WEBVTT 00:00.080 --> 00:01.120 Okay, here we go. 00:01.160 --> 00:02.840 New agent is what I press. 00:03.000 --> 00:03.560 And it's. 00:03.600 --> 00:05.600 What type of agent would you like to create? 00:05.640 --> 00:07.880 I'm just going to say blank agent. 00:08.200 --> 00:11.280 Uh, create a name that reflects your agent's purpose. 00:11.280 --> 00:14.400 Let's call it first agent. 00:14.720 --> 00:19.120 And, uh, keep chat only off and press create agent. 00:19.520 --> 00:20.960 And here it comes. 00:21.720 --> 00:27.800 And what we see here, first of all, a system prompt which is defaulting to your helpful assistant. 00:27.840 --> 00:28.880 You know exactly what that means. 00:28.880 --> 00:30.280 Now the first message. 00:30.280 --> 00:30.600 Hello. 00:30.600 --> 00:31.400 How can I help you? 00:31.400 --> 00:32.840 Today is how it begins. 00:32.840 --> 00:38.800 You can see it says you can add variables using double curlies, very similar to expressions in N810. 00:39.280 --> 00:44.880 The voice we've got is Eric, who is apparently smooth and trustworthy, will be the judge of that. 00:45.120 --> 00:51.120 The language is English and we can select which model and Gemini 2.5 flash is the default. 00:51.160 --> 00:58.480 Let's see if it if it allows us to to pick Gemini three flash as the oh look, it has the GPT OS in 00:58.520 --> 00:59.760 there as experimental. 00:59.800 --> 01:05.780 Now they have Gemini three Pro but not Gemini three flash, which for me was was only released yesterday 01:05.780 --> 01:06.420 I think. 01:06.420 --> 01:07.860 So that's not a surprise. 01:07.860 --> 01:09.860 We'll stick with this for now. 01:10.140 --> 01:14.860 So with this, this is giving us the the the all of the setup. 01:15.020 --> 01:17.420 Now it's time for us to give it a try. 01:17.460 --> 01:18.020 Okay. 01:18.020 --> 01:19.100 Let's give this a shot. 01:19.100 --> 01:22.260 We're going to press preview to see if we can run this thing. 01:22.340 --> 01:23.140 Let's give it a try. 01:24.980 --> 01:25.620 Okay. 01:25.740 --> 01:31.260 Up comes this avatar thing and there's a cool AI agent button. 01:31.700 --> 01:32.700 Let's give it a shot. 01:34.780 --> 01:35.140 Hello. 01:35.140 --> 01:36.180 How can I help you today? 01:36.460 --> 01:36.740 Well. 01:36.740 --> 01:37.460 Hello there. 01:38.940 --> 01:39.420 Hello. 01:39.460 --> 01:40.900 It's a pleasure to assist you. 01:40.940 --> 01:42.020 How may I help you today? 01:42.300 --> 01:43.660 Tell me a fun fact. 01:47.140 --> 01:49.940 Did you know that a group of flamingos is called a flamboyance? 01:49.940 --> 01:53.740 I always thought that was a rather fitting and delightful name for such colorful birds. 01:54.260 --> 01:55.420 It certainly is. 01:55.460 --> 01:57.660 And now tell me a joke about flamingos. 01:58.940 --> 02:00.540 Why did the flamingo cross the road? 02:00.540 --> 02:02.420 Because it was the chicken's day off. 02:03.240 --> 02:03.560 Haha. 02:04.040 --> 02:04.520 Very good. 02:04.520 --> 02:05.200 Thank you. 02:06.640 --> 02:07.680 You're most welcome. 02:07.680 --> 02:09.400 Is there anything else I can help you with today? 02:10.680 --> 02:13.520 So I think that was really impressive. 02:13.560 --> 02:19.200 That of course is credit to Gemini two five flash for being the model that was that was responding to 02:19.200 --> 02:19.760 us there. 02:19.760 --> 02:27.320 But credit to 11 labs for making that so fast and so seamless and so quick to feel like we were actually 02:27.360 --> 02:29.600 speaking with a voice agent. 02:29.760 --> 02:35.640 So I think it's fabulous that it's so easy to have that kind of frictionless experience. 02:35.800 --> 02:36.480 All right. 02:36.480 --> 02:40.000 Let's take it a little bit further to have like a business conversation. 02:40.000 --> 02:44.960 Let's say you are a helpful assistant for an airline. 02:45.280 --> 02:48.840 You assist customers. 02:49.360 --> 02:50.160 I spell that right. 02:50.240 --> 02:55.400 Assist customers with their travel needs. 02:57.560 --> 03:09.020 For your reference, a ticket from New York to London a return ticket. 