Up next in 10
How to get ahead of the curve and future-proof your business with Power Platform and Azure OpenAI
7K views
Oct 30, 2023
In this session, you will some wonderful case studies and how these have been used in different businesses. You will delve into the capabilities and integration possibilities of these cutting-edge technologies. Discover how you can leverage Power Apps to rapidly build custom applications, harness the language generation capabilities of ChatGPT, and unlock the potential of Azure OpenAI for advanced AI functionalities. Software Architecture Conference 2023 (https://softwarearchitecture.live/) #SoftwareArchitectureConf23 #CSharpTV #csharpcorner
View Video Transcript
0:00
I'm really excited to share what I have and then
0:05
how the software architecture in terms of Power Platform and Azure OpenAI can work together
0:10
You get some amazing things. I have no much demo here, but some use cases and those use cases
0:19
how you can actually intertwine with Power Platform and Azure OpenAI. That is something we'll talk about today
0:26
Again, thanks to Stephen, he mentioned that I'm working many years from
0:33
in Microsoft Partner and Microsoft is pretty good to work there. That is something. We'll talk about how we can get
0:42
ahead of the future proof your business applications with Power Platform and Azure OpenAI
0:48
Now, everyone is talking about Azure OpenAI and all, but how we can integrate and have
0:54
many customers are talking about how they can use Power Platform Azure OpenAI
0:59
I have to give some kind of caveat because not everything will
1:05
work because of the different regions, but I will give some inputs what I have for you
1:10
Now, what we can talk about is quite simple about the Azure OpenAI service and
1:18
difference between Azure OpenAI and OpenAI. They are two different entities. Although Microsoft have bought a lot of good services with OpenAI
1:28
which is a small company, now they have using all the services of Azure OpenAI
1:33
and main differences, and why we as Microsoft people, we use OpenAI as a slightly more stronger way to
1:41
use as enterprise applications, how we can use it. Then we also talk about how the Power Platform
1:48
can utilize Azure OpenAI. This is something we will talk about. Then we see some patterns
1:56
which use cases and which particular application you can work with the Power Platform
2:03
You at least have an idea. Again, in 35 minutes, it's very difficult to show everything
2:09
but I am sure that at least you have some beginner information about the architected design
2:16
and how the Power Platform and Azure OpenAI work together as cohesive applications
2:21
Again, some resources I will share towards the end. Again, I will share all the things like that to
2:29
Simon and he will share with you on that later. Now, first thing is about Azure
2:36
not open AI, Azure AI. Azure AI is there for many, many years
2:42
What we are trying to do as a Microsoft employing also for all the end-users and the customers
2:50
that we are to try to have three different clouds, right? We have our Markets 365
2:56
we have Azure and 365 Dynamics, right? So these are the three clouds we work together
3:02
And the applications. So applications can work together as your power platform or some third-party applications
3:11
if you want to integrate as a partner solution, right? Now, what happens is that many customers or the other customers
3:19
they want to integrate just not as a low code, but also part of the third party applications
3:27
or maybe they want to integrate with Azure. Obviously, with Azure AI, we can integrate
3:33
We can use the low code solution with your bot services, with cognitive search, with a form recognizer
3:42
or with any third party application you can integrate. So it's kind of a limit for that
3:48
Also, as a developer, you can integrate with many applications. Entire service we have
3:57
we have cognitive services, we have custom vision, we have language things
4:02
everything we can integrate. But what is new, again, what is new is Azure OpenAI
4:07
which we are talking about today. Now, what we have in the last couple of years
4:14
maybe this year as well, that we have a lot of uncertainty, right
4:17
A lot of, a lot of changes happening. We have wars happening and then a lot of inflation
4:23
So what we're happening that as a normal user, not as a company, a normal user
4:30
how to think about the cost, right? So inflation pricing process, very, very big thing for normal people, right
4:38
General politics about the wars happening and supply chain is actually affecting everyone in the world
4:44
right and the third thing which is everyone right now is not anxious
4:49
so we're having a lot of challenges at this point of time and how
4:53
the Azure and AI can help you right so what I want to show you that
5:00
many many customers have these samples to you that many companies like E&Y, PepsiCo
5:08
Progressive they are using they are using a lot of lot of services for
5:13
your customers. Example, they are using a manual processing previously for one of the very big
5:21
