How Fireflies Voice Agents Automate First-Round Calls (Full Transcript)

[00:00:00] Speaker 1: And we are live. And I'm seeing already some folks joining. Welcome, Kim, Oleg, Elena, Lucas, Jeff, Robert. A lot of people joining already. Mari, Joshua, Hevchak, Aimeen, Craig. Welcome, everyone. Feel free to drop here in the chat where you are joining from. And if you know already, what you're gonna use voice agents for, or if you don't know already, what do you plan to use it for? Love to know. And we'll tailor the session based on everything that you guys send here. Hey, Christina, welcome. Jeff, Jeanette, Deborah, Carolina, welcome. Good to see you all here. We have some team members together with us here today. Craig joined from South Africa. Also, not only about if you have voice agents or if you plan to use it, but if you know about voice agents. If it is all very new and fresh for you, let us know in the chat as well. Who else is here? Harris, Ralph, Richard, Jay. Hey, Brad, welcome. Glad to see you here as well. Jade's gonna use voice agents for recruitment. This is the leading use case that I'm gonna show Jade. So glad you're here. And then we can show more later. But just gonna give another minute for more people to join and we can get them started. While that, keep dropping the chat, more about you, your use case. Hey, Eric, Manuela. Bertie, Sinha, welcome everyone. Glad to see you all here today with us. It's the first time Debra is hearing about voice agents. Prepare to be surprised then, Debra. It's pretty amazing and mind-blowing. Oh, such an interesting use case from Amin. So build profiles for artists based on strategy calls.

[00:02:29] Speaker 2: Interesting.

[00:02:32] Speaker 1: Glad to hear, Richard. As I mentioned, we're gonna leave the demo with a screening voice agent, but then we can open for any other case here as well. With that, I think we can start and already answering Jeanette's questions here. Yes, this will be recorded. We're gonna post in our resource library later today and also send via email. I'm seeing already some questions coming in. So you can drop also your questions here in the question mark icon is our Q&A feature. You can drop there and I will cover all the questions later. And yeah, so welcome everyone. This is a live of Sours. This is very different from our webinars that we're hosting here in the community. If you haven't been in one of the webinars, be sure to always follow our event space notifications. We are hosting either twice or once a month. Actually, the next one, Jeanette, is about autofill, right? The one coming up. I need to post it, but yes, we're gonna hold it later in the month. So of Sours, it's a more actionable session. So we're gonna show just a short demo and you can drop your questions and your use case, we can build it together. So not only I'm gonna tell you what is about Voice Agents, I'm gonna just quickly give an overview, but we're gonna show it, we're gonna build it together. So prepare to open Voice Agents there with you, open your Fireflies app and build together with us. But first, what it is, Voice Agents? So Voice Agents, it is your first round screen, kind of an autopilot for some meetings that it is low effort and repetitive. And by the end of this session, we'll be able to build one and share it ready for customers and our internal team members that have. I'm Pedro Hernandez, the Fireflies Community Manager, and I'm here with Chan, our Senior Product Manager that will help us with the demo.

[00:04:55] Speaker 2: Hey folks, thanks Pedro, looking forward to the demo. I can't wait to show you guys what Voice Agents can do for you. Yeah, but a quick intro is I'm Chan, I handle Voice Agents, I'm a PM here at Fireflies. And yeah, without further ado, let's get started.

[00:05:14] Speaker 1: Yep, so what we're gonna cover today, what it is a Voice Agent, we're gonna build one from scratch and then open for a Q&A, simple as that. Again, keep dropping your use cases and questions in the chat or here in the question mark icon, it is a Q&A feature, be sure to address all later. So why building Voice Agents? Why use it? So the goal of a Voice Agent, it is the first round filter that can save you time, not a replacement for your meeting and human interviews. Think about the things that you constantly have and schedule as many mentioned here, recruitment, the first sales call, so all these are Voice Agents can handle for you. And then a human can take the next steps and have a better meaning because now we have context, you can build templates and ask things that you need to when a human actually speak with another, then you have the full contents and have way better meetings in your team. So precisely, it is an AI that holds a real spoken conversation meeting on your behalf. You can build your questions, you can add your knowledge, you can set your guidelines. It talks for you, it holds the conversation naturally. You follow your questions and your knowledge. And as I mentioned, it fits those repetitive calls that you might have, qualification calls, product feedback, user testimonials, and it can enhance back with the structure. So using it also together with AI skills, Sean is gonna show you later how to do it. It can not only deliver the full meeting for you with a transcript summary, but also if you're using for sales as some mentioned here in the chat, you can get bands, you can qualify the leads. So using all the full power of that AI skill delivers to you. So using both together, it is the best scenario. And I'm gonna show later with the demo, you can also, if you don't know how and what for use voice agents for, you can also leverage ask Fred to it. So we're gonna show this later. And the payoff is clear. Few first round calls on the calendar, few repetitive meetings that you have, and all that same questions being replied without you having to do so. And then later having the structure notes that you can compare. That is plainly. So before we build one, again, if you have questions about what it is, keep dropping there, but you're gonna see soon live how it works and precisely what it is. We're gonna also share a live demo here in the chat for you to test yourself. But before that, a voice agent, it is based on three bases for the setup, its settings. You're gonna add the questions, you're gonna add a knowledge and you're gonna add instructions. So for the questions, just sharing some tips. If you wanna open already your voice agents, be sure to have your link ready for it. So let me drop here. Go ahead and open and start thinking about four to eight questions that you usually ask your customers or candidates, your leads. And always one question at a time, keep it natural, focus on what action is gonna help you take off that call. And for the knowledge and everything that I'm saying here, Chen will go into details and I will also share some templates in the chat while he shows the demo live. But for the knowledge, since we're gonna build it soon, this is just to give you some ideas. So give the agent only the information that you actually need. So when you think about, I'm gonna create one for sales. So I'm gonna add everything about my offer. No, be very precise. Add your sales pitch, add your sales guide and don't make it too broad. Be sure to be always specific. So details, must have, criteria, frequent ask questions, approve answers, boundaries and this connects with instructions. It is not only good for you to add what the agent should do, but what it should not. So Chen's gonna show how to set the tone, pacing but be sure to add in the instructions, the word rails. So what it should not say, when to recommend a human follow-up, the voicing you can add the voicings to here. I'm not able to fully answer this. A human will follow up with you later and be sure to have it mentioned the next steps after it ends the call. So just some three easy tips for us to start thinking about it. And now I will handle to Chen to make it one live here, build one live together with us. Live demoing, how to pick one, how to write the questions, based on knowledge, set the right rails, everything that we discussed it. And I seen him asking, so we're gonna show how to set the language, how to brand it, how to share. And then we're gonna drop a demo here in the chat so you can test yourself as well. And while he show each of the steps, I'll keep dropping here in the chat, some templates that you can use. But by the end of the session, I will also drop a whole playbook with all the prompts that I'm sharing in the chat. And then that you can use to build your faster. With that, Chen, thanks again for being there here with us and feel free to take over and live demo.

