Customer Agent @ the HubSpot Admin HUG 🐙

Welcome everybody. Thanks for being here. Thanks for bearing with us through this ongoing platform change. Sharing my screen. Here we go. This is another event of the HubSpot admins hug. This is our homepage. Please be aware of it, bookmark it, keep an eye on it. We will be scheduling quite a few events in short order now that some of the dust is settling. But clearly not all the dust is settling. You are here. We are talking about customer agent. Super excited to dive deep into that. Next week we are having a Show and Tell. So we will definitely have our technological decks in a row by then or else that's going to be a real challenge. And then in March we're talking about lead scoring and reporting and and other events coming soon. So keep an eye out. We also have a YouTube page so if you ever wanna. If you ever miss an event, if you ever can't log into an event, if you want to share an event with someone. If you wanna rewatch an event you Already saw this YouTube channel is your source of truth for that. Currently a nice tidy 200 video is in there so enjoy it. Love it. And this one, this session today will be there soon. Ish. We gotta figure out how that's gonna work and then. But it is being recorded again. We are having a show and tell coming soon. We would love to have these more frequently than we have been. Every time we do a show and tell it is an incredible event. You all are working on great things. You're all smart people. You're all doing stuff in HubSpot. If you ever do anything you are the slightest bit proud of or you think is interesting or you overcame a problem or you fix something please submit this form let us know. We would love to to have you present at the show and tell sometime. Once we have four or five speakers we will. We will schedule another one of those. We have a space on community.HubSpot.com it's great. Should check it out. We have a group on LinkedIn and also like a channel a handle on LinkedIn. So if you're on LinkedIn that's where a lot of the action happens. Connect with me Indiana. We we talk about HubSpot stuff all the time. Boot camps probably all filled up now right? You could do the Connect call Boot Camp or the Discovery call Boot Camp. CRM customization for developers. Boot Camp Rev Ops is still open. Okay. I would actually highly recommend that one to you all admin types and video strategy. That's cool. So check out the boot camps. They're free. They're six week cohort based education things. Love them. And again, we are here every Tuesday. I don't know if I remember to say that I'm a little flustered, but we are here every Tuesday at this time. And if there's ever a HubSpot topic or speaker or something you would like to see covered or done or addressed here at the admin hug, feel free to fill out this form or again, connect with me. Connect with Deanna. We love to hear from you. We love to hear your ideas. We want to make sure this is as relevant and valuable as possible to you. And I know I went fast and there's more I could have said, but all this stuff is on adminhug.info. so check it out. Deanna maintains that list and it's, it's lovely. So that's enough of that. Thank you so much to the 128 of you who have made it in here and I apologize that the chat is not working the way we want it to. And I will continue working on that in the background. But for now, Vijay, take it away. Tell us some happy things about Customer Agent.

Thanks, Kyle. Let me get scarce. I'll share my screen. Make sure to share the audio here. Hey everyone. I hope you all can hear me. Kyle, can you hear me? Can you give me a thumbs up? Awesome. Excited to be chatting with you all. So we have the Customer Agent product team here. My name is Vijay. I lead the Customer Agent product team. I have Bhargav and Kriti on the on the Customer Agent product team as well today joining to talk about all things Customer Agent. First of all, thank you for taking the time to come hang out with us. We hope to make this time useful for you. And this is an agenda. We'll start with a brief intro of Customer Agent for those of you who don't know, give you a sense of what the agent is about. But we'll spend most of our time actually walking through a live product demo. Between me, Vargo and Kriti, we're going to walk you through the entire product demo, including all the new stuff we've sort of recently launched and also later on give you a sneak peek of what's coming next. And then if you all have questions, feel free to sort of throw them in the chat. We'll try and monitor them and we'll have about 10, 15 minutes at the end to actually answer all the questions you have. So with that, let's get started. So what is Customer Agent the way we Think about customer agent is there's one single unified agent that helps businesses generate leads and scale support powered by the HubSpot customer platform with all the shared context and data. And we are really focused on solving for two core use cases. We call them like scale support and generate leads. When it comes to scale support, the sort of the core value prop is how you can scale support without scaling headcount. And this is something which is the most popular use case today where the agent is fully automating all tier one support. Right. Really freeing up support teams do more high value tasks. And this is not a thing on the slide. This is actually happening in the real world and you hear from the customers later on and the agent is available 24 by 7 across all the channels where your customers are so really meeting customers where you are and really deliver an experience that is very delightful and highly personalized. Again, leveraging the power of the customer platform. And the most important thing is you can get started in minutes. It just takes a few steps to get started. These agents are so easy to get started and get them to train and get them to answer questions compared to the historical chatbots which were very tedious and hard to set up. The second use case we sort of solve for is the generate lead use case which is currently in a private beta is where we help convert more visitors into qualified pipeline. This directly fits into our AEO strategy where we are seeing more bottom of the funnel, middle of the funnel being more important than top of the funnel. And that's where the customer agent will come in, where it'll start to engage with your customers, answer questions, but also sort of take them through a qualification process and help them sort of turn your visitors into pipeline. And being able to deliver this across channels, be it chat, email, SMS, WhatsApp, et cetera and really helping you actually have more high quality conversions. And that's where we sort of, we want to evolve this product into so with that. So these are code use cases and over the course of the demo you'll start to see how we'll sort of talk about this. So let's maybe talk about a bit of our customers. The agent has been available for a couple of years. We have thousands of customers use this on a, on a weekly basis and we have hundreds of customers who love the product. And I don't say this lightly, we usually have another customer for whom the agent today is their best supporter. Right. Because the agent is solving hundreds of queries, resolving a big part of their ticket pipeline so they can really Focus on things that really they care about, really focus on these high value, high value tasks and really starting to see sort of behavior of customers who are using the agent to really change as well. So let's actually just to bring this to life, hear from a customer who's been using the agent for a while and like a one minute clip I want to play.