customer in Europe. And with it, not Azure Open AI, but Azure AI, they're saving
5:27
a lot of money, right? So cost is a very, very big factor
5:32
Second is a PepsiCo. Everyone knows about PepsiCo and they have lot of
5:35
lot of models. They have lot of data scientists and Having those kind of Azure AI cases
5:43
they actually speed it up using their new models, with a lot of new changes in the algorithms
5:52
Progressive is another company which is using chatbots. Now, chatbots are there everywhere
5:56
They're saving money for $10 million. There are many, many other use cases I'm talking about
6:02
but this is the thing which people know what's happening in the world
6:07
They have more cases I can share with you, But these are some few samples
6:11
Now, one thing which we have to take note is that Azure AI
6:17
how can work together? What I have to appreciate here is Azure OpenAI
6:24
can work with your chat GPT at human-like language generation. We can have Dolly for image generation
6:32
We can have Codex, we can have a code generation and chat GPT which everyone knows what's there
6:38
So if you are under the rock, you don't know. So again, OpenAI is something which everyone is talking about
6:44
Now, what I want to share about that OpenAI, again, my entire talk about OpenAI and Power Platform
6:52
So how OpenAI is actually revisiting all the customers, just not in Asia, but all across the globe
7:01
So we are talking about the co-pilot services. We're talking about the, if you see in LinkedIn, if you write the chat messages, it creates content, right
7:16
So it will help you to write your messages, right? I actually write two weeks back, I was talking to someone and then it gave me all the content, this amazing thing which I've seen
7:26
So try these things. Now, what Microsoft is trying to do is trying to integrate the OpenAI, which is a startup
7:33
and to integrate with all your Microsoft services. Whether it's a power platform, whether it's your dynamics
7:41
whether it's your co-pilot, everything is there. So what we are trying to achieve is that all the people who have no knowledge they can also build applications by not writing any single end or code That is thing which we are trying to achieve
7:59
Whether it's modern work, whether it's a data and AI, everything can work together
8:03
Again, people who have no knowledge, they can also write applications. Now, one thing which I've always get this question
8:12
from many, many customers and also from community people, is that what a difference between Azure OpenAI
8:18
and OpenAI differences. Again, you see the sticker slide, you can take a photo
8:24
but I will share this slide deck to you. But a lot of major factors
8:29
a lot of other differences, but one major thing is security. Security is here
8:37
Now, what happens is that when you use OpenAI, chat GPT and openAI.com
8:45
it is free for you. Now, why it's free? Because people want to use it and all
8:51
But if you want to use enterprise, obviously, it's not that safe
8:56
It has its own security controls. But when you use Azure OpenAI
9:03
you always have all the goodness of security from Azure. That is something which we are very boast of
9:12
that all the security coming from Microsoft 365, Dynamics is coming from there as well
9:18
If you want to have secured an application, try to use OpenAI
9:23
That is something which I always see. Other services are talking about integration
9:30
Now, OpenAI have nothing about the integration and about the security. There's nothing mentioned on that
9:40
But when you use Azure, obviously, we can have our own VNet
9:45
our own firewall and we can rest it the ip addresses right this is not available in open
9:51
ai i mean the normal open ai right and the third again there are many features but then which is
9:56
why i will always share is that um everyone can access the open ai right although you can pay
10:03
the paid content right which is like a cost there but when you use your models um whether it's a
10:11
Da Vinci or Dolly, you always have to create an application and then using the application
10:19
you can build that application with any app service, whether it's Azure services or Power Apps
10:26
So you have to create a service. So you have an account for that
10:30
So these are the things which are something which we are very proud
10:36
The one thing which we are working on is about the number of regions
10:40
Right now, we have use cases, regions, which is Western and Eastern America
10:47
Now, just in this month, July, we have released Asia. So you can create Azure Open UI services in Asia data centers, regions
10:57
So something which we have already expanding our footprint in Asia. Now, that is everything about Azure OPI
11:05
Let's talk about the low-code. Now, one thing which we have always seen is that we were having lots of challenges in just in a few years
11:17
The problem is that we will have a shortage of the users
11:25
So we always have a developer shortage. And creating a new application always takes time
11:32