[00:12:04] Speaker 2: Absolutely, thank you so much, Pedro. Just a quick background before we jump right into the demo. So most of you here are Firefly's customers and over the time you've seen Firefly's join your calls and takes notes for you. A couple of months back, we also released a talk to Firefly's where you can talk to Firefly's in your meetings. So the natural course was, okay, it takes notes for you as a silent participant. Now it talks to you in your meetings. What can Firefly's do next for you? That's pretty much the ideology behind Voices is where it can take calls for you. So it's not looking to automate the rich second level calls, but the ones that can be repetitive, which is first round calls like screening, discovery, check-ins, testimonials, and stuff like that. So as part of this journey, we've interviewed a lot of customers. So if you go into any voice agent platform at the moment, the setup is pretty not straightforward. It's complex. You'll have to go through a decision tree. You'll need to walk through a pretty crazy setup. That's pretty much the TLDR of it. But what we've built here at Firefly's is we made it super simple for the non-technical folks to create voice agents, essentially commoditizing it. With that being said, I just want to jump right into the demo. I want to show you how simple it is to create a voice agent and we'll walk through all the capabilities that we have. Let me just quickly share my screen.

[00:13:30] Speaker 1: Okay, there you go.

[00:13:38] Speaker 2: Okay, so for voice agents, we have seven different templates that you can choose from that ranges across screening, discovery calls, progress check-ins for scrum teams, user testimonial agent for marketing, performance review for internal teams, user research, and customer support as well. And you can also create custom agent where you can build your agent from scratch for any purpose whatsoever you might need. Setting up a voice agent is super simple. For the sake of this demo, we will go through the recruiting use case, but I'll be happy to show you other use cases as well, but pretty much think of it this way where the creation flows pretty much the same for all voice agent templates. So screening interview agent, let's go into a template. So you'll be able to find this by selecting voice agents on your left-hand pane. And the fun fact is the moment you create your first voice agent, if you are a paid Fireflies customer, you get 100 free AI credits, which is essentially 100 minutes for you to test voice agents free. So let's go into creating a voice agent. So the setup is super simple, especially for screening agents.

[00:14:45] Speaker 1: So sorry to interrupt you, just because before we started, since more people might be interested in that. So Jeanette asking if there is a cost to build one.

[00:14:58] Speaker 2: Sorry, I didn't understand the question.

[00:15:00] Speaker 1: But if there is a cost for build one, so if people can freely build many here together with us.