I am of the idea that customer support no longer exists. AI took over. Whatever AI on customer support cannot take should automatically be passed to customer success. I don't see how customer support teams should still be considered a valuable team in a company. I think of all the AIs even over intercom by the way. So if you want to tell this to your people, I believe after testing potentially seven different customer support AIs Brisey is potentially the best by far. We have around 100 customers right now and there's only one customer success person and I intend for this team not to grow more. We have basically delegated, I would say 95% of tickets to Brissy and we read all of them to make sure that everything is on point. And after what, two months of reading every single ticket it is 95 accurate.

Right?

So it's very on point

easy. Here for folks is the name of the customer agent and we are starting to see this not just in customers telling us but also in our own sort of tracking where we track resolution rates as a sort of measure of quality. And over last year we went from a low 20% resolution rate to right now I think we are close to 70% resolution rate which means for a customer using the agent, almost 70% of their issues are resolved without a human being in the loop where the agent is completely sort of answering the questions. And this is game changing for so many support teams. Both support teams who don't really have a lot of resources or support teams who have a lot of resources but then now they can deploy the resources in different ways. We have teams actively doing things like account management and more customer success sorts of use cases. So that's a quick TLDR on what the customer agent is and how customers are using it today and the value they're already deriving it from it. So with that let's actually dive into a demo. I'm going to hand it over to Barga who's going to talk about the setup experience and how you can train the agent to really suit the needs of the business. Margav, take it away.

Yeah.

Thanks Ujay.

Hey.

