people across my company who prefer different formats, whether that's a Google Doc, whether that's a PDF, or even Notion for the people who are loving Notion. In this case, I'm just going to download it as a markdown. Great. For people who don't know what a markdown file is, think of it and actually, maybe if I screen share, it's gonna show it. Probably not. Think of it as essentially all all of the very trimmed down information in a Google Doc, but it's a much smaller file. And so as a result, it's way easier for these AI systems to then be, to then be processing this information. Cool. Now that I'm moving over from Claude, I'm actually going to be changing over. Oops. Your screen share stop. Apologies. Great. Here we are. So now I've moved over, into Claude Design. Personally, Claude Design is one of my favorite tools within Anthropic. As someone who is, one of the design minded people at our company, I love empowering and enabling my entire organization to create great assets, that are also, reflective of my company. Great. So now I've got the thirty day checklist, right, as we just said. And I want to now make this into a welcome guide. So here I have, my prompt that I prepared for you guys. Once again, I'll be sending this afterwards. Turn this into a one page welcome guide. I have all of the instructions. Note one thing here is this is including my design system. We can talk about this later, but what this essentially means is that all of the fonts, all of the logos as this build, are going to be coming from a central library that I have defined those instructions. Great. This might take, a little bit, to build, and so, being oh, never mind. Way faster. Fabulous. Welcome name. See you on Monday. What we can do is we can add comments into this. So let's say my new hire name is Rachel. I send it to Claude. She's gonna do it. If I want more advanced prompts, update all names to Rachel, make this very friendly with emojis. What's great, once I said is that, we can essentially have this in the background be running all of these updates, and we can add in lots of different prompts together. The only thing we have to keep on doing is just clicking this little refresh button. Great. Welcome, Rachel. Added some emojis. It then uploaded all of these. One powerful thing as well. I've got multiple people on my people team and I want them to be involved in these processes. And so the amount of collaboration that you have in these platforms, is really, really helpful. Cool. So I'm gonna stop this. And now let's walk through, to and I'll screenshot change slides. Perfect. To handling how do we handle PII? Well, this is a topic that came up in a lot of your, in a lot of your questions is how do we handle sensitive data with AI systems? I think number one, it's important to, understand that you the at the baseline, you should never be pasting names, social security numbers, medical, medical details, or anything potentially like HIPAA, HIPAA related or HIPAA covered into a consumer AI tool. If you wouldn't want to screenshot it and accidentally send it to one of your fellow employees, then you probably shouldn't be pasting it into a chatbot. This is there's two ways to be safe. Number one, is make sure to use your organization's approved AI tool. In many cases, for example, with ours, we are an enterprise customer of Anthropic. And as a result, they strip out all identifying details, and they handle, they have all of the security and data security, compliances. And as a result, I still feel pretty comfortable. If you are still a little bit more conservative, and you're still a little bit worried, always strip any identifying details. For example, like I said, a name, an SSN, or anything, and put in placeholders, for example, employee a in brackets. Great. And so now going over, you know, to what does enterprise tier mean in practice? So the business or enterprise version of whatever your company uses, for example, Claude for CoWork, ChatGPT for enterprise, Microsoft Copilot, Gemini, etcetera. These are very different products than the free app that you have on your phone. And so when you're evaluating a vendor or if you're asking your IT and security team, which I would always recommend, let me skip over to this slide, is ask your IT and security team. Is this data being used for training? How long are they retaining that data? That means how long they're keeping it in their AI systems. Have we signed any agreements? Have we signed any protections about ourselves? And then on the other end, and we'll talk about this actually in another, in in another, conversation because we've unfortunately run out of time is a lot of these AI systems have these things called connectors. I'm sure you guys have seen the MCP connector to, let's say, Slack for messaging or email or even into Notion. As a result, what that means is these AI tools are plugging into your other systems, your Slack, your Drive, your Calendar, and even, you know, like your HRS product. As a result, we not only have to think about what is the data that we're putting into these chatbots, but how are these chatbots then sending that data into the rest of these connectors? Are they approved? Is this asking for confirmation? Is this wrong or not? I'm now going to switch over, you know, back to that demo to show you an example of what good looks like, with PII. Gonna move over to my onboarding plan. As a reminder, I just hired my new CSM at my organization, and I'm going to now give an example of why it's important to double check what's being put in. Cool. So here's the example I'm gonna use. Using the onboarding plan above, write a welcome email from your manager to the new hire. Very, very common. Send it to x with these instructions. Let's see if Claude is protecting me. Perfect. Great. So as you guys can see, it pulled up this little UI, this little design in which I get to see what the email looks like. Most importantly, see how it hasn't sent it yet. And in fact, it's asked me, should I send it? Should I not? Should I copy it? Should I open an email? Should I connect it via Gmail? Once again, these are some of these connectors. Right? And so in general, I would recommend to everyone that this is an example of what a good AI setup looks like. It asks for your permission. It asks for confirmation. And then before it actually does that work. If you're in an AI platform or using these AI tools that are acting on your behalf without asking for permission, then I would say this is not safe for HR tools. Great. I'm gonna be shot stop screen, sharing. Unfortunately, we ran out of time to talk about my one of my favorite topics, which is, Claude code prompts, for employer relations and investigations. But in general, I hope that this, was really helpful. In the last couple of minutes, I would love to answer whatever questions that are top of mind for the audience. Let's see. Can we get a copy of the presentation? Yes. Josefina, I'm glad you love this. Susan. Yes. I will be fall I will be sending a follow-up with some of these prompts. Can you, Susan, great question. Can you use AI to send emails on your behalf? It's a great question. I would say this is something back to my point of asking your IT and security team their posture on this. Personally, here at Warp, we do think a lot about, because we deal with sensitive data. We are talking to customers as well as prospective customers. Is this a good reflection of us or not? As a result, I am constantly reviewing all of those outputs, but I am, in general using these clod or Chachi b t, or Chachi b t places to then send an email just the same way as you I showed that little, that little design where you could send it over via an email. Great. Roshni, how do you manage multiple hires onboarding at the same time? Fabulous question. I think we have the benefit in which, Warp one runs on Warp, our HRS and payroll platform. And so we design a lot of routines and automations. I do that myself as well as the platform has that. I think in general, it depends on the use case. How am I, for example, creating in bulk? And I think, like, these AR platforms are really good at not only creating just one onboarding workflow, but you can actually send it five, and it'll take the same amount of time to create it. Great. Let's see. What prompt do you use to make sure the required links to federal and state documents are included and kept current for a new hire checklist? Great question, Janet. I'll also be sending this, back to you. One thing that's really powerful about these AI systems is nowadays, they can actually search. And so you can ask it to look up research on what is the what's like the updates on, for example, I've now crossed 50 employees and now I need to I'm federally regulated to have a parental leave program. Right? You can essentially ask these AI systems. They'll pull the research and then just ask it, to include those links. Think of AI systems as generally having a chief of staff or a or an assistant or an additional person on your team. So the TLDR is order them. And then if the work is not good, give them that feedback. Great. I know a bunch of people are popping off and we're over time, but it was a pleasure to chat with you guys all today. My name is Gabriela Garcia. I'm the chief, people officer here at Warp. We will be doing a couple of follow-up sessions after this. I'm also very excited to hear what topics are important to you guys. In the next couple of sessions, we'll be discussing benefits and enrollments, how to design benefits, benefit benefit plan design, how to help your employees choose what type of benefit plan is a good fit for them, and then we'll also be talking about the employee relationships and, relations and investigation questions. Thank you all, today. If you have any questions, you all have my email. It was a pleasure to chat with you all today. Have a great rest of your Wednesday.