[00:15:08] Speaker 2: Absolutely, yes, yes, go ahead, go ahead to build once. It's totally free. And unless in other ways someone takes a call with the voice agent, you will not be charged. And even the cost for voice agent is super affordable. It's one AI credit per minute. So like I said, right? So if you create your first voice agent, you get 100 AI credits for free. And that's like 100 minutes. So you can take up to what a lot of 100 minutes cost, I mean, a total of 100 minutes and you can just have voice agents to test and see if it fits into your workflow. So to answer your question in a nutshell, it does not cost anything to create voice agent. Just when you just take calls with your voice agent, credit starts to get consumed. Hope that answers your question. Anything else Pedro or can we get into the demo? Go ahead. Absolutely. So like I said, the screening interview agent. So the screening interview agent also has one extra feature that none of the other agents have, which is one click agent creation. But before that, let me just give you an idea on what constitutes in creating a voice agent. So there are three aspects of creating a voice agent, super simple, questions, instructions, and knowledge base. So questions, what are the questions that the voice agent needs to ask your end user? That's what you enter in parts of questions. This entry does a simple text and the voice agent understands the order and it asks the questions in a neat, natural way. Next we have knowledge. So if the end user asks the voice agent a specific question, what information does it need to refer to to answer that specific question from the end user? That's where the knowledge base comes in. So you enter information that is essentially needed for the voice agent to refer to when it's answering questions. And the best part is you're not only entering knowledge base as like pretext, you can also connect your knowledge base, which is connected to settings and you can add PDFs websites as well. And the third aspect is instructions. So every single Firefly's voice agent template and even custom agents are hard-coded to the industry best practices in terms of behavior. But let's say you have very specific instruction that it needs to follow in a specific call, you can actually add that here. So three aspects, like I said, questions, instructions, and knowledge base. And once you enter this, you're pretty much done creating a voice agent. So now let's come back to the special feature that I was talking about for recruiting. So for recruiting, you see this banner here. We have one-click agent creation flow where the moment you click this dropdown, let's say you are from a certain company. And in this case, I'm here at Firefly's. What Firefly's automatically does, it goes through your careers page. It automatically fetches the open roles as part of your careers page and list them down here. So all that you need to do is select one of these roles and click order fill. And we automatically build the questions, instructions, and knowledge base for you. So you don't need to be worried about having all this information beforehand. So the knowledge base is already filled with the company information, instructions is with the hiring best practices whatsoever. Let's say you can't find a role. So we also have other 200 inbuilt common roles out there. It's not just the open roles, but other roles as well. But even if you cannot find a specific role, let's say I enter something like agriculture or something like that, that is not part of the common roles. You can always type that and click order fill and Firefly's automatically does that for you. So in this case, let me just select one of the roles. So I'll just select product manager platforms and I click order fill. It just takes a moment for Firefly's to understand what it needs to do. And you can see that it's automatically going to fill that data. Just takes a moment. There you go. So Firefly's automatically fills this information for you. So just select the role and Firefly's fills the data and all that you need to do is just click create. The version is created. That essentially took like 10 seconds. So now we also have customization options. So you can add your own logo, give a voice agent name to it, enter your organization name. You can choose from multiple languages. We support all popular languages. You can decide the session duration that ranges from five minutes to one hour. You can choose multiple voices, pre-shipped voices that we give. And the most coolest aspect of voice agent is that we support voice cloning. So you can simply record a 30 second voice directly in Firefly's and clone your voice so you can make your voice agent truly yours. You can also choose the tone between professional, casual and talkative. Change the conversation speed from 0.7x to 1.2x and decide what's the best way to customize your voice agent going forward. So I'm just going to change my name here. And we also have multiple settings. And by the way, you can also preview the voice by testing multiple different combinations. So we also have additional settings, more advanced settings for users where you can control the visibility of the voice agent. So this is something that you're creating for your personal use or for your team use. If it's for your team use, you can enable this where other team members will also be able to see this voice agent in your Firefly's workspace and make changes to it. You can decide what medium you want to use and decide what meeting privacy will it be when the voice agent conducts a meeting. So let's say a voice agent takes a call for you. At the end of the day, it lands as a Firefly's meeting. So what should the privacy of that meeting be is something that you decide here. And also a recap email. So who gets a recap email? Is it going to be just the owner or is it going to be participants or is it someone from your team as well? And here's the most interesting aspect, right? So getting started and talking to a voice agent from an end user perspective, which we will see shortly is something that you have full control over. So you see this option. So this is where you decide what information that you want users to enter before they talk to the voice agent. So if you select name and email, they'll have to enter both their name and email, have their email verified by a one-time password and start talking to the agent. But if you don't do that, all they need to do is just enter their name and start talking to the agent. So we've learned this from a lot of customer feedback, especially from sales, where they said that a lot of customers were reluctant to give their email and that also added as a friction point for us. So which is what we want to solve by giving the option to the customers where they can choose between name and email or just the name. So I'm just going to leave it to name here. And once I hit create, I'm pretty much done. So this is pretty much your voice agent creation process. So you have multiple sharing options, which I also love to show you. So you can try your voice agent live by clicking try it live. Copy the link. So this is the link that you will share with your end users or candidates in this aspect. You can also embed your voice agent in your website, download a QR code where you can just share the QR code and people just scan it and talk to your voice agent. You can add the voice agent meeting to your calendar events from Google and Outlook calendars. And you can also share your voice agent to socials like Teams, WhatsApp, X, and LinkedIn as well. So I'm just going to copy my link. So I'll just close this. So once I am done, you can find all the voice agents that you've created in your My Voice Agents section. You can also find, when you just hover over these agents, you can also find the total number of calls completed by those agents. And you can simply manage your agents here. So you can also copy their link. You can clone their agent preview and get all the sharing options that I told you about from this page. So without further ado, let's talk to the agent and see how that sounds like. So let me select a good voice. Okay, it's correct. That's fine. So this is the agent UI. Like I said, I've only selected name as an agent owner. So I'm just going to share this URL with my users and all they need to do is just enter their name and they can talk to the agent. So they get an idea of how much, what is the duration of the call. And here is an interesting aspect, right? So if you folks remember, I created the agent in English, but as an end user, they can decide which language that they want to talk to the agent. So this is dynamic. So let's say you are a recruiter who is hiring globally or you are a sales rep looking to do discovery calls across users across the globe. When you share your agent, the agent is dynamic as in the user decides which language that they want to talk to the agent. And even if you've configured your settings in English, you've entered the instructions and questions and knowledge base in English, the agent end user will be able to change that and talk to the agent in a language they are comfortable with. So it's truly global. With that being said, I'm just going to go ahead with English right now and let me just get into the call. So it's like a waiting room where users can just configure their mic, decide what mic that they want to talk to the agent with and stuff like that. And they just click join meeting to talk to the agent.