Hi everyone. Very excited to show you how to set up the customer agent and train it. So first of all, where do you find the customer agent? You just have to go to service and click on Customer Agent and you will land on this page. This page just gives you an overview of the features and some customer testimonials, how they're using it and just some basic ideas of what the customer agent can do for you. There's also a video here that you can watch. But yeah, you just have to click on set up your agent, give your agent a name. I'm going to call this one Peter Rabbit and select a personality Friendly. And then, yeah, so you can, at this point you can choose to add some knowledge or content to the agent which it can use to answer questions. And so something like typically a customer would do is go to their website. Let's assume that we're building a website for this Peter Rabbit world and I just add the URL here. Click on import related URLs which means that all the sub URLs on that web page will also get added to this page. Up to 5,000 pages. You click on Create Agent and that's it. Your agent is now set up. You still need to deploy it, you need to add more stuff to it, but is like good to go in terms of just testing the really basic functionality. So for example, if I go here, I can see that I can already test the content that I added to the agent, but I'm going to talk more about how you can train it further. So for that I have set up another. I'm going to go to this portal which already has a lot of these things set up. So yeah, so after you've set up your agent, like after you finish this step, you will as you start using the agent, so you will have this get started guide where you can set these things up. You can test your agent, set up, handoff, deploy your agent and Vijay and Kriti are going to talk about some of these things. But once you've started using the agent a bit, you will see that the page changes and you start seeing, you know, different call to actions and different metrics and stuff like that. But for now we're going to focus on the training bit. So when you click on Manage, this is where the bulk of the features are. The first section is Identity, where we've already given the agent a name, a personality. Right. If you're using Voice, then you can also select the voice here. But we're not going to focus on that for this demo. The second feature here, which is a new one is called guidelines is a way for you to have more control over the agent's behavior in terms of the tone it should use. So for example, I've added some instructions here about the kind of language it should use, the tone it should use. I've said warm, confident tone, whatever is appropriate for your brand. Basically, I'm going to. So this is in a published state right now, so I'm going to go back to the draft state to show you like some of the features here. You can also give it specific instructions about the response style. Maybe you prefer the agent writing in paragraphs or in bullet points, or you want the agent to give detailed answers or concise answers. So they're already like. They're also like templates that you can use. So I've selected the concise template here, but I'm going to. Let's say you want a more balanced approach. You don't want the answers to be too short. You can just select that and. And the instructions are already updated. You can also give the agent scripted responses. So these are for scenarios where you might not want the agent to draft its own response. So think of anything like for example, a greeting. You always want the agent to greet a certain way or close the conversation a certain way. So you can add all of that here again you have templates that you can use greeting, closing. Or if you have irate customers, like what should the agent do? Maybe it should use a specific language to talk to them. Or when it's talking about your policies, it can again use specific language that you ask it. So it's going to use it verbatim, it's not going to draft its own response. And you can also add guardrails. So again you have templates for these. I've added both of these here. So topic deflection. For example, if the visitor brings up any of these topics like pricing, legal, medical, financial advice, I don't want the agent to answer these questions, so I'm going to hand it off. To basically avoid these conversations. The agent is going to avoid these conversations and you can also set up a handoff which we'll talk about later. You can also write custom instructions here for anything that does not fit into those boxes. So for example, I've written if the customer is based in Germany, use a formal tone. If they're in Spain, use a casual tone. So another cool feature is that when you click on Optimize, we tell you if the we'll firstly optimize the instruction and if it's an instruction that we do not support, we actually tell you that we're not going to support it. So instruction removed because it relies on customer geographic location which the agent cannot access or use. So this is a, like, this is a great way for you to validate whatever you're entering is actually useful. So once you've added all these instructions, you can just hit publish and it'll build one of the check here if there are any issues, they'll be reported and you can just hit publish. So now your instructions are live. If you want to go back and edit them, you can just go back to the draft mode and keep, you know, make as many edits as you as you want. So the other feature I want to talk about is adding knowledge. So this is the core knowledge that the agent is going to use to answer all your questions. And there are lots of ways in which you can add knowledge to the agent. So when you click on this Add content button, you will see that you can add any of your HubSpot content, your knowledge base, website, landing page, etc. You can also upload files. A cool feature that we recently launched is that the customer agent can now also read structured information. So if you upload a CSV or an Excel sheet which has all your product skus, it's different attributes information about all of