And compliance is, again, a big issue. So what we have seen from the, again, it's not from Microsoft, it is coming from this Forrester and the things, that we need around 750 million applications in the next three years
11:47
And obviously, we can have so many developers, right? I'm talking about the hard code developers who write C-sharp code, Node.js
11:54
And that's why Microsoft is focusing completely for low code, no code
11:58
which means all the users can create their own applications without writing single end of code
12:05
That's something I'm very bound to share with that, right? And 70% will be created by the enterprise, right
12:14
Many people said, is our platform only for low code? Is it designed for enterprise
12:21
The answer is yes. It could be complex because there are a lot of nitty-gritty things will be there
12:29
But the answer is, if you want to integrate your Power Platform
12:34
with enterprise applications, the answer is yes, you can. But you have to think about some use cases for that
12:40
I will talk about use cases there. But one thing which we are very proud of
12:45
that we are number one in terms of the low-code. I guess we have a system like OutSystem, Mandix and all
12:52
but we are number one in terms of the usage. and 33 million users are there
12:58
Every month we have millions of users. Which shows things that many people are using Power Platform
13:05
We're not telling about the third-party applications like I mentioned here, but the usage and the resource which Microsoft provides
13:13
is massive, right? And we have around 97% of the people are using, Fortune 500 people are using 91% of this thing
13:27
So Fortune 500 people, 500 companies are using 97% of this, which means that many, many people are using the Power Platform
13:38
Now, it could be low scale or could be enterprise. So it could be that number is not there
13:44
but you see the value that why we are pushing for low code because many, many applications can be done by the business users
13:52
Okay. And that is a we are needed for that. So I just want to share about this thing
13:57
Now, let's talk about some use cases, right? I just want to give some resources later on
14:04
But let's talk about actual use cases, which many customers are utilizing
14:09
So there's one German company called a Starbucks. And you know, in the construction field, if you make a small pillar or something, and
14:20
and you're not careful, then you have a huge cost for that. Right
14:25
So what this company is doing that using AI model, right? And they're talking about the risk management solutions
14:33
So anything which is a construction companies or hospitals, schools, are we on the land is okay or not
14:40
So they have a lot of algorithm and they are using with Azure OpenAI services
14:45
And what the best thing about that is that within three months
14:50
they are predicting around 80% accuracy. That is a beauty of OpenAI
14:57
Something which is really, really good. You see the accuracy is so much
15:01
And again, we are always improving the AI models there. Carmex is a very big company in US
15:10
And there are a lot of the tools they do. And what they're trying to do is that they want to utilize
15:17
the OpenAI services. So just like Transcripty, you ask some question, a specific car and about all these features and all
15:28
it will take a lot of time. But if you just type a particular car and about all these features
15:35
just within few seconds, it will give you all the things, whether the mileage is there and how much of
15:42
the value after you sell the car everything can be done within few seconds as a text summary So all the research will come within a few seconds And that is the beauty of again I talking about different use cases
15:56
One use case is about construction. Now we're talking about retail, about automobile services, right
16:04
So this is something which we are integrating these services with large customers, right
16:12
Then third one is another very big company, which is called Treland
16:17
And they're using a lot of algorithms. There are a lot of data. And the main thing for them is to have a data security and the residency
16:27
The data should be within that region, not outside. So here, what they're trying to do is that they want to have that data integrity features for their own enterprise customers
16:38
right so the plan is actually migrating these solutions with their own cloud services they're
16:46
using some other another cloud and because that cloud doesn't have that service and they are
16:52
migrating that solution to azure open ai so i don't say the name of the other cloud service but
16:58
they're using azure open ai that is the thing we are actually trying to and uh engulf for the lot
17:04
of AI services from other third-party applications. Migration is something which is very, very crucial
17:11
The last use case, again, that is a very big oiled field in America
17:20
What they want to do for their own use cases in the public side
17:24
and also in their internal side, they want to create some chatbots
17:30
and they're using something called cognitive services. Okay. So these are the use cases and what they after integration
17:38
