Dinakar
Great and uh hello everyone and welcome to today's webinar on AI augmented app testing strategies to accelerate delivery my name is dinaka and I'll be the host for today and before we go get.
Dinakar
Started I would like to go over a few housekeeping rules your line is currently muted however you can submit your questions using the Q&A option at the bottom of your screen we will answer.
Dinakar
Your questions at the end of the webinar in the Q&A section please not note that this webinar is also being recorded and we will we will share the link of the recording with your.
Dinakar
To share with your colleagues later with that said let me quickly introduce the speaker for today Avinash is a renowned thought leader a recognized keynote speaker and the co-founder of Pcloudy he carries.
Dinakar
A rich experience of for about 20 plus years of experience in product development and testing he brings with him a passion for in emerging technology and quick adoption both of which have solidified his.
Dinakar
Reputation as a leader in the app testing space. Without further ado, let's hand it over to our speaker, Avinash Tiwari.
Avinash Tiwari
Thanks Dinakar, uh, thanks for the introduction, um.
Avinash Tiwari
So good evening good afternoon or good morning depending on where are you joining from and a very warm welcome to all of you I really excited to be in front of you and thanks.
Avinash Tiwari
For joining in such large numbers it shows the interest in uh this topic and and the kind of change kind of impact it's creating within the industry uh as dinakar mentioned my name is.
Avinash Tiwari
Avinash and I'm one of one of the co-founder of Pcloudy which is part of opy and over the last two decades I have been at the intersection of product building Tech and software.
Avinash Tiwari
Quality and I've seen so many business Transformations Tech Transformations but uh having seen all that I can I can say that we are clearly at an inflection point where there are huge uh changes.
Avinash Tiwari
Happening around uh Us in every area of our life and that's because of the advancements in the AI technologies that we have seen last couple of years and that's one of the reason I'm.
Avinash Tiwari
I'm super stoked today to share what I know and also uh to know from you uh through your feedbacks through your questions.
Avinash Tiwari
So please feel free to just drop in your thoughts suggestions feedbacks share or questions uh during the session uh I would love to know about what your thoughts are and.
Avinash Tiwari
Also if there are questions love to answer towards the end of the session uh so use the Q&A panel as freely as you can.
Avinash Tiwari
So having said that uh a quick round of introduction for the product some of you who are uh who might be new to PE cloudy uh Pcloudy is a unified platform for continuous automated.
Avinash Tiwari
Testing and we are one of the first un unified platform to bring uh the automation creation as well as the whole automation infrastructure in terms of mobile device Cloud browser Cloud at a single.
Avinash Tiwari
Point combined with test management so that gives uh organizations an ability to uh run their test continuously uh in a agile fashion and.
Avinash Tiwari
Then execute it on a large infrastructure that we provide so from that perspective uh uh I would welcome all of you to go and try the product and and see it for yourself um.
Avinash Tiwari
We have have been in Industry from last uh 7 years almost uh we are trusted by some of the largest companies in the world and.
Avinash Tiwari
So far we have served close to, plus customers globally we are recognized by many of the analysts as well as uh our users have given uh extremely generous uh ratings which shows the love.
Avinash Tiwari
For us uh on on Portals like G2 crowd and you can see the badges on the slide here and we're very proud to say that we have crossed almost 8 million plus minutes of.
Avinash Tiwari
Testing on the platform so you can uh imagine the amount of uh testing happening daily on our on our cloud in terms of both uh automated execution as.
Avinash Tiwari
Well as manual testing of uh different kind of applications uh be it mobile or web uh so with that uh introduction out of the way uh I wanted to uh focus on a quick.
Avinash Tiwari
Recap of AI Revolution that we are witnessing right and needless to say I mean AI is everywhere we all are experiencing in in different sphere spheres of Life be the students the the the.
Avinash Tiwari
Education system to programming to uh management functioning uh to our medical uh uh systems everywhere we are seeing use of AI.
Avinash Tiwari
So AI definitely is going in a space of democratization where it's easily accessible to everyone and it's making a real impact in our lives but was it the same uh some time back we.
Avinash Tiwari
Have been talking about AI for quite some time and this is me five years back uh presenting in U uh software testing conference one of the largest testing conference in India and the the.
Avinash Tiwari
The topic was is AI forcing us to REM reimagine the future of testing and uh uh we have seen seen the shift but 5 years ago AI wasn't.
Avinash Tiwari
So simple right um we the computer vision models or the machine learning models around supervised learning were quite popular uh but accessing those models building those models from scratch was very hard it required.
Avinash Tiwari
Deep technical skill set and uh the use cases especially in the testing space or the development space uh were not very common uh from from the from the perspective of available models that existed.
Avinash Tiwari
That point of time so we could use uh AI in a very limited way at that point of time but come 2022 and November 20122 and until you have been living under the rock.
Avinash Tiwari
You know what I'm referring to it's the launch of open AI chat GPT and that kind of changed the game uh from that point of time we have she we have seen a tremendous.
Avinash Tiwari
Growth in the space and uh this graph that you see above the line are the open source model that came into the uh came into the space and the below the line are the.
Avinash Tiwari
Commercial models so you can see there are multiple AI models that exist in the market and so many are open source which gives a lot of options for developers to play with which didn't.
Avinash Tiwari
Exist five six years back when we started our journey um and now the AI models can be accessed using simple API calls uh.
Avinash Tiwari
So in that sense accessing AI building applications on top top of AI uh has changed completely and that's the reason we are seeing such rapid progress uh into all the uh spheres of our.
Avinash Tiwari
Life uh where uh we we are seeing use of AI uh quick uh word on where do we exist on the hypee cycle I mean this is a good way to understand what's happening.