that. The agent can read through that information and answer questions about your products and services. There are also other ways. I showed you this when we set up the agent, you can add a URL and crawl all the associated pages. You can also create short answers where you can just type questions and answers. These are all the different ways in which you can add knowledge. But going back to Vijay's point about how this agent uses the Context in the HubSpot context and the HubSpot platform to personalize the experiences for your customers. You can also give the agent access to your CRM data. So for example, I've given the agent access to all these different properties. And when you give the agent access to these properties, you can also give it instructions about like how to use this property. You can give it view access or edit access. If it's view access, it'll just, you know, view the properties. If it's edit, then it can also go and update those properties. And the last thing you would want to train the agent on is actions, which is a way for the agent to actually go and do something in a system that you've connected to the agent. So think of it as making API calls. So if you have an order management system where all your orders are stored. The agent can go into that order management system and fetch the status of an order to answer an order related query. So just to give an example, I've set up this order status action and it's pretty simple to set this up. You just give this a name, you can write a description which the agent will use to understand what this action does. Then there are these trigger phrases where the agent will look for in a conversation to know that when to use this action. For this action, I want the agent to ask for two inputs, because these are the two inputs I need to send to the API. So I've asked the agent to ask for order number, email address, and then I've configured the API here which has all my order information. And then lastly, I've given the agent some instructions on how to use the API response, because the API response can have a lot of things. So this action is already published. I quickly want to show you what a conversation like this looks like. So this is Help desk where I already had a conversation on the Zenplant website for this demo. So that conversation is basically in this inbox here. So here's a customer who's saying, I haven't received my order. Could you please tell me the eta? The agent asks the customer for the email address and their order number, because those are the two things that we configured here. And then the customer shares the email address and the order number. And yeah, the agent basically hits the API and gets that ETA from there and answers the question. So that's how actions work. So one thing you might have noticed here is that there are two things here like resolve issues and generate leads. That's because the agent, we support both the use cases. So this is a front office agent that can handle any kind of customer facing conversation. And most of those conversations fall into one of these buckets. So Generate Leads is where you can configure the agent to ask qualifying questions and generate leads for you. And so we give you this action called Lead Qualification. This comes out of the box and all you need to do is so just configure this. This is a way for you to tell the agent what your qualification criteria is and what the agent needs to do when someone meets that criteria or partially meets that criteria or doesn't meet their criteria. So in this case I've added, so this is a hypothetical business called Zenplant where, you know, we were selling plant care services. So in the qualification criteria I've said if the subscription type, if someone is interested in the professional subscription and they're based in Boston and the kind of plant care they're looking for is indoor and outdoor. Then, yeah, if someone meets all of that criteria, then I want the agent to book a meeting with them and I've told it which meeting link to share. So this link comes from your calendar, your meeting scheduler, so this is like one of the meeting links there. You can also set up a round robin on this calendar so that the meeting gets assigned to your different agents. For partially qualified, I want the agent to not offer a meeting, but update the lifecycle stage in the CRM to marketing qualified lead. And if someone is not qualified, I just want the agent to close the conversation. So I'm going to hit publish. And now the agent is equipped to handle any conversations from prospects. So I'm going to quickly switch over to this inbox, which is what I've set up for my lead generation conversations. So here's a customer, he's saying, I'm looking for a plant care service for my office building. And so the agent at this point already realizes that this customer is interested in buying services. So it detects that intent to buy and has therefore immediately gone into the lead qualification mode and is asking for their email. So once they share the email, the agent now asks for the three qualifying questions, the city they're in, the kind of care they're looking for and what kind of subscription they're interested in. The visitor or the customer gives shares the information. And so this person is in Boston looking for Indocare and wants the professional subscription. So they're qualified. So that's why the agent has now offered to book offer them a meeting. They can click on this link and then book a meeting with the sales rep. So that's how lead qualification works. So yeah. So if you. So just to do a quick recap, you set up your agent, give it a name, personality, give it some instructions about, you know, your business, what kind of language it should use, set the guardrails, add knowledge to it in many different ways and then just set up actions. If you, if you have a need for actions and if you want it to generate leads, then just set up the lead qualification action as well. So once this is done, your next step is to test it, optimize the setup. And I'm going to hand over to Kriti to talk about that step.