[00:24:31] Speaker 3: Hi Chan, thanks for joining. How are you doing today?

[00:24:35] Speaker 2: I hope everybody's able to hear the agent. Pedro, can you hear the agent? Yeah, go ahead. Awesome.

[00:24:41] Speaker 3: Hey Chan, it looks like I can't hear you at the moment. Could you check if your microphone is muted and try unmuting it? There should be a microphone button on your screen you can use. Take your time.

[00:24:53] Speaker 2: Just got that started. Sorry about that. I'm back now. What's up?

[00:24:58] Speaker 3: No worries at all. Glad you're back. How's your day going so far?

[00:25:04] Speaker 2: It's going good.

[00:25:04] Speaker 3: How about you? I'm doing well. Thank you for asking. Really glad we could connect today. So just to set expectations, this will be a structured conversation where I'll ask you a few questions to understand your experience and fit for the product manager platform's role at Fireflies. There's no rush. Feel free to take a moment before answering any question. Does that sound okay?

[00:25:30] Speaker 2: Absolutely, let's go for it.

[00:25:34] Speaker 3: Great, let's dive in. So to start off, can you describe a time when you had to define a platform product strategy that balanced the needs of multiple internal product teams with the potential for external developer consumption? I'd love to hear about the vision you set, how you gained alignment across the organization, and what key challenges you had to overcome along the way.

[00:25:56] Speaker 2: Absolutely. So when I joined Fireflies, I joined Fireflies as an integrations product manager. When I joined, we had like around 40 plus integrations. But after I joined, I wanted to ramp up the developer program. We wanted more users building on top of Fireflies and not just our shipping integrations in-house. So on top of that, I wanted to expand more support. Like I said, we've made our API documentation public. We've added a very comprehensive developer program. We work with a lot of partners to do co-marketing. We reached out to them, showed value. We brought in mutual customer requests and we brought it in front of them to get them onboarded. And right now it stands at 120 integrations, which is almost more than 2X of what Fireflies was at. So it was a combination of building in-house playbooks where we were able to ship integrations faster and also other external factors like building a comprehensive developer program. So that's pretty much what we've scaled integrations going forward in a nutshell. So that automatically reflects to platforms. Does it make sense?

[00:27:07] Speaker 3: Yeah, that makes a lot of sense. Going from 40-plus integrations to 120 is a significant scale-up. And it sounds like you tackled it from both the internal efficiency side and the external developer ecosystem side simultaneously. I'm curious, when it came to aligning internal product teams around this platform vision, were there any notable tensions or pushback you had to navigate? And how did you handle that?

[00:27:36] Speaker 2: I wouldn't say there was a pushback, but it was not easy getting things done real fast. But that's usually started with clear communication and clear priorities. So I have a heartache where I don't believe that cross-functional breakdown usually happens because of disagreement. It's because people are working on different sources of information. So my job is to avoid that from happening. So once you get that sorted, get your priorities right, get your documentation right, and communication, do a lot of over-communication, that pretty much solves that piece.

[00:28:11] Speaker 3: Got it. That's a really practical perspective. Making sure everyone's working from the same source of truth rather than assuming disagreement is the root cause. That makes sense. All right, moving on. Fireflys operates in a pretty rapidly evolving AI space. How do you stay ahead of emerging technologies and incorporate them into your platform roadmap, especially when you're also dealing with technical debt and resource constraints?

[00:28:38] Speaker 2: Got it. So I have to jump. It's just a demo call, but it was a pleasure talking to you. I will see you around. Thank you so much.

[00:28:47] Speaker 3: Thanks so much, Chan. It was really nice speaking with you. Wishing you all the best, and we'll be in touch. Ending the call now. Thank you for your time, Chan.

[00:28:59] Speaker 2: So that is a voice agent call. So like I said, you can specify the session duration, but for the purpose of the demo, I had to cut it short. Also, users will also get this post-feedback screen which comes directly to us, so we keep iterating and we keep improving the voice and experience to make it more natural, ensure there's no hiccups for users. So that's there. And right after a call is done, so the first thing we do is we create a call. And right after a call is done, so what I've just shown you up to this point is just 40%. So what happens after the call is where the magic is. So here is a sample voice agent call. At the end of the day, it's an experience that all of you know about. It's a Fireflies meeting. So right after a call is done, you get your summary, you get your transcript, you can use AskRed to ask specific questions. So all you need to do is just create your voice and have it sent across to your users. They can take calls and you can just sit back and watch your summaries just land into your Fireflies account. And it doesn't end there. Again, you can use AI skills. For example, here is a sample user testimony agent. We have AI skills that almost 200 plus AI skills that ranges across various categories. So I just got a skill that is a customer satisfaction scorecard and it gives me a score on whether the customer was actually liking the product and having fun using it. So you can also use multiple AI skills that can run on demand or automated basis once you create them. So you have multiple skills that range across scales, marketing, recruiting, HR, that you can attach to your voice agent calls. So it's not just about getting the notes and your transcript. You can extract specific information. You can also create custom skills to extract skills, sorry, information that's very custom to your workflow. And of course you can use Ask Fred to decide to get even more information and just have a conversation with your transcript or your meeting. So again, the story doesn't end here. So we have a suite, like I said, of 120 integrations. And we also have a comprehensive MCP server where you can use our MCP in product to fetch information about your voice agent meetings. But that being said, like I said, we have around 120 integrations ranging across multiple categories. So our integrations work in a way where there's less HITL, which is human in the loop. So which means all they need to do is just create a configuration based on rules and it runs automatically. So let's take ATS in this case because I just showed you a recruiting voice agent. So we have multiple ATS support like Ashby, BambooHR. So let's say I create a configuration in Ashby to create new candidates or map nodes to existing candidates. It automatically fills in the scorecard by skill. So skills can be used to fill scorecards. You know, the notes can be sent using the integration. So all of this happens automatically and all that you need to do is just create a voice agent, have your integration configured and Fireflies takes care of the rest for you. So this whole experience is truly agentic, not just getting the call and viewing the transcript, but it actually creates candidates for you in your ATS, maps the nodes, fills in your scorecards. So it does 90% of the work for you. All that you need to do is just manage your agent and sit back and just watch it take the calls for you. So the same thing applies for sales and discovery as well, right? So once you have your CRM configured, it can create new leads for you. It can create new contacts, create opportunities, map the nodes to them, whatsoever. And this reflects and mirrors across multiple integration categories. So that is voice agents in a nutshell. I'd love to see you create or come up with more use cases so that we can, you know, explore those things together.

