WEBVTT 00:00.080 --> 00:04.320 Okay, it's time to build the final subagent of the program we're going to amplify. 00:04.360 --> 00:06.160 We're going to account executive. 00:06.200 --> 00:16.640 We create a new workflow and we're going to call it the demo booking Subagent B booking agent. 00:16.680 --> 00:17.640 There we have it. 00:17.680 --> 00:19.400 We're going to add a first step. 00:19.440 --> 00:22.000 It's going to be on a chat. 00:22.040 --> 00:23.480 We'll change that later. 00:23.920 --> 00:28.480 And we are now going to add a agent shockingly. 00:28.680 --> 00:31.120 Uh, and uh that all looks great. 00:31.120 --> 00:31.640 For now. 00:31.960 --> 00:33.200 Give it a chat model. 00:33.200 --> 00:34.600 I'm going to pick OpenAI. 00:34.640 --> 00:37.400 You pick whatever you want and we'll come in here. 00:37.400 --> 00:40.520 I'm going to pick GPT 5.2. 00:40.800 --> 00:42.600 Uh, a good strong model. 00:42.600 --> 00:44.920 But again it's totally your choice. 00:44.920 --> 00:48.080 I'm going to say I want the reasoning effort to be low. 00:48.760 --> 00:54.680 And that's a sensible starting position for us for our last AI agent. 00:54.680 --> 00:58.600 And we're now going to add a tool and we're going to add in one new integration. 00:58.600 --> 01:00.580 But it's an easy peasy one. 01:00.580 --> 01:06.100 We are gonna add in a tool, which is one of the first ones people show you with, and which is the 01:06.100 --> 01:07.660 Google Calendar. 01:07.940 --> 01:09.860 An easy thing we can put in here. 01:09.900 --> 01:11.540 Google calendar tool. 01:11.580 --> 01:15.020 We have to select a credential, create a new credential. 01:15.180 --> 01:22.460 This is one of the OAuth tools where if you're on a cloud, they've already got the OAuth two setup, 01:22.460 --> 01:27.060 the OAuth two app set up so that all you have to do is press connect with Google. 01:27.100 --> 01:30.260 It comes on in here with the usual kind of pop up. 01:30.300 --> 01:36.020 Choose an account, choose this account, continue and connection successful. 01:36.020 --> 01:37.700 I can now close that window. 01:37.860 --> 01:38.860 It's green. 01:38.900 --> 01:40.860 We're connected to my Google calendar. 01:41.020 --> 01:42.780 It's it's crazy simple. 01:43.220 --> 01:44.700 All right, back we come here. 01:44.700 --> 01:46.140 We can now choose this. 01:46.140 --> 01:47.380 What do we want to do. 01:47.380 --> 01:53.140 So we can either choose to to look at individual events and get my events from the calendar. 01:53.300 --> 01:56.100 Or we can look at overall calendar availability. 01:56.260 --> 01:58.580 Um, so we want to think about what we want to do now. 01:58.730 --> 01:59.010 Okay. 01:59.050 --> 02:04.610 So the purpose of this agent is going to be to negotiate a good slot when I can give a demo to the person 02:04.610 --> 02:08.410 that's reaching out, that's being that's come in to, to communicate. 02:08.410 --> 02:15.050 And, uh, I'm mindful of the lethal trifecta and making sure that since this information is going to 02:15.050 --> 02:20.490 be available externally, I want to be careful to avoid exposing private data. 02:20.690 --> 02:25.970 Now, the typical way you might do this is by exposing my event information so that the agent can negotiate 02:25.970 --> 02:28.010 a slot knowing all of my events. 02:28.010 --> 02:33.890 But but again, my my lethal trifecta alerts my spidey senses ringing there that I probably wouldn't 02:33.890 --> 02:34.530 be comfortable with. 02:34.530 --> 02:40.250 All of my calendar events are exposed to the internet, so maybe a better approach instead is only to 02:40.290 --> 02:45.090 make available, uh, whether or not I'm available for a particular event. 02:45.090 --> 02:45.490 So. 02:45.490 --> 02:46.530 So that's what I'm going to do here. 02:46.530 --> 02:48.570 I'm going to make the operation availability. 