Thanks, Bhargav. All right, let me share my screen.

Cool.

All right, awesome. So I'm going to be walking you through testing and optimizing for the agent. So what testing allows you to do is to spot any issues potentially before customers do and try different approaches on what's better content or potential triggers for human handoff. It's really a way to help you build confidence that your agent will be able to handle things that the way you want it to. So this is an example of live chat where you can make sure that live chat is clicked and I'm in live mode right now. So you're actually seeing how your end visitors would truly see the live conversations. So I've just asked it a few questions of do you have any roses in stock? It gave me a response to clarify what email address and it created that with created a conversation associated with that email as a record and then you'll see the answer and responses back and on the right what you're seeing is for each response just generated what was the sources that were used and if there's a CRM record created for your visitor it will also include that as Barga also mentioned in the properties and yep. So this is just an example of live chat. I'm going to actually also show you email. Just one second. So email is email testing. What is really important about that is you don't actually need to be, you don't need to have an email channel that's actually connected in order to test it with the tester. You can preview the full email conversations from the customer's initial question through your agent's response as it would appear in the real inbox. And why this matters is because customers typically expect more thorough, well written responses and you don't really get the chance to ask as many follow up questions in real time with email. So prior to email testing you only had the ability to turn on email and see how things will go or you can just test in chat. So now you can make sure that the agent will get it right, whether that's the tone, whether that's the completeness of your answers or formatting and that that's more important since email is more permanent. So you can navigate to tester in any of your tabs. I can show you in a second. But what you're seeing here is the responses in an email format that our agent gave back. Same thing with clarifying the city. I did that with Boston, I did that for indoor arrangements and then the type of subscription I'm interested in starter and then it gave a response back here when what you also can do is if you didn't like that answer, you can click connect me to Human agent. But on the right what you're seeing is all of the insights that are being used to generate. So here you'll see all of the sources used to give those responses and it's also showing you lead qualifications. So here it's saying what the agent qualified based on the criteria. You can also improve your responses by clicking improve response. And here what it'll let you do is to either create a new short answer which is a new source that the agent will use in the future. You can add a handoff trigger, add action or manage your sources directly, or write a new KB article which is super cool because while you're testing, you're able to directly train and improve the agent's performance. Okay, I will continue now to show you knowledge gaps. Okay, cool. I want to talk a little bit about the coaching experience that shifts more to closing knowledge gaps. So here in the knowledge gaps reports you can see the topics where the agent struggled and did not know the answer to the end visitor conversations. And it in this case we have geography which was a topic that the agent couldn't answer in the last 30 days. So if I click into that and I drill into the specific question, it's what is the capital of Istanbul? The agent didn't know the answer before and it actually handed it off to a human. And a human agent has given the answer. I'll show you what that looks like in a second. So if I click here, you can see all of the conversation and this response was me as the human agent on what the answer was. And what you see here is that the agent learned from that answer and recommended it as a short answer and says, hey, there's something you might want to add as your knowledge. So I created it as a short answer. In this case the answer is Ankara and I can create and resolve that. So this is a really powerful experience that shows how AI and humans can work to get to work together to get better. So if you are in help desk, you can see those actual conversations or inbox and you can respond by as a human agent. And that will become a recommended AI answer here for you to resolve. The next thing I wanted to talk about was source performance. Source performance is a new capability we just launched that basically allows you to see what content is helping or hurting your agent's performance and you have more visibility into what's driving successful outcomes. What we've done is connected agent outcomes with resolutions and the conversation volume of your end visitors to direct specific knowledge sources that are driving them. So you can see Every article file doc that your agent is referencing along with how each of them are actually performing. And we are classifying these into four categories. You have top performers which are the sources that are consistently driving successful outcomes and you want to keep them updated. We also have underperformers which are frequently used content but that doesn't have the best frequently used content but with poor outcomes that needs potential editing and then underutilized bucket which is valuable content that needs better surfacing or better editing. The actual source and low impact are sources that are not contributing much and might need to be candidates for removal. So you can access this from the Performance tab and it's right here the new sources you don't need any other information other than having your knowledge sources already configured. And what you'll see is in the key metrics which are times referenced in the conversation along with the resolution rate of when that sources were used. Up here in the top buckets and then on the bottom. What you can do is click into this actual source. If I click this, it'll take me to the Sources content sources tab that Bhargav was showing with that source and you're able to click that and edit it directly. Just going to go back to show you more things. You can also click View conversations and see all of the conversations that that in that bucket that the source was leveraged in. And you can click delete or update as I mentioned. And the other thing is this just helps you to understand which source was being used correctly and are customers satisfied with the answers. And everything is kind of more actionable for each single tab as well. Like I mentioned, if I click here I can go and edit that and then for the low impact it's to delete. I'll pass it now to Vijay.

