WEBVTT 00:00:00.000 --> 00:00:01.620 Today I'm going to show you how Microsoft 365 00:00:01.620 --> 00:00:04.780 Copilot gives you access to the best of both 00:00:04.780 --> 00:00:08.640 worlds with Anthropix Cloud and OpenAI's ChatGPT. 00:00:08.820 --> 00:00:11.000 My name is Shervin Shafi and I'm a Copilot Principal 00:00:11.000 --> 00:00:14.560 Engineer at Microsoft. Welcome to Collaboration 00:00:14.560 --> 00:00:17.440 Simplified. Now in the spirit of full transparency, 00:00:17.679 --> 00:00:20.559 this is April of 2026 and these are frontier 00:00:20.559 --> 00:00:23.399 features at the moment and later on in this video 00:00:23.399 --> 00:00:25.579 I'm going to cover exactly what you need to do 00:00:25.579 --> 00:00:28.219 so that you can get access to frontier features. 00:00:28.300 --> 00:00:31.640 So hold on for that. But if you have access to 00:00:31.640 --> 00:00:34.359 Frontier features and Anthropic, all you got 00:00:34.359 --> 00:00:38.420 to do is go to new chat in your M365 Copilot 00:00:38.420 --> 00:00:41.659 application. And if you notice on the top right 00:00:41.659 --> 00:00:44.579 -hand side, typically you're set to auto, which 00:00:44.579 --> 00:00:47.520 means that Copilot decides which model to use 00:00:47.520 --> 00:00:51.039 best. Quick response, think deeper, but new, 00:00:51.200 --> 00:00:56.500 now you have access to Opus by Anthropic. And 00:00:56.500 --> 00:00:59.979 you also have access to multiple models by GPT. 00:01:00.280 --> 00:01:02.880 And if you select that, you'll see all of them 00:01:02.880 --> 00:01:05.959 listed here. So let's go ahead and select Opus. 00:01:05.959 --> 00:01:08.439 As I was preparing for this, I was thinking to 00:01:08.439 --> 00:01:11.560 myself, what is a good prompt that can help exercise 00:01:11.560 --> 00:01:15.019 Opus's capabilities? So I came up with this using 00:01:15.019 --> 00:01:17.659 my recent meetings, notes and emails about this 00:01:17.659 --> 00:01:20.819 project that I've been working on. One, create 00:01:20.819 --> 00:01:24.420 a one -page executive Word document. Two, a five 00:01:24.420 --> 00:01:27.799 -slide PowerPoint deck. And three, an Excel tracker 00:01:27.799 --> 00:01:31.060 with open action items. Now, I don't know if 00:01:31.060 --> 00:01:34.680 this is the best use of Anthropix Opus Cloud. 00:01:34.939 --> 00:01:36.900 Now, you can see after a little bit of time, 00:01:37.019 --> 00:01:40.239 it came back with this executive summary that 00:01:40.239 --> 00:01:43.480 it gives me the Word link for. It came back with 00:01:43.480 --> 00:01:46.400 this PowerPoint slide deck that I can access. 00:01:46.700 --> 00:01:50.060 And then also this Excel tracker. And so the 00:01:50.060 --> 00:01:53.400 Word document looks something like this, which 00:01:53.400 --> 00:01:56.719 looks pretty nice. The PowerPoint looks something 00:01:56.719 --> 00:02:00.079 like this, which also looks really nice. And 00:02:00.079 --> 00:02:02.959 then the Excel tracker looks something like this. 00:02:03.079 --> 00:02:05.200 So this is the first use case, being able to 00:02:05.200 --> 00:02:10.719 use Opus by Anthropic instead of OpenAI's ChatGPT. 00:02:11.229 --> 00:02:13.849 Now, let me know which one works best in what 00:02:13.849 --> 00:02:17.250 situation, because that's subject to your experimentation. 00:02:17.370 --> 00:02:20.169 But I will say that in the next part of the video, 00:02:20.270 --> 00:02:23.110 it's very clear which one shines when and where. 