[00:32:48] Speaker 1: So Pedro, over to you. Thank you again, Chan. Thank you for the great demo. And yeah, there's not much more to say, but only then we're going to start addressing the questions. If you have any specific questions, keep dropping the Q&A. And before we start the questions, be sure to leave the session here today with one action for you to think. What is your most repetitive first round meeting call? And then try it with a voice agent. Just build one, share the link and see how it goes and let us know. But with that, let me start getting the questions here and be sure to stick around until the end of the questions. I will drop the playbook. But if you notice, if you got the notification already, there's a new post in the community with the full playbook, with all the templates and prompts that are shared in the chat. But I'll be sure to share here later. And if you scroll a little bit up here in the chat, I will send again. Here's a live example of a voice agent that we use to get customer stories. So in our blog, we have many customer stories that are using Fireflies with a lot of interesting use case. So we use that to get a template and then post it later. Chet, I see your hand raised. Go for it, Chet.

[00:34:26] Speaker 2: Absolutely. Like I said, do share, like Pedro said, do share your use cases. But what's more important to us is also what features do you like to see in the existing voice agent suite? So a couple of features that are coming is something I wanted to share, which is video support. So right now you get the audio for voice agents. We're also working on video support where you get the entire video of the entire experience that your users have to Fireflies. So as you might have seen, we are not running this on any video conferencing platform like Google Meet or Zoom. It's our very own in-house platform. So we handle the entire infra. So we are bringing in video support for all of you. So a lot of recruiters wanted to know if their candidates are taking the call. They wanted to know how their candidate's behavioral feedback is and stuff like that. So we are bringing in video support. So that is something that we will ship shortly. And we want to see more features coming in, requests coming in so that we can tackle them one by one. We have a great roadmap coming in for voice agents. Customers are loving it. They're saving a lot of money. They're saving a lot of time. But use cases are great, but we'd love to get in more feedback on what features would you like to see at voice agents as well. And if you have any questions whatsoever in terms of voice agents fitting into your workflow, you can drop them in our community or drop us an email and I'd be happy to jump in.

[00:35:50] Speaker 1: Also, actually, there is a question that is a feedback here. So Kim is asking if Recruitery will be added as an ATS integration. Do you know, Chen, or is anything planned? Recruitery, Recruitery, not sure.

[00:36:07] Speaker 2: Got it. So right now we have Ashby, BambooHR, Breezy, GEM, Greenhouse, Lever, Workable, all the top ATS integrations. But like I said, if you have any other features, especially integrations, and if you are willing to accommodate voice agents, we'd be happy to build an integration in less than a week for you. So I've taken a note of that and we can have that ship, but we'd love for you to test voice agents and see how it can fit into your workflow so we can also work on getting the integrations live for you. So in shipping integrations are not a problem. Like I said, we have a very great team back home. We can get integration done in less than a week, but we'd love to see you try voice agents. And if you are willing to go forward with voice agents as a customer, we'd be happy to ship an integration, especially ATS, in less than a week for you so that your workflow is uninterrupted.

[00:37:03] Speaker 1: Thanks, Chen. And there are three use cases that I will voice over, but then you can show the templates. So an anonymous person asked how to leverage Firefly's AI agents to handle sales and support queries. So for sales, Chen, you can show the templates that you have, but for support, a good idea is I'm using a platform and when I send a support, an AI will reply through email. So it is an AI replying to email and they even say, hey, if you want to talk to a human, you can just reply and ask to talk with the person. But you can also do this with a voice agent. So when they send a ticket, you can send an email back saying, hey, you can reply to ask to talk to a person. But if you want to talk first with our AI agent to get to know more about your case, then we can reply. We can solve it faster. You can use the voice agent for support queries to understand better what is happening. And then when a human actually reviews, it has the full context. Because we know if you work with support already, the first questions or first tickets are very basic, like, oh, my screen is not working. And then you can use the voice agent to say, hey, okay, I understand that your screen is not working, or what you're using, it is desktop, it's mobile, what's the feature, when you use it. So it is an example how to use for support. But Chen, if you want to show our sales examples that we have, and while you show that, Kin is asking, if we plan to create a conversational AI recruiter, that's precisely what voice agents are, Kin. So Chen can just show back the examples that we have for recruiting as well.