02:48.570 --> 02:50.250 I pick my calendar here. 02:50.370 --> 02:56.690 Uh, and I'm going to say the model should choose the start and end time. 02:56.850 --> 03:03.950 And just to be clear, I'm going to give a manual description of this, uh, whether, uh, the user 03:03.950 --> 03:13.150 is available to give a demo, uh, from the given start time to end time. 03:13.150 --> 03:14.950 That makes super clear what's going on here. 03:14.950 --> 03:17.670 These times are defined by the model. 03:17.830 --> 03:20.350 And let's go with that. 03:20.550 --> 03:24.110 That is the Get Availability tool set okay. 03:24.150 --> 03:28.790 Next the final tool here is going to be one to actually create the event. 03:28.790 --> 03:30.950 So back we go to the Google Calendar tool. 03:31.510 --> 03:36.350 And this time it's going to be something creating an event. 03:36.510 --> 03:37.870 We'll pick my calendar. 03:38.150 --> 03:38.830 There it is. 03:38.830 --> 03:42.230 We'll let the model choose the start and the end. 03:42.270 --> 03:43.990 And we're going to add in some other fields. 03:44.270 --> 03:48.510 And look here is where I'm going to to be a little bit lazy. 03:48.510 --> 03:54.390 And what I'm not gonna do is I'm not going to have it actually invite the person that we are going to 03:54.390 --> 03:55.590 be having this demo with. 03:55.630 --> 04:00.330 And there's a reason for that, which is that we are going to be rolling this, this, this course out. 04:00.330 --> 04:03.050 I'm hoping lots of people like yourself are going to be taking it. 04:03.050 --> 04:08.730 And I am fearful that if I'm not careful, someone is going to accidentally run this and invite a bunch 04:08.730 --> 04:14.610 of senior CFOs at companies like Accenture to to fake sales meetings with them. 04:14.610 --> 04:19.690 So I'm going to fall short of actually inviting a third person to it, third party to it. 04:19.690 --> 04:25.770 But hopefully it's super clear to you how you can do that should you wish to do it for your business. 04:25.890 --> 04:30.250 Uh, but otherwise you can assume that you would either automate it here, or that you would have that 04:30.250 --> 04:35.210 as a manual step, that you would actually do the final stage of inviting a third party. 04:35.250 --> 04:40.010 When you see the invite in your calendar and when you get a push notification. 04:40.370 --> 04:40.970 Okay. 04:41.010 --> 04:45.290 So we'll add in a description field that the model can set. 04:45.290 --> 04:49.370 But for we won't add in the invitees, uh, for for this go round. 04:49.370 --> 04:52.210 And you could certainly take that as a challenge to yourself. 04:52.210 --> 04:56.760 And then make sure, of course, that you're testing it thoroughly before you actually start inviting 04:56.760 --> 04:58.600 people out there on the internet. 04:58.600 --> 05:03.080 But but because we're doing this in a very careful way, it should be very robust. 05:03.680 --> 05:04.360 Okay. 05:04.560 --> 05:11.360 So with that, we might as well, uh, give it a manual description as well, a create an event to represent 05:12.040 --> 05:18.240 a demo, uh, at a time that is available. 05:19.160 --> 05:20.200 Uh, okay. 05:20.240 --> 05:21.680 That's all great. 05:21.920 --> 05:24.200 I think we're in really good shape here. 05:24.520 --> 05:29.520 Let's go back and, uh, set up our prompt and then give this a try. 05:29.560 --> 05:30.960 All right, here we go. 05:31.000 --> 05:32.440 We're going to our agents. 05:32.480 --> 05:35.520 We are going to add a system message. 05:35.800 --> 05:37.520 We're going to put in here. 05:37.840 --> 05:42.080 You're provided with a person who would like to receive a demo of the account product and their preferred 05:42.120 --> 05:43.000 time slot. 05:43.040 --> 05:47.440 Use your tool to check the calendar's availability and then book the demo in the calendar. 05:47.440 --> 05:49.200 Reply with the results. 05:49.760 --> 05:51.160 Um, okay. 05:51.280 --> 05:55.020 And now I'm going to say require output format. 05:55.380 --> 05:57.100 Let's get out of here. 05:57.100 --> 05:58.740 Let's go to the output parser. 