Thanks Kriti. Let me share my screen again. Perfect. Cool. So now that spoke about how do you sort of train the agent and how do you sort of test the agent and evaluate. So I'm going to be focused on talking about how do you deploy the agent, how do you get this agent which is now trained and tested in front of your customers. Right. And we recently launched a bunch of new functionality that I want to sort of talk to you about. The first one is what we're calling like on ramps or deployment options on how do you get this from our customers. I'll also talk about updates we've made to human handoffs. I've seen that there are a bunch of questions around. Okay, what happens when you hand off to human? What controls do you have. I'll talk about that and also talk about some settings we've recently launched so that the agent worked really well on these channels. So in terms of channels itself, so today the agent works on pretty much all channels on chat, which is it can be a website chat, mobile chat, can be a chat that's embedded inside like a product. It works on all those channels. It works on social channels like WhatsApp, Facebook messenger, it works on email. And we also support this channel on a bunch of third party apps which use the custom channels, including Instagram, TikTok, Slack, etc. Etc. The agent works on all those channels. And finally, we also support Voice, which is currently in a private beta. Right. So if you're interested, you have to sort of request for it. So those are all the different channels you can deploy the agent. But I think the thing we've learned over the course of the last few months is I think deploying the agent right now in a lot of these channels is like all or nothing. Which means you either make that agent available to all your customers or you don't make make it, or you don't deploy the agent at all. So I think a lot of the work you've done is to give you more optionality to be able to do that. So the first feature I want to quickly talk about is what we're calling working hours, right? Where you can say you want the agent to work in a specific times of the day, right? So earlier it is all hours, it's available 24 by 7. But a lot of customers want this available only during certain types of the day when, maybe when the team is not available during offline hours, right. Or specific times only when the team is available. So this gives you a lot more control over when the agent is responding to those, to these questions and when the agent is not, then all these are sort of automatically assigned to a human agent, right? And this is a really powerful way for you to get started where you can say, okay, make the agent available only during offline hours when you typically don't get a lot of traffic. So it's a great way for you to test out the agent and then sort of deploy it. And then as you build confidence and as you sort of continue to sort of improve the agent performance, et cetera, based on what Kriti said, then you can make this available to a broader set of audience. So that's the first one, Right. I think the other thing is we've integrated this agent with workflows, right? And this is Actually really powerful. I'll show you a couple of examples here. I have a workflow setup which says based on the type of customer who's reaching out, you can decide which type of customers the agent is handling, which type of customers your human team is handling. In this case I have a branching logic based on the type of customer. Have a property which says type of customer. And then for all my free customers, I'm saying assign that to a customer agent. Right? But for all my VIP customers, these are like high value customers. I want a human to answer them. And then you can use workflows in like really, really powerful ways. This is one example. And the other example is like specific topics. I think there's someone on the chat who asked things like password resets, right? Or maybe certain types of queries. Here I said based on like a ticket description. Or you can also use things like ticket categories. Say assign those to specific things to a customer and others to a human. In this case, I said anything that is to do with plan care, servicing, pricing, etc signed to a customer agent. But more sort of risky questions around billing or legal etc. Talk direct them to a human. So I think this really gives you a powerful way to use workflows to orchestrate the types of conversations you want the customer agent to answer and the types of conversations you want the human to answer. And you can do a mix and match of these different criteria. This again is one of those things where you'll get more control over when the agent should answer, what time of the day using working hours. Right. And what type of customers or what type of queries that the customer should sort of customer agent should answer. I think what we're going to evolve this experience into over the course of next, next quarter or so is you're able to do all these things simultaneously. Right. The other feature we're actively working on is what we're calling percentage rollouts, right. Where you can roll out the agent to a certain percentage of your customers, not all of them, maybe like 5%. So it's a great way for you to get started and try out the agent. So those are like two things we've launched recently. I think that'll be very helpful for admins as they're getting as they want to use the agent on channel like email or other messaging channels, where right now it's either use the agent or you don't use the agent. And I think a big part of sort of the deployment is the handoff experience. Right. Is what happens if the agent doesn't know an answer, right? Are there certain types of queries we want the agent to sort of not answer and hand it off to a human. Right. So the first part of the handoff experience is custom triggers where you can say in this case, anything to refunds or demos or speaks to Sal, etc. Order cancellation, etc. What happens in this case is that triggers the handoff and it sends to a human. So we're going to update this experience over the course of next, maybe next couple of months to be very similar to what Bhargav showed earlier around guidelines where you can actually give it free form instructions. And that makes it a lot more powerful. You can say you have a little more control over if things like if the customer has a certain sentiment, you can say hand it off to a human, et cetera. So I think this gives you a little more control. So the first part here is like setting up the triggers. The second part is once you set up the triggers and once you define, okay, which types of conversations the human should handle, the next is you can define different handoff sort of process here. So we have a few of them. We recently introduced this async handoff. The first one is live handoff, which means it directly sends you to a human agent where the team is available. So here you can hand off all these agents to a specific team or specific member of a team, right? But you can also use for more complex routing, you can also use workflows again here, right? Or if you're using skill based routing, you can also define who the agent should hand off to based on a set of conditions, right? For example, if you are, if you have a customer who's reaching, reaching out to you from a specific country or a specific type of question, like a billing question, maybe you have a billing team, right, who needs to answer those questions versus a different team or a customer with a specific, from a specific geography, who talks in a specific language, you're able to sort of really configure all that as a part of the workflow and just like integrate the workflow here. So this is like a super powerful experience for customers because it gives them a lot of control over who the agent should hand off to. The other thing we also recently sort of launched is this setting where once you hand off to a human, what should happen? You want to keep the chat open or when you want to close the chat, especially