[00:38:59] Speaker 2: Okay, so I'm just adding on top of what Pedro said. So for sales, it truly depends on if your workflow is inbound or outbound. So voice agents work best for your inbound use cases, where you can add your voice agent as part of your website. So let's say you have a product, you have interested customers coming in, you can just embed the voice agent on your website, and it can take calls for you, answer your prospects' questions, a lot of use cases on top of that. And also with respect to customer support calls, like I said, the knowledge base really kicks in when it comes for customer support agents. You can add your entire knowledge base, your documentation and whatsoever, and have that attached to your customer support agent. So every time a user has an issue, you can just send your customer support agent, and it can take the call for you. So I think Kin's question was more about not a voice conversation, but a text conversation, if I'm not wrong, Kin. So right now, it's purely voice. We don't have a conversational AI recruiter in the cards at the moment, but that's something that a lot of customers have asked, to be honest. If we see more demand, we'll surely bring that into the play. But for now, it's purely voice and not conversational. Forgive me if I've understood your question wrong. Feel free to drop in a correction, and we'll be happy to answer that for you. So that being said, let me just show the sales use case. Yeah, there we go. So the voice agent creation flow is pretty simple and straightforward for recruiters, where it's just one click. But for sales use cases like a discovery call, you'll have to enter the information yourself. So we are working on bringing in more automations here, but it's just simple, even straightforward. It's just less than two minutes, even in this aspect, where you enter the questions, knowledge base, and instructions. So I'm going to create a discovery call agent here for Fireflies. So I am going to add a couple of discovery questions here. So this is specifically for Fireflies as a product. So I am going to add a knowledge base about the company. So you can see that I've added information like what the company does, what the product does, the security posture, the common objections that customers ask for, and how to deal with it, and stuff like that. So I'm just going to keep it basic, but you can obviously add your knowledge base and get that done. So I'm just going to keep the instructions normal. For demo's sake, I'm just going to add a couple of instructions. So there we go. I already have the Fireflies logo, thankfully. So I'm just going to call this chance sales agent or something like that. So here I have the instructions, questions, and knowledge base. I'm going to give it a nice different voice for variety. I'm going to keep the session duration to five minutes, so our leads are not bored. Just going to keep the tone as casual, so everybody feels comfortable talking to the agent. And that's pretty much it. So I'm going to tweak the settings a little bit. So what I'm going to do here is I'm going to send the recap email to both the owner and the participant. So even after the leads are done with the call, they get the recap email. And also the meeting privacy, I'm just going to set it to myself. So I am going to enable the visibility, so the rest of my team members can also see it. Just click create. We are pretty much sorted. I can copy the QR code and embed code, like I said, and have it embedded as part of my website. But that is pretty much it, folks. So I can just copy the link and have it sent across to my leads. I can have it embedded, use a QR code where the leads can simply scan and take it. Or let's say I have a call right now with one of my leads, but I cannot join. I can simply add the voice agent to the calendar event, and the lead can simply just talk to the voice agent going forward. So a lot of interesting aspects that you can do with voice agents, especially for sales. And right after your call is done, you get your meetings, and you can use integrations to configure your CRM. So I was just showing this in one of the calls webinars, where you can configure your CRM to create contacts for you, map nodes to companies, map nodes to opportunities, create tasks for you in your CRMs. So imagine this, you configure all these settings. Let me just, this is, skip. So let's say I want to send me a HubSpot. So you can configure multiple settings from here, which is basically a one-time configuration. You can create tasks, decide which pipeline deals that I need to send the data to. And here's the thing, right? So imagine creating a voice agent and having it embedded as part of your website. And when this option is enabled, let's say the prospect has certain action items for you. Pipelines can automatically create that action item as part of your CRM. So how easy and cool is that? So you just enable this and have your configuration saved, and you are pretty much done. So you can also get your transcripts done. Okay, I'm just going to add both summaries and transcripts. I'm just going to click Continue, click Save, and I am pretty much done. I just have to create my config, and I am sorted here. So you get your calls right away after, and your tasks and your CRM hygiene is also maintained. So that is Fireflies for Sales. But like I said, right, voice agents right now is a broad spectrum. You can also create a custom agent. A lot of users that we've seen are doing a lot of interesting stuff with voice agents. I'd love to see what you build with voice agents. A lot of users are cloning themselves. So they've done voice cloning, so you can create your own voice clone. It's just simply reading out this para and recording the voice, and Fireflies creates a clone voice for you, which you can apply to your voice agents. Have it sent across, use it as your AI assistant in case you're not able to join any meetings. So the opportunities are limitless, TLDR. I'm just going to stop here. But I'd love to see what you create with voice agents. So that is sales, and like I said, the possibilities are endless. Over to you, Pedro.

[00:45:35] Speaker 1: Yeah, so let's get some questions here. And just because this was a recurring and frequent question, just make sure everyone understands what is the cost to build, run, and how many agents they can have it.

[00:45:51] Speaker 2: Okay, so I also saw another question where they said, do we support GHL integration? So for those of you who don't know, GHL is nothing but GoHighLevel. So we do support GoHighLevel CRM. So it's here. But yeah, to answer your question, Pedro, voice agents, like I said, is powered by AI credits. So AI credits, one AI credit per minute. So you can create as many agents you want. So even if you create 100 agents, it's all going to consume from your common team pool of AI credits. So it is bundled on top of your Firefly's existing subscription. So let me show you the AI credits plans.