05:58.780 --> 06:01.180 You know, I like structured outputs. 06:01.340 --> 06:02.740 Let's just have it like this. 06:02.740 --> 06:05.500 But what we'll have is simply success. 06:08.460 --> 06:08.980 Or whatever. 06:08.980 --> 06:09.940 We'll be optimistic. 06:09.940 --> 06:12.340 And then we will have, uh. 06:12.740 --> 06:13.620 Summary. 06:15.700 --> 06:19.500 Summarize the outcome here. 06:22.700 --> 06:25.420 Uh, okay. 06:25.700 --> 06:28.540 Uh, and, uh, that should be great. 06:29.060 --> 06:37.580 Let's now just go and add in some if here if routes based on the results. 06:37.780 --> 06:39.100 We know what to do now. 06:39.140 --> 06:40.300 We won't make a mistake again. 06:40.340 --> 06:41.900 This should be an expression. 06:42.140 --> 06:43.020 It should be. 06:44.460 --> 06:45.580 I'll let you say it out loud. 06:45.900 --> 06:49.300 JSON output dot success. 06:50.580 --> 06:53.960 Close brackets like so. 06:54.880 --> 06:55.920 Uh, okay. 06:56.160 --> 06:58.080 And then this goes to Boolean. 06:58.080 --> 06:59.640 Goes to is true. 07:00.120 --> 07:00.960 All right. 07:01.000 --> 07:02.200 That looks right to me. 07:02.360 --> 07:05.880 And now we'll add in two push notifications here. 07:06.080 --> 07:07.600 And we should be ready to try this. 07:07.640 --> 07:07.920 Okay. 07:07.960 --> 07:09.720 And I've done all of that. 07:09.760 --> 07:13.320 I've put it in the push notifications on the true and the false. 07:13.520 --> 07:15.760 So that and I've also added in the stickies. 07:15.840 --> 07:21.800 So we've now got a nice little setup there where based on what comes in, it will try and book a meeting 07:21.800 --> 07:23.280 and then report on it. 07:23.480 --> 07:24.800 Let's give this a shot. 07:24.840 --> 07:25.000 Wait. 07:25.000 --> 07:27.560 Just before we run it, let's make this feel even more real. 07:27.600 --> 07:29.400 Go into the calendar. 07:29.440 --> 07:31.440 Here we've got the, um. 07:32.000 --> 07:34.320 We've got the fields that we are populating. 07:34.320 --> 07:37.960 And there's, of course, a bunch of fields we're not populating intentionally because we don't don't 07:37.960 --> 07:39.880 want this to go to go wrong. 07:39.880 --> 07:41.560 But I might just populate a few more. 07:41.560 --> 07:43.000 Let's populate location. 07:43.000 --> 07:47.200 This is where you could put your zoom, uh, your zoom link so that people can have a video call with 07:47.200 --> 07:47.280 you. 07:47.280 --> 07:48.840 So let's put in something fake here. 07:48.960 --> 07:53.110 Uh, like zoom us a b c d e f. 07:53.310 --> 07:53.950 Imagine that. 07:53.950 --> 07:55.150 That's actually my zoom account. 07:55.150 --> 07:56.350 So it's going to go in there. 07:56.390 --> 08:00.070 Let's add another field and let's have this be the summary field. 08:00.070 --> 08:01.790 And we'll let the model populate that. 08:01.990 --> 08:04.470 And that's just going to make it a little bit more well rounded. 08:04.470 --> 08:07.950 And with that let's open the chat and give this thing a try. 08:07.990 --> 08:08.510 Okay. 08:08.550 --> 08:19.710 So now in the chat I'm going to say book a demo with Tom Jones singer at a 2 p.m. eastern on Jan five, 08:19.750 --> 08:20.870 2026. 08:21.470 --> 08:22.950 All right, let's see what happens. 08:22.990 --> 08:24.950 The AI agent goes out there. 08:25.110 --> 08:27.670 It's going to be looking at my availability. 08:27.830 --> 08:28.630 Things are happening. 08:28.710 --> 08:29.950 Tools are being called. 08:30.190 --> 08:32.390 Stuff is in the works. 08:34.830 --> 08:36.030 Drum roll please. 08:38.550 --> 08:39.030 It's back. 08:39.070 --> 08:44.230 Trying to make the structured outputs, and the workflow is executed successfully. 08:44.230 --> 08:45.590 And they're there. 08:45.790 --> 08:46.990 I hope you heard that. 08:47.190 --> 08:48.670 We have liftoff. 08:48.830 --> 08:52.770 Okay, so apparently it's created something in my calendar. 08:53.170 --> 08:54.010 Let's go have a look. 08:54.050 --> 08:55.810 Well, let's go and take a look at the calendar. 