when the, when the human is not available. So that's for that specific use case. So async handoff is for teams where they don't really have a team's live to answer questions, this has actually been a big request for customers. A lot of times they don't really have a human to take over and answer these questions. So Async handoff really sets the expectations to the visitor that the agents are not available right now, but someone can follow up with them over email, right. And it gives you a little more control over how quickly you want to answer. It's the same settings apply here. You can figure out, you can decide who to sort of hand off to and or based on what conditions. Again, this gives you more, more power, more control. And the final one is don't hand off at all. Right. So in this case the agent will continue to sort of answer questions, right? But then they will inform the support team or any team who's looking at this that the agent is not that someone can follow up, but for a visitor, they can continue to answer and ask questions and the agent will continue to answer them. I think the one other thing I want to talk about as a part of handoff process is where do you see all these handoffs if you're using help desk? So we have a really solid experience where you have a customer agent view here and you can see all the customer agents or conversations in one place, right? And then you can also see if it's escalated. For example, in terms of Async handoffs, we add this escalation property and so you know the types of tickets that actually need your attention immediately. Outside of that, you can also see all the conversations the agent has closed by going to all closed and then actually looking for the agent name. In this case it's Luma. So I can see all the conversations the agent has closed. I can click into it and I can get a. I can see how the agent has actually answered these, these questions. So this is a great way for you to see how the agent is actually is doing. So that's the other thing I want to quickly sort of touch upon. And the final one is channel settings. So we know that these channels are different. Like for example, the way the agent sort of behaves on a messaging channel is going to be different on images on than email channel because the expectations of these channels are very different. So we've launched a few configurations here to help. The first one for conversation for all messaging channels is the ability to sort of capture email addresses. Right. So in this case we said don't ask email, but you have a lot more control over when the agent should Ask email either at the start of the conversation, after answering the first question, before handing off to human, etc. Now I think we recently launched inactivity behavior where the earlier, if the visitor sort of leaves the conversation, we automatically close that conversation after 24 hours. But now you're able to set that to whatever time you want, right? 15 minutes, 30 minutes. You can also do a custom time as well. So this again improves things like time to resolution because you don't have to wait for 24 hours, you can close it immediately. In terms of email, so there's a bunch of new configurations in email. One is email signatures. So I think what the agent does is agent looks up the signatures that are already set up on an email channel. So in this case if you go to the inbox and go to your email settings, you will see signature there and we just like follow the signatures. The second is reply behavior. If there are multiple recipients on an email thread, you can choose if the agent should reply to only the sender or to everybody on the thread. So you have more control over this. And this is the other big one which is ignore lists. These are email addresses the agent shouldn't respond to. Again, this is based on a lot of feedback we've gotten is if there are certain domains or certain email addresses. A lot of times we've seen customers forward email addresses and they don't want the agent to respond. So you have control over like domains or types of emails. You don't want the agent to respond at all. What, what happened in this case is it just directly sends them to a human. And finally inactivity behavior. Again, for email, you can choose what this should be. Should it be 8 hours, 24 hours or whatever custom time you want. So that's the, those are the new configurations we have. So I think in terms of deployment, the new things we've launched is ability to give you more control over who your agent engages with, what time of the day, more control over once it hands off to a human. Right. Who should it hand off to and then more settings to ensure the agent works really well on these channels. So that's a quick sort of TLDR on the overall sort of customer agent experience Today. I'll also quickly talk about a couple of other things that are sort of coming down. The first one is so we right now have a 28 day free trial experience. So this trial is applicable for all customers who are I have not using Customer Agent actively right now or have used it in the past, but they've stopped using it, but they want to try it again, right. So for them there's a 28 day free trial that's available for them. And what happens is you have the full experience of the customer agent and you can use it how much ever you want. And once the trial ends we automatically pause the agent and then once, once you add credits then you can go and sort of manually restart the agent. I think the other thing about this client experience is cool is the trial only starts once you deploy the customer agent, not once you set it up. So that's again like a very customer friendly way of doing trials is not just when you set up the agent, but once you set up the agent, once you test it, once you're ready to go and you deploy the agent on a specific channel, that's when the trial starts. So that's a little bit about the trial experience and finally what's next. So I think we'll continue to sort of evolve our deployment process giving you more control over how do you deploy, right? Who do you deploy to, what times of day, what percentage of customers, what segments of customers, etc etc. So I think we feel that's going to be like a really big sort of unlock for a lot of customers and admins to get started on customer agent. The second one is to support more complex business types. This is a request we get all the time is a lot of our customers, especially with as we move market, have multiple brands, they operate in multiple regions, have multiple product lines, right. Etc, etc and we are sort of evolving the customer agent to sort of support these types of more complex businesses based on where the agent understands who's reaching out and answers the questions accordingly. And you will have more control when you sort of set up the agent to be able to sort of define that. Next is new agent optimization tools. We know the the customers who are seeing the most value are the customers who are continuously optimizing the agent, improving the agent today. So we're going to give you more tools to be able to do that easily and quickly. And then finally deeper CRM integration. Bargo sort of spoke to you about how the agent today taps into your contact properties and to personalize interactions. So we want to sort of take that to the next level including integrating with all your CRM sort of data, including custom objects, not just for personalization but also for answering questions. A lot of customers store information like business information here. The agency will tap into that to answer questions that we order information, what have you. Right. So this is like a quick tldr over the course of next six months or so, we will deliver on these things and we believe that customization will be a lot more sort of relevant for larger businesses once we do that. So with that, Kyle, I think we are done with what we've prepped for. Happy to sort of take any questions live, but I think we've also been answering questions Async as well on Zoom Chat.