and have some research and so design, what they actually designed is that they want to have
17:43
from open air service to train their chatbots. What chatbots, maybe their internal services
17:51
Okay. Maybe they want to have some kind of oil rich, how much oil is coming through
17:56
So those kinds of services they're doing. Second thing is they have a public website as well
18:03
They want to integrate Q&A with that service as well. So public and also internal
18:09
They want to do that. So using that particular method only for a few months
18:14
they're able to have huge time savings. Rather than having a human person
18:19
always entering the chatbot features or entering these things with some of the third-party applications
18:25
they have wasting their time. But now with open AI service, they're using that a lot of services with a few seconds
18:34
So again, I'm talking about some scenarios, some kind of use cases, which makes sense if you are trying to use the open AI services with Azure or with Power Platform
18:48
Okay, now we talk about some architecture designs. right now I talk about a couple of services we can integrate. The first one is obviously a document
18:59
process automation service, right? This is a very large customer we are talking about
19:05
It's again in the US and they have a lot of, lot of services. And it's, again, it's an oil
19:13
cash company there. And they have a lot of documents, right? So documents from different
19:18
vendors about different companies, different countries also have. So they are using some document generation, right
19:28
And form recognizer, right? This is Azure service, right? Form recognizer. Now, what's happening is that earlier there was no open AI service
19:39
So they were always using Azure functions and Cosmos DB because the content is huge, right
19:46
And they found that the service is not that straight and there's no content generation
19:52
is not happening that frequently. So what they did, they create a new service
19:57
and they call it Azure Open Service. They use Azure function and the data is coming from your Cosmos DB, right
20:05
So all the prompts, somebody asked some questions about a PDF document
20:10
or something that states can create a document from that year, generating the quality matter
20:18
for a specific company. Now, having that kind of content is very tedious
20:26
but now thanks to OpenAI, we just type something and then it will generate
20:30
the content for you. Now, it could be your text yzer, could be your face recognition
20:37
could be anything else, right? So that's why this is one of the design patterns
20:41
I'm talking about, And if you feel that if the data has to be shown as a web applications for your end users
20:48
obviously you can do. This is for consumers. This is for your backend systems
20:55
This is for consumers. So this is some design pattern which we can use, integrate with Azure services with your OpenAI
21:03
The second use case, again, a different design pattern is using the caller
21:14
This is a European company where they want to have very large telecom in Europe
21:22
What they want to do is they want to have some text translations because they have Polish
21:30
they have French and all. That thing is not working that seamlessly
21:35
About voice calls is not that clear. They have some call center
21:40
Now this line is not very clear, but we always have a call center
21:44
We always have a storage set. The main challenge we're facing is that if somebody has voice call
21:51
the voice call is not very clear because they're using some service and the speech to text
21:57
it's not working fine for them. What they did, they tried to integrate
22:02
Azure OPNS service. And then found that after a few, again, again, we have to train for that
22:09
After a few trainings, they found that it's much easier to have this kind of contact center
22:15
and speech recognition center, which giving a lot of transcripts. So again, think about just not a telephone
22:22
but think about a large enterprise customer where they have a video call and the transcript
22:28
and they want to generate. So previous method didn't work for them, so thereby they integrate Azure OpenAI service
22:36
And this system actually stored into their CRM system. So this is, again, few of the use cases where OpenAI can work together with Azure
22:48
or it could be your third-party application. Maybe you have your own custom website and you want to integrate
22:54
The answer is yes, you can integrate with this. So this is one of the architecture design which we can utilize
23:00
Okay. Now, one thing which I have again, last geared towards the Power Platform
23:07
Now, we have a lot of customers. We keep asking that RO, we have a lot of product customers and we want to integrate
23:14
the PVA, the Power Virtual Agents, but not using the public sites
23:21
but we want to have our own internal. Have one very large customer
23:26
is an insurance customer and then want to have their own data
23:29
and they want to see how it works for them and all. So what we did with partners
23:35