03:11.540 --> 03:14.740 Costs 599. 03:15.620 --> 03:17.820 Okay, so we've given it some context. 03:17.860 --> 03:19.420 We've given it some information. 03:19.420 --> 03:21.500 Why don't we change the voice? 03:21.700 --> 03:23.060 Who should we have here? 03:23.340 --> 03:26.900 Uh, we can choose, uh, Miss Walker. 03:26.940 --> 03:28.740 Warm, reassuring and round. 03:28.780 --> 03:29.540 Jason. 03:29.540 --> 03:31.700 Calm and soothing. 03:32.540 --> 03:33.780 Uh, Stokes. 03:33.780 --> 03:34.700 Relaxing. 03:34.740 --> 03:36.020 Casual and warm. 03:36.340 --> 03:37.180 Who would you like? 03:37.180 --> 03:39.020 Who do you want, Mrs. Walker? 03:39.020 --> 03:39.460 You say? 03:39.860 --> 03:42.700 All right, Mrs. Walker, but hopefully you did. 03:42.700 --> 03:43.740 That's who we'll choose. 03:43.860 --> 03:45.420 All right, Miss Walker. 03:45.460 --> 03:46.780 Miss Walker, it is. 03:47.020 --> 03:49.460 And, uh, that's, uh. 03:49.500 --> 03:51.380 Let's see how this goes. 03:51.580 --> 03:57.860 Uh, and, um, we will, uh, we will we will press the preview button and give this a whirl. 04:00.860 --> 04:01.380 Hello. 04:01.780 --> 04:02.900 How can I help you today? 04:02.940 --> 04:04.060 I'd like to go on a trip. 04:05.180 --> 04:06.220 That sounds wonderful. 04:06.540 --> 04:07.880 To help me plan your trip. 04:08.000 --> 04:11.880 Could you please tell me where you'd like to go and when you're thinking of traveling? 04:12.160 --> 04:12.360 Yeah. 04:12.400 --> 04:15.400 I'd like to go from New York to London next week sometime. 04:17.880 --> 04:22.000 A return ticket from New York City to London costs $599. 04:22.160 --> 04:24.640 Would you like me to check availability for next week? 04:28.000 --> 04:29.080 So there you have it. 04:29.120 --> 04:31.840 It's a nice interaction, obviously. 04:31.880 --> 04:33.800 It's able to use the context. 04:33.800 --> 04:36.800 It's able to use the information we put in the system prompt. 04:37.040 --> 04:39.880 This is relatively easy stuff. 04:40.000 --> 04:43.000 It's just nice that it's so quick to put it into action. 04:43.040 --> 04:43.320 Okay. 04:43.360 --> 04:47.600 So we're now going to look through some of the more advanced features of 11 labs by going across the 04:47.640 --> 04:48.600 nav here. 04:48.800 --> 04:50.200 Start with workflow. 04:50.240 --> 04:51.720 This it says it's new. 04:51.920 --> 04:53.320 This should look familiar to you. 04:53.360 --> 04:55.720 It looks quite like the canvas in N810. 04:55.920 --> 04:58.200 And indeed it's where you can put in nodes. 04:58.200 --> 05:02.680 And you can have things like use of tools to connect different agents to each other. 05:02.720 --> 05:08.040 Now, we're not going to really use this because we're going to be using N810 for our workflow because 05:08.170 --> 05:13.450 we know that and we're going to invest our time and energy in making that really, really strong. 05:13.490 --> 05:15.170 And this is a little bit more raw. 05:15.210 --> 05:17.450 This tool as it says it's somewhat new. 05:17.530 --> 05:19.530 You can experiment with it if you'd like to see it. 05:19.530 --> 05:25.450 Going to the templates button you can choose something like Qualification Flow and see here how it looks. 05:25.450 --> 05:31.450 You've got an agent responsible for deciding whether or not the user is asking a technical issue or 05:31.490 --> 05:32.490 a billing issue. 05:32.730 --> 05:37.610 And if you click on these these issues, you can see that this is something which is able for the LLM 05:37.650 --> 05:42.530 to make a decision based on a condition, the user has a technical issue. 05:42.610 --> 05:49.170 If so, the LLM can choose effectively by using a tool to pass control over to a different agent, which 05:49.170 --> 05:54.690 has a different prompt and which is going to then it's going to be using Miss Walker, the person we 05:54.690 --> 05:57.210 just spoke to, to answer the technical support question. 05:57.370 --> 06:00.730 Uh, actually, they, uh, it looks like all of them are using Miss Walker. 06:00.930 --> 06:07.570 Uh, and, uh, this this is, uh, this is how you would then divert a billing question to something 06:07.570 --> 06:09.210 with a different prompt. 