08:55.810 --> 08:56.770 Here is my calendar. 08:56.770 --> 08:58.610 Here's Monday, the 5th of January. 08:58.610 --> 09:00.730 There appears to be a booking at 2 p.m.. 09:00.770 --> 09:01.490 Here it is. 09:01.490 --> 09:04.330 It's for an account AI demo with Tom Jones. 09:04.330 --> 09:06.010 It's on my zoom link. 09:06.050 --> 09:07.410 And there's the details. 09:07.410 --> 09:08.410 It's all there. 09:08.450 --> 09:11.130 It all happened just as we'd expect. 09:11.130 --> 09:12.370 Fantastic. 09:12.650 --> 09:14.890 Okay, that's the last subagent. 09:14.930 --> 09:20.770 We just have to package it up to be a subagent, and then we can move on to the final agent of the course. 09:20.810 --> 09:23.690 Okay, so back here I'm going to remove the chat message. 09:23.690 --> 09:25.850 It's not going to be triggered by chat message. 09:25.850 --> 09:27.170 We're going to have it triggered. 09:27.170 --> 09:33.010 And again I press tab to bring up these I'm going to trigger not as a executed by another workflow but 09:33.010 --> 09:34.970 again a webhook call I'm going to do. 09:35.130 --> 09:37.650 So I'm going to do HTTP post again. 09:37.970 --> 09:43.330 And I'm going to have it respond using the respond to webhook node as we did before. 09:43.570 --> 09:47.770 Um, and now I'm going to connect this in here. 09:48.050 --> 09:55.550 I'm going to open up my AI agent, I no longer have a connected chat trigger node I want to define below 09:55.550 --> 09:59.830 and an expression and I will punch in my expression in here. 10:00.150 --> 10:04.590 Please book a demo with dollar name. 10:04.630 --> 10:06.030 That's the incoming JSON. 10:06.030 --> 10:12.110 We look in the body of the post, we get the name and then the body time slot, which is going to be 10:12.110 --> 10:14.110 some description of when. 10:14.390 --> 10:17.670 And so then this will, this will appear a bit like that. 10:17.870 --> 10:23.270 Uh, and uh, and that's still going to fit nicely with this system message. 10:23.270 --> 10:24.070 So that should work. 10:24.070 --> 10:24.390 Great. 10:24.390 --> 10:25.430 Let's save it. 10:25.430 --> 10:26.870 And now go back here. 10:26.870 --> 10:30.510 And finally we have to add in the responses at the end okay. 10:30.550 --> 10:34.470 So I'm going to add in here a respond to webhook. 10:34.950 --> 10:40.070 And we will have this thing respond with uh some some text. 10:40.430 --> 10:42.070 That text can be an expression. 10:42.070 --> 10:49.060 And what we'll have it respond with is going to be uh, just exactly what came out here, this Summary 10:49.220 --> 10:49.860 right here. 10:49.860 --> 10:51.420 We'll have that be. 10:51.460 --> 10:52.620 Be what it comes up with. 10:52.820 --> 10:54.100 Uh, that is perfect. 10:54.420 --> 10:56.540 Uh, and, uh, yeah. 10:56.580 --> 10:59.780 You know, I basically we want exactly the same thing for this other leg as well. 10:59.780 --> 11:02.820 And of course, there'll be a way to bring them together and do it once, but we'll just do it this 11:02.820 --> 11:03.100 way. 11:03.300 --> 11:05.420 This seems fine enough to me. 11:05.420 --> 11:06.940 Respond to webhook. 11:07.580 --> 11:11.780 Uh, and, uh, uh, make this be, uh, sorry. 11:11.820 --> 11:14.700 The the, um, respond with text. 11:14.900 --> 11:20.740 And again, we come into the AI agent, we make this an expression and we take the summary. 11:20.780 --> 11:22.060 We drop that in there. 11:23.220 --> 11:24.260 I must have missed that. 11:24.260 --> 11:25.220 Try that again. 11:25.220 --> 11:26.420 Summary in there. 11:26.660 --> 11:28.180 And perfect. 11:28.220 --> 11:35.540 We now have everything set up to be responding back to the webhook when this webhook is called. 11:35.940 --> 11:40.460 And that then means we can now save this and publish it. 11:40.660 --> 11:44.340 And we have now published our two Subagents. 11:44.340 --> 11:47.780 Finally, we are ready for the last agent of the course. 11:47.780 --> 11:48.740 Let's go and do it!