that we use Power Automate to fetch the information and we integrate with the Azure OpenAI service Right now Azure OpenAI service won work till you have your common services working as well
23:48
Now, I can give more details, but again, to save time that if you want to have more questions
23:55
ping me, I will share my LinkedIn, ask me questions. I can share more knowledge on that
24:01
But just that this customer want to have their own data in Azure
24:06
The main question is that using a chatbot, the question can be asked by the storage
24:14
and then cognitive search. Think about a medical. This is a medical company where they want to have
24:20
all the data as questions and all, and they use cognitive services
24:24
a cognitive search, and then using a prompt, prompt engineering, they're fetching the questions
24:28
and the user can actually ask the questions. Then again, as a chat GPT functions
24:33
it worked for them. So again, this is some use cases which Power Platform and as your Azure developer, as an AI developer, you can integrate
24:46
And I have a repo link with that and Simon will share that link with you later on this
24:54
Okay. Now, before I proceed, I want to share something very, very important topic, which is called a learning thing
25:03
Right now, power platform, a lot of people know Azure OpenAI. Maybe a lot of people don't know
25:09
Right. So I'm going to share some learning material with you and you can access
25:16
those things in Microsoft Learn. And also there are a lot of videos which are there with that
25:21
Right. The first thing first is about Microsoft Learn. Right. So I hope everyone knows what is Microsoft Learn and you can use just not
25:30
of a power platform you can use for Azure, you can use for your Dynamics
25:34
you can use for data and AI, everything you can in that link you can see
25:40
It can go from beginner to expert. Let me share with you
25:48
Then for the power platform specifically, we have many certifications. If you are started with the power platform as a beginner
25:57
I will highly recommend to use PL-100. Then you can go for associate levels
26:04
and then go for PL-600. So it gives you slightly more understanding
26:15
about how the Power Platform can work. Now, if you're new to Power Platform
26:20
please take time. Don't rush. And these exams are not that difficult
26:27
are quite difficult. So take your time, read things, go watch videos
26:34
practice some papers, online papers, and then if you're not an Amcity, the exam cost
26:41
will be quite high. So please save time, invest in your learning, and then write exams
26:49
I have seen many, many people fail the exam because they didn't have prepared
26:56
Take some time for that. Second is something called Azure. Again, what I will highly recommend
27:03
so this resources which is called in your screen, is that how we can work on the Azure AI
27:12
Learning Azure AI, anomaly detector, how to use cognitive services, how the API works and also lot of things are there inside
27:24
That is my things. I have some time, Simon, a few minutes more
27:36
Simon, can you hear me? Yeah, you got 10 minutes, Aroh. Yeah
27:44
You got 10 minutes. Yeah, I got 10 minutes, awesome, awesome. Yeah, so I want to share something with this
27:54
Okay, so I'm gonna share some other, few minutes. I just want to get the knowledge about how things works. There's one course
28:05
as you open eye. And I will share this link with you as well, you can have a look at this
28:16
and yeah. Okay. So this is the one course I have completed this thing and I found this is one of the
28:27
very amazing course. And I will, here I could share this link, right
28:41
Oh, okay. Yeah, there's something which I always go for. Obviously, you can go for Microsoft Learn
28:50
but I have a habit of learning and practice. So what I do, I always create samples
28:57
I use some samples, and then I try to practice using that particular model
29:01
So it's something which I go for. And if you want to learn about OpenAI's
29:13
So this is another very good module
29:24
They have a lot of content there where you can understand how things work
29:29
They have a couple of videos as well. Try to see this thing because it will make life easy if you are new to OpenAI
29:43
Okay. Yeah. That is something about OpenAI
29:54
But if you want to have more understanding, I will highly recommend go to a lot of videos
30:01
which I will share with you. things again from Azure features about how
30:43
we integrate your own data with the OpenAI. A lot of content is there
30:49
You just search for OpenAI, a lot of good content are coming through
30:53
I always appreciate the people who are contributing and sharing the knowledge for this
31:00
I'm actually going to run three sessions on OpenAI, part one, part two, part three in the next two, three months
31:13
And if you are interested, you just buzz me. I will actually share the links with you later on
31:17
So I'm starting from next month onwards. But if you want to have more knowledge
31:22
go for this resources which I shared with you
#Distributed & Cloud Computing
#Enterprise Resource Planning (ERP)
#Open Source