00:02:23.250 --> 00:02:25.590 And for that, what we're going to do is we're 00:02:25.590 --> 00:02:28.469 basically going to select researcher on the left 00:02:28.469 --> 00:02:31.590 -hand side here. And notice that when researcher 00:02:31.590 --> 00:02:35.879 comes up, you'll see this. auto GPT responses 00:02:35.879 --> 00:02:39.000 refined by cloud which is a frontier feature 00:02:39.000 --> 00:02:42.919 and that is actually shown to you when you select 00:02:42.919 --> 00:02:45.800 this drop down here it says the same thing in 00:02:45.800 --> 00:02:48.120 auto and then there's also something called the 00:02:48.120 --> 00:02:51.599 model council GPT and cloud deep reasoning let's 00:02:51.599 --> 00:02:53.520 look at a couple of slides to understand this 00:02:53.520 --> 00:02:55.939 a little bit better in this auto mode it's also 00:02:55.939 --> 00:02:59.520 known as critique in researcher responses are 00:02:59.520 --> 00:03:02.699 generated by GPT and cloud they work together 00:03:02.699 --> 00:03:06.159 with GPT drafting the initial response and then 00:03:06.159 --> 00:03:09.400 cloud reviewing for accuracy completeness and 00:03:09.400 --> 00:03:12.780 citation integrity before it's delivered so essentially 00:03:13.479 --> 00:03:16.620 OpenAI creates the summary and then Anthropic 00:03:16.620 --> 00:03:19.199 reviews it for completeness and accuracy. So 00:03:19.199 --> 00:03:21.520 they work in collaboration with each other. Now, 00:03:21.580 --> 00:03:23.939 on the other hand, the Model Council simultaneously 00:03:23.939 --> 00:03:27.919 runs a prompt across multiple AI models within 00:03:27.919 --> 00:03:30.539 Researcher. And this summarizes where the models 00:03:30.539 --> 00:03:33.759 agree, where they disagree, and what each uniquely 00:03:33.759 --> 00:03:37.080 adds, making it easier to validate answers and 00:03:37.080 --> 00:03:39.719 spot the gap. Switching back to Researcher, in 00:03:39.719 --> 00:03:42.180 auto mode, let's type in a prompt. Tell me about 00:03:42.180 --> 00:03:44.759 Home Depot and Lowe's and how they compare and 00:03:44.759 --> 00:03:47.340 they differ. First, it comes back and it asks 00:03:47.340 --> 00:03:49.879 you for some clarifying questions and how long 00:03:49.879 --> 00:03:52.030 you want your... report to be and so in this 00:03:52.030 --> 00:03:54.770 case i just said go ahead and make it a short 00:03:55.039 --> 00:03:57.939 and next when the reports actually completed 00:03:57.939 --> 00:04:00.939 you'll see that it says generated by GPT but 00:04:00.939 --> 00:04:04.460 refined by cloud and in this case here's Home 00:04:04.460 --> 00:04:06.740 Depot versus Lowe's and here's the executive 00:04:06.740 --> 00:04:09.680 comparison as well as category specific strengths 00:04:09.680 --> 00:04:12.780 and so on so a very good report and it's critiqued 00:04:12.780 --> 00:04:15.719 by cloud but written by chat GPT now if we go 00:04:15.719 --> 00:04:18.339 back to researcher and this time we select the 00:04:18.339 --> 00:04:21.160 model council and we put the same prompt in and 00:04:21.160 --> 00:04:23.480 we send it off let's see what happens this time 00:04:23.769 --> 00:04:25.910 So in this case, when it's done, you'll see that 00:04:25.910 --> 00:04:28.149 on the left -hand side, you have the GPT report 00:04:28.149 --> 00:04:30.110 and the right -hand side, you have the cloud 00:04:30.110 --> 00:04:32.290 report. You can select these arrows and then 00:04:32.290 --> 00:04:35.069 actually get to the full report. So here's the 00:04:35.069 --> 00:04:38.529 GPT one, and you see the full report in all its 00:04:38.529 --> 00:04:41.009 glory. It's super detailed. And then you can 00:04:41.009 --> 00:04:44.069 also switch to the anthropic one and see its 00:04:44.069 --> 00:04:47.810 report in full detail as well. So tons of information. 