[00:46:37] Speaker 1: Oh, yeah, there you go.

[00:46:38] Speaker 2: Let's go to billing. So yeah, you can choose between a plan of 50 AI credits to 10,000 AI credits per month. And if you have a custom volume that you want to discuss about, our CSMs are here. And we also have a team who can get in touch with you. So it totally depends on what the total number of calls that you might have, right, per week. And we can give you an estimation on how much credits you might need to run your voice agents. So how many roles are you hiring for? How many voice agents you might need? It all depends on deciding how many credits that you might need. So it all comes to – we also have auto-upgrade, where even if you run out of credits, auto-upgrade, when it's turned on, you automatically get credits added to your account so that your workflows are not affected. So it all comes down to what your workflow demands. And like I said, we always have a plan that's right for you. But it's always going to stick between one AI credit per minute. It's not going to change. We are still evaluating the pricing for video. But as for audio, it's going to be one AI credit per minute. So 10 minutes of call with your voice agent, it's 10 AI credits. And to give you an idea, 200 AI credits are $20 per month. So you can do the math. I also have a voice agent calculator, pricing calculator, which I will share once this call ends, and Pedro can share it as part of the community.

[00:48:12] Speaker 1: Yeah, I will also add to the recap. But something that we were discussing as well, Chen, before the session is how it is effective. So it might seem that using more and spending more credits is going to cost more, but actually it's saving time. So imagine how much time you're saving and how the use is going up. So the more that they use, it is better because you're having a better adoption and you're also saving time and saving money, actually, if they use more your voice agent.

[00:48:47] Speaker 2: Let me show you something, folks. So I just found the voice agent pricing calculator. So here's the thing, right? This voice agent pricing calculator, I'm going to send it as part of this chat. And you can try to estimate the cost and the time saved as well. So here is a sample input that I have here. So how many repetitive calls that you have every week? Let's assume that I have around, let's say I'm a recruiter. I'm trying to hire here for Fireflies. Let's say I have 15 repetitive calls every week. So average duration of these calls, so what is the average? It's usually 25 minutes. So the moment you enter this data, you'll get an idea based on how many hours you save per month from voice agents. So 25 hours saved per month. And what is the AI credits per month package that you would need? That's 1,500 AI credits. So cost saved per month. So this is based on wages in the US. So it's like I've saved like $989 rather than if a hiring person would conduct the interview. And the package, the cost of the AI credit package. So like I said, we need 1,500 AI credits. The cost of AI credits package per month is nothing but $136. So use this voice agent calculator to play around, figure out what works best for your workflow. And even if it does not, we'll be able to support custom pricing and give you volume discounts if you're looking for that as well. So I hope this helps, Pedro.

[00:50:20] Speaker 1: Yeah, thanks for sharing. I'll definitely add that. And be sure to try everyone here. And two types of questions that we got frequently here in the chat and the Q&A, I briefly mentioned that. But what is based on? What is the background of voice agents? And if you plan to expand, people are asking if it's open AI, if it's cloud, or any other solution. And what do you plan to expand on? And just because it's kind of connected, if you can speak other languages other than English.

[00:50:58] Speaker 2: Okay, so we use a combination of modern providers. We use Anthropic. We use Levent Labs. We use different providers for multiple use cases. So right now, like I said, right, so it's a mix based on the type of use case that you want. But, yeah, as far as vendor, we try what works best for us in terms of the use case. So it usually ranges between Levent Labs, Anthropic, and OpenAI. So that is pretty much what powers our voice agent. And we are also evaluating other vendors constantly. Like I said, we keep evaluating the feedback that comes to us from users on a daily basis. We keep checking what we can do to improve voice and make it super human. So to give you some background, when we started this initiative back in early 2025, the latency was pretty crazy. So from what we've seen from the early stages, we've come a long way in terms of doing a lot of evals testing, you know, customer interviews and stuff like that. So right now, I believe that voice agent is in a really good space where it can conduct a conversation naturally. So that's what's top important to us. So we handle all the backend side of things for you. So our models vary. So all that matters to us is to deliver the best experience for users, and that's a variable for us.

[00:52:28] Speaker 1: This was a question that we got in the comments. But just to double down on that, how it works, the user identification when going through the voice email.

[00:52:39] Speaker 2: Yeah, like I said, that's a caveat. If you choose not to have users enter their email, it is just going to be name. So the participants have just identified based on their name. So they'll have to enter their name before talking to the agent. But if you really want to have participant information and you're very particular about that, you'll have to enable the email aspect of it.

[00:53:03] Speaker 1: Two questions are kind of similar. If it can schedule the next meeting, like fully automate a follow-up, or if it can escalate the need for a follow-up with a person afterwards?

[00:53:20] Speaker 2: Okay, so it cannot schedule a meeting, but that's something that we are working towards, which is to do post-meeting actions. But it can definitely escalate stuff. Let's say you have a Jira integration or a project management integration, like linear, and if you have Firefly create tasks for you, it can create tasks accordingly. And on top of that, it's totally agentic. So once you have those product management integrations turned on for your voice agent, it's just going to create tasks for you going forward. So that's a workaround for this. But like I said, post-meeting workflows are something that we will invest in going forward, seeing how the adoption goes for the traditional voice agent use cases.