Great. Yeah, way to nail the timing. Thanks so much for being here. We are going to go through the the Q and A and friends. Another setting in Zoom that I apparently messed up is. Oh, no upvotes are here. Are they? Oh, good. It was turned on. Great. All right, so Deanna fixed that. We'll have it on from the beginning in the future. So, yeah, hop into the Q and A, look at the open ones, upvote the ones you're most interested in. We'll work most upvoted to least upvoted in the 9 minutes we have left. 8 minutes after I give you a second to vote. All right, great. I can't wait more than 10 seconds. We've learned this every week. All right. Ryan asks. Oh, and Vijay is typing an answer. Vijay, if you want to just talk out loud. Oh, it did?

Yes. I think Ryan asked this question about. Sorry. I think Ryan asked a lot of really good questions today. So thank you, Ryan, for participation. I think he asked a question about if you use workplace in terms of human. Anders, would. Would you save credits? I think the answer is yes. So if you use workflows, what happens is even before it hits the customer agent, we sort of route it to a human agent. Right. So I think that's why I think workflows will be very, very powerful because then you can really use the workforce to orchestrate the type of interaction you want the customer agent to even handle and you want humans to handle. This could be types of customers, it could be specific topics, what have you. So, yeah, nice.

Oh, weird.

Can the customer tap into order object? So not right now. So that's something we want to work towards. Right. Is once we integrate into custom objects, the agent should be able to tap into any sort of custom object related property information and answer questions based on that.

Cool. Jeanette asks, how does customer agent compare to other ticketing systems for the support?

I mean, I think as I said earlier, I feel like we are sort of best in class when it comes to customer agents specifically. But if you want to integrate customer agent and help desk together with help desk is sort of our solution. Really focus for sort of build ground up for our support teams. I think there are a lot of interesting things we support. Actually some of that I was not able to cover today. But we have this feature called Reply Recommendations, right. Where what it does is the agent actually recommends answers to your support reps. Right. It's essentially sort of trained on the same thing and the agent recommends answers to your support reps who can then use it. So but that's a Help desk exclusive feature. So we have a lot of these features which are sort of Help Desk exclusive features. So I think putting together both Customer agent Help Desk together, I think I would say that we are one of the best in the business.

That's great. Sorry, I have no idea how any of this works. Oh, I understand what's happening. While you are answering with your mouth, the others are answering with their fingers and so things keep disappearing. Bhargava and Kriti, if you want to come off mute, are there questions you have answered that you would like to talk out your answer for the group real quick before we call it a day?

Yeah, I can start. There was a really good one about improved reporting and something we didn't really talk or touch much into is in the future we're working on like a topic explorer type of view where you can actually see all of the end visitor conversations organized in a way where these are topics that are performing well and topics that are maybe not performing that well and tie that with coaching ability as well. So I think a lot of people were wanting to know what other improved reporting would be out there. And like that's when we're actively working on and I think you know, one of at least gave an idea about for like an export ability. So I think that's something we can definitely take away and consider.

One question I saw multiple questions related to lead qualification and how you can route things or use it in campaigns and stuff like that. So my answer to all of those questions is that workflows is a really powerful tool. So think about how you can use Customer Agent in tandem with workflows. So whenever the customer agent qualifies a lead or like talks to a qualifies a lead in the sense that not like asks all the questions and then qualifies them as qualified, not qualified or partially qualified, it creates a contact and updates this contact property called Customer Agent Lead status and you can then create workflows based on that property like use that as a trigger and then do a bunch of things with these contacts. You can enroll them in campaigns or hand them off to the prospecting agent. So whatever you want to do with these contacts, there's a lot you can do and achieve with workflows. I would encourage you to try that out for sure.

Amazing. Thanks so much. We are coming up at time here real quick the three of you. If folks want to connect with you, do you have a way you like them to do that? Are you LinkedIn people? What kind of people are you?

I think LinkedIn works. We can also give our email address. I don't know if that's.

That is up to you. Other. Other product people sometimes do that. I never lead straight with volunteering that because it's a talkative group with lots of opinions. You'll get input.

Okay. In your inbox, Kyle, I listen to whatever you have to say. Whatever your recommendation is.

Right.

I think the thing is we are all available. We love talking to customers and partners all the time. Right. Roadmap is built off everything. We keep hearing from them. So we are all ears. So yeah.

Amazing. Well, thanks so much for being here. I think we will call it here and thanks everyone for muscling through. We will have fewer technical difficulties next time. Vijay and team, thanks for being the guinea pigs here. You did great. Everybody go enjoy your Tuesday.

See you next week. My friends,

Welcome to the island hug no flashes facts and a whole lot of love get it done Acts of gentle we're all in by choice. Fixing fields and hacking time making magic with no headline power drop gold Tiana brings fire in a crew that's here to lift you higher we swap the whims and share the pain north solo in this game it's the acting hug where legends walk in from chaos to clarity that's how we win Study got a challenge accepted Portal perfection full. Quick fix Queens workaround kings custom coded everything tabs open coffee strong you've been solving all along we've got the know how we've got the map even HubSpot Prada pulls up to chat you're not just her you help shape the tools admins rewriting all the rules thanks for building with the hub hug thanks for powering the hug the h campaign for too much there's a whole crew hold inside it I move together powered by insight.