00:04:48.050 --> 00:04:50.269 We're not going to go through this all, but what's 00:04:50.269 --> 00:04:54.339 unique about this model council is that it actually 00:04:54.339 --> 00:04:58.079 puts these side by side so the first section 00:04:58.079 --> 00:05:01.220 says having done extensive research both cloud 00:05:01.220 --> 00:05:04.819 and GPT underscore that these are the things 00:05:04.819 --> 00:05:07.839 that they both agree with and then as you scroll 00:05:07.839 --> 00:05:11.600 down it says what each model brought to the table 00:05:11.600 --> 00:05:15.379 so cloud brought these unique things to the table 00:05:15.379 --> 00:05:17.920 and then GPT brought these historical contacts 00:05:17.920 --> 00:05:21.430 and then as you scroll down it says Where did 00:05:21.430 --> 00:05:25.949 the reports differ? And so the top areas of agreement 00:05:25.949 --> 00:05:29.470 between GPT and cloud were these. And then as 00:05:29.470 --> 00:05:31.930 you scroll down, the top areas of disagreement 00:05:31.930 --> 00:05:35.110 were these sections. So what do they agree on? 00:05:35.189 --> 00:05:37.750 What do they disagree on? Now, one last thing. 00:05:37.850 --> 00:05:39.829 If for some reason you don't see these frontier 00:05:39.829 --> 00:05:42.649 features, talk to your IT admin or share this 00:05:42.649 --> 00:05:46.290 clip with them. So as an IT admin, you have to 00:05:46.290 --> 00:05:50.620 go to your Microsoft 365 Admin Center. and you'll 00:05:50.620 --> 00:05:53.339 see this menu item called copilot and then there's 00:05:53.339 --> 00:05:55.759 this settings menu so when we select settings 00:05:55.759 --> 00:05:58.779 you have to go to this section called view all 00:05:58.779 --> 00:06:01.779 and when we select that the first thing you got 00:06:01.779 --> 00:06:05.420 to do is scroll down to copilot frontier and 00:06:05.420 --> 00:06:09.600 when you select that you decide if you don't 00:06:09.600 --> 00:06:13.160 give access to anyone if you give access to everyone 00:06:13.870 --> 00:06:16.850 or if you're just giving access to specific users 00:06:16.850 --> 00:06:21.550 next there's a section here that says ai providers 00:06:21.550 --> 00:06:25.050 operating as microsoft sub processors which is 00:06:25.050 --> 00:06:28.670 anthropic so when you select that you decide 00:06:28.670 --> 00:06:33.009 if anthropic is either enabled for all no or 00:06:33.009 --> 00:06:36.009 specific users and this is on by default for 00:06:36.009 --> 00:06:39.290 some tenants and specifically in europe you have 00:06:39.290 --> 00:06:41.490 to go and actually enable this because it's off 00:06:41.490 --> 00:06:44.160 by default So that's another place you got to 00:06:44.160 --> 00:06:46.500 go to. And if you're curious what a sub processor 00:06:46.500 --> 00:06:49.259 means, we actually have a documentation that 00:06:49.259 --> 00:06:52.019 talks about Anthropic as a sub processor. And 00:06:52.019 --> 00:06:54.399 if you scroll down a little bit, it talks about 00:06:54.399 --> 00:06:57.839 that Anthropic will operate with Microsoft oversight 00:06:57.839 --> 00:07:00.660 through contractual safeguards and appropriate 00:07:00.660 --> 00:07:04.410 technical measures. and that such use is covered 00:07:04.410 --> 00:07:07.209 by our enterprise data protection which is key 00:07:07.209 --> 00:07:10.589 because if you go back to our M365 app and you 00:07:10.589 --> 00:07:14.829 go to new chat and you select cloud you'll see 00:07:14.829 --> 00:07:17.769 that we still have this green shield on which 00:07:17.769 --> 00:07:20.910 means that you have Microsoft's best security 00:07:20.910 --> 00:07:23.329 measures that are our enterprise data protection 00:07:23.329 --> 00:07:26.329 and similarly if you go to researcher and you 00:07:26.329 --> 00:07:30.139 select The model console, you'll still see that 00:07:30.139 --> 00:07:32.720 we have the green shield for that as well as 00:07:32.720 --> 00:07:35.120 for auto. If you enjoyed today's video, make 00:07:35.120 --> 00:07:37.319 sure to give it a like and consider subscribing. 00:07:37.720 --> 00:07:39.800 Let me know if you have any questions and I'll 00:07:39.800 --> 00:07:41.180 catch you on the next one.