[00:54:02] Speaker 1: And since you mentioned also voice cloning, there's a question here about to improve it with test ourselves. And I was also surprised by the results and the feedback was positive. But do we plan to add maybe more phrases or anything different to make the voice more accurate?

[00:54:22] Speaker 2: Yeah, so here's the thing, right? So at Firefly, we take privacy and security as the cornerstone of what we build anything, right? So a lot of voice agent platforms, they let users upload a few audio files and to create like the best version of the voice clone. But the problem is there is a chance for misuse. And I might take Pedro's recordings, a couple of Pedro's recordings, and create a voice clone myself, which is not ethical. So even though we have a disclaimer, I'm not sure if that would be secure for our users, which is why we wanted users to record a live voice note. They just have to read out a simple sentence and clone their voice. But I understand if there is feedback in terms of the quality, we can always look at improving the text or seeing how we can evaluate our models and make them even more better. Where the shorter the number of texts that they have to read, the better the quality is something that we try to achieve as a ratio. But like I said, right, to make it perfect, we do not want to compromise certain things in security and privacy, which is why we haven't pursued that route. But of course, we take feedback on how voice cloning can be improved, and we will do that going forward.

[00:55:47] Speaker 1: And going further with automations and more advanced use cases, can it also connect the API or NCP? What we can do after the meeting is over? Like what are the next steps that we can send externally?

[00:56:07] Speaker 2: Okay, so that is something as part of the roadmap. We don't have an API endpoint to expose voice regions at the moment. Right now, it's purely app-based. It's something that we'll have to go through. But I've taken the feedback. I'd love to know more on how do you plan to implement this in your workflow from an API perspective, and we can figure things out going from there.

[00:56:30] Speaker 1: And just one last question here that Jade sends, and it's a little bit tricky, but I just want to connect with the playbook that I'm dropping again here in the chat. On what can happen, Jade? The best case is also just reach out to our support, send the voice agent link, send the user. They can check what precisely happened. But be sure to check the instructions, the knowledge base, and everything like the settings of your voice agent. I'm sending here in the chat a playbook that we built to improve that, and maybe you can avoid with a better setup. So, Chen, go ahead as well.

[00:57:13] Speaker 2: Yeah, so most of the times, 90% of the time, it's a network issue on the user's end. If that's the case, feel free to drop us a text, and we jump in right away to give you the entire support. The idea is it's always network-based, so we have a lot of intelligent retry mechanisms where the voice agent tries to reconnect itself for a period of time before it decides, okay, this user has completely cut off and the network is bad. So all that fail-proof mechanisms are in place, fallback mechanisms are in place, but it's usually almost always because of the user's network. But if that's the case, and if that's not the case, we'd love to investigate it further. Feel free to share the meeting link, and we'll have our support team take a look at it. But this is a very spot or incidents that we haven't seen really come across often.

[00:58:13] Speaker 1: And just because we still have time, Chen, and it was a good question, how to add... Actually, the question is if we have a notion integration for the knowledge base, like for the source, but can you show quickly just how to add any live public knowledge base so we don't have to keep pasting files here?

[00:58:33] Speaker 2: Absolutely. So the Firefly knowledge base is centralized, which is... By centralized, I mean that it's not just used by voice agents, but also other aspects of Firefly, like Sales Assist, which is a platform where Sales Assist helps you in your sales calls by answering live questions where you're giving in live suggestions whatsoever. So when you click add knowledge base, you can add your knowledge base here. So you can actually do that right away from creating a new knowledge base or choosing an existing knowledge base. So click new base to add a new knowledge base so you can add multiple documents. So the way I see it, it's pretty segregated into two aspects of it, static and dynamic. So static knowledge base is pretty much your documents where it does not change after you upload it, PDF docs, sheets, CSV text and HTML. So we support all sorts of files and the dynamic ones are your webpages. So you just enter your knowledge base, like your documentation website. And also you can just enable include subpages so that it scrapes in all the subpages included from that knowledge base as well. So once it's done, it automatically gets in that data. So the version would reference that information going forward in all sorts of calls. It's not just text, it's going to be this. So I'm going to show you where knowledge base is actually located. So you see this knowledge base here as part of settings. So you can create a knowledge base here and it will be used by other aspects of the product as well. So we centralized this. So it's a piece of information that's used by other Firebase aspects as well. So voice agents is just one of it. So you can create the same knowledge base here or do it directly from voice agents, it's totally up to you. I also see someone ask, okay, they've asked if we support a notion integration. Not at the moment. We support notion as a connector for ask thread. We support notion as an integration to send notes, but we do not have support for knowledge base at the moment. We will bring that shortly. What you can do in the meanwhile is export your notion pages or just make your notion pages public and add them as websites here. So that way Firefly has access to those. So it's a workaround, but it works. Just make your notion website or page public and just add them as part of the URL and you are good to go.

[01:00:57] Speaker 1: Thanks again. Thanks so much, Jan, for all the questions answered, all the live demos and for everyone here that joined today and be sure to check our upcoming events as well. We have another one next week to show you some tips on how to use ask thread and how to get the most out of it and also check the playbook. In the playbook, you're going to find a section that if you're not sure how to leverage voice agents, you can ask thread. So there's a prompt there. Just drop that in ask thread and it can give you some ideas on what type of voice agent you can view to your company. Thanks again, Jan, and hope to see you all in the community.

[01:01:42] Speaker 2: Awesome. Thank you all. Bye.

[01:01:46] Speaker 1: See ya. Bye.