06:09.210 --> 06:14.110 So this would be an example of a template that would allow you to set up multiple different voice agents 06:14.110 --> 06:20.470 and have the conversation be routed by the first agent, depending as it wishes, based on a description, 06:20.470 --> 06:22.750 a condition that is given to the LM. 06:23.070 --> 06:29.030 So it's pretty self-explanatory that that is the, uh, I'm going to put that in the bin now like that. 06:29.030 --> 06:35.990 But that is the workflow, the the sort of miniature Nan functionality built into 11 labs. 06:35.990 --> 06:37.590 The next tab along is branches. 06:37.590 --> 06:43.110 This is where you can specify different versions of your agents and try out different possibilities. 06:43.470 --> 06:49.270 Knowledge base is where you give your agent some background information, some knowledge that it can 06:49.270 --> 06:51.750 use to to give it subject matter expertise. 06:51.950 --> 06:53.590 So right now let's do this. 06:53.590 --> 06:58.550 Let's press Add Document and choose Create Text right here. 06:58.550 --> 07:02.350 And I'm going to call this Apple Product details. 07:02.510 --> 07:03.990 You may have seen I already did this. 07:04.390 --> 07:09.790 And I'm going to paste in something that I got ChatGPT to write, which is just a quick ton of information 07:09.790 --> 07:12.210 about Apple products from their website. 07:12.210 --> 07:13.370 Create text. 07:13.410 --> 07:14.130 There it is. 07:14.170 --> 07:19.090 We've just built Apple product details as part of an agent's knowledge base. 07:19.570 --> 07:20.130 Okay. 07:20.370 --> 07:23.930 Now we're going to go back to our agent and change the system prompt. 07:23.970 --> 07:31.290 You are able to answer questions on Apple products. 07:32.130 --> 07:36.490 And let's give this a try to see if it can look into its knowledge base. 07:36.690 --> 07:37.010 Okay. 07:37.050 --> 07:39.890 Let's pick another another person here. 07:39.930 --> 07:41.210 We've had Miss Walker. 07:41.410 --> 07:46.250 Let's now let's pick, uh, um, Jason. 07:46.250 --> 07:48.850 Calm, meditative and soothing. 07:49.170 --> 07:50.330 That sounds good to me. 07:50.530 --> 07:53.330 Uh, Jason, we have, um. 07:53.370 --> 07:53.970 Okay. 07:54.130 --> 08:01.010 And now let's come back here and press preview, and we will give Jason a call. 08:03.490 --> 08:03.810 Hello. 08:03.810 --> 08:04.890 How can I help you today? 08:05.490 --> 08:05.810 Hello. 08:05.850 --> 08:07.770 I'm interested in buying an Apple Watch. 08:07.810 --> 08:09.450 What's the latest version? 08:09.450 --> 08:10.250 And the price? 08:11.970 --> 08:16.670 The latest version of the Apple Watch is the Apple Watch Series 11, which has a reported launch price 08:16.670 --> 08:18.990 of around $399. 08:19.230 --> 08:23.950 There are also other models available, such as the Apple Watch SE three, which is more affordable, 08:24.270 --> 08:27.910 and the Apple Watch Ultra three, which is designed for rugged use. 08:29.710 --> 08:30.630 Well, thank you very much. 08:34.550 --> 08:35.230 You're welcome. 08:35.230 --> 08:38.910 Do you have any other questions about the Apple Watch or any other Apple products? 08:39.870 --> 08:40.390 All right. 08:40.390 --> 08:41.230 We'll leave it there. 08:41.230 --> 08:42.190 So there you have it. 08:42.230 --> 08:46.230 It was able to consult a knowledge base and answer questions on it. 08:46.230 --> 08:49.470 And this is the first glimmer of the thing called rag. 08:49.470 --> 08:53.990 That is going to be our other topic for this week, that we're going to build out in a more robust and 08:53.990 --> 08:57.790 industrial strength way in N810 and connect the two together. 08:57.790 --> 09:04.230 But you can see that there's this this very simple implementation through the UI that you get with, 09:04.230 --> 09:10.670 with, uh, with 11 labs, that makes it so easy to give your agent subject matter expertise, domain 09:10.670 --> 09:13.350 expertise, just using documents. 09:13.350 --> 09:14.630 So you should give this a try.