Avinash Tiwari
With this new technology any new uh technology that comes it goes through this whole hype cycle it's a good representation how it's very difficult to predict how AI is shaping up.
Avinash Tiwari
Right now uh I mean I don't want to take that chance of predicting but this is a good way to understand where do we uh stand.
Avinash Tiwari
Right now so so once uh in last one year there has been inflated expectation and that period is kind of over where uh we are reaching towards trough of disillusion illusionment which means.
Avinash Tiwari
Now organizations are realizing the real use cases where AI can be implemented and also in a lot of cases the outcomes that were expected uh uh is is not materialized.
Avinash Tiwari
Right now the cost of implementing at a lot of places is high very high uh and and we know that the compute cost involved in building models and deploying models.
Avinash Tiwari
So it's going through uh this truff and uh it's predicted that Beyond this truff uh people will realize the real use cases where AI can be implemented and.
Avinash Tiwari
Then we'll see a real productivity gain um so so the the point that I'm trying to highlight is this is uh the time to really assess what are the areas where uh uh AI.
Avinash Tiwari
Can be implemented lot of prototyping small uh uh utilities small business cases can be implemented and then we can figure out what what are the real use cases where real productivity Gaines can happen.
Avinash Tiwari
So this is a good representation of where do we stand right now so beyond this fluff right I just wanted to cover some Basics uh terminologies and and some of the uh points that.
Avinash Tiwari
I'll be talking about so just wanted to clear some of the points here or or the basics here some of you might be knowing already.
Avinash Tiwari
So just bear with me uh so EI as we know has existed for quite some time I mean machine learning computer vision uh have been there for quite some time and we have seen.
Avinash Tiwari
The uh use cases emerging from last uh five six years uh we have seen you have used uh YouTube recommendation Netflix recommendations uh spam filtering in Gmail autocorrection of emails I mean some of.
Avinash Tiwari
The models have been in for quite some time and we have we have been part of that Journey from uh quite some time but it's the generative AI which is which is about creating.
Avinash Tiwari
New content right earlier models were mostly about classifying existing content classifying existing content and predicting on top of it here generative AI is all about creating new content uh in different formats that's the.
Avinash Tiwari
Real game changer which we have seen in last one year or so and and that is uh helping us move at the rapid bace uh in the AI space and this is based on.
Avinash Tiwari
Uh large language model as we all have heard this term llm um so llm basically is about uh analyzing sequence of text its context its reasoning and.
Avinash Tiwari
Then predicting the next words so for example if I T type sky is llm will predict sky is blue.
Avinash Tiwari
Right and as as we type more words it will add uh the next sequence and and it also understand the context in which you are ask asking uh or providing the the sentence uh.
Avinash Tiwari
Now llm itself so if you talk about chat GPT uh we talk about chat GPT a lot uh chat GPT is not llm uh to be more precise chat GPT is an application which.
Avinash Tiwari
Has been built on top of llm right uh so that clarification is very important now llms are basically the uh Foundation models.
Avinash Tiwari
Right and and what I was saying what I was saying about llm is true for foundation model that Foundation models are the actual uh uh models here which are responsible for analyzing the text.
Avinash Tiwari
And then producing the next sequence of words so llm is just an instance of the foundation models and Foundation models are trained on billions and trillions of uh parameters from the text from the.
Avinash Tiwari
Data which is available across the internet through books there are different sources based on which it has been trained.
Avinash Tiwari
So GPT 4 I mean no one knows the exact number but it is supposedly having 1.7 billion trillion parameters uh which is extremely large and it takes a lot of time for any model.
Avinash Tiwari
To reach that stage in terms of training um so these are some of the basic terminologies that uh I wanted to kind of address.
Avinash Tiwari
So that we we understand and and use the same language uh going forward now in terms of in terms of progress that we have made uh especially around gen and uh llms based application.
Avinash Tiwari
Uh this is a good representation of the journey which is uh happening and this is this journey is uh across last one or one and a half year uh starting from 2022 November and.
Avinash Tiwari
Uh 2024 as we speak today uh so during the initial launch of chat GPT everyone uh figured out uh that if we give a particular kind of uh uh question uh llms can respond.
Avinash Tiwari
With different answers so the prompt engineering became one of the the key Concepts uh within the Gen space and once people understood uh prompt engineering and started playing with it uh using it for.
Avinash Tiwari
Different purposes uh we saw a flood of llms especially in the open source space uh and llama 2 was one of the uh major releases in the open source space which kind of gave.
Avinash Tiwari
So many uh other open source uh LMS its way uh into the market but those uh llms coming into the Market uh opened up the space and uh developers soon soon realized that we.
Avinash Tiwari
Can build real applications on top of it I think that was the uh moment where real shift started happening uh in the Gen space and that kind of uh led to the emergence of.
Avinash Tiwari
What we call it llm Ops uh which means uh emergence of tools uh Dev tools uh uh different kind of databases hosting Frameworks for llm generated applications.
Avinash Tiwari
Right and uh initially again it was uh applications which were built like cool projects for example resume generator or summarization of text but soon uh again developers started looking at uh using these llms.
Avinash Tiwari
For their Enterprise needs so for example how do I build something which is dedicated or which can help me in my uh work area.
Avinash Tiwari
Right uh not just mine my organization as well and that that created a lot of disappointment as well because llm was uh quite generic in terms of its data and if I use Enterprise.
Avinash Tiwari
Data I run the risk of sharing the data to the uh uh outside world so there were a lot of challenges that came across but.
Avinash Tiwari
Then they came rag which which is retrieval augmented generation and that again gave a boost to the overall application development.
Avinash Tiwari
So rag in a very simple way is where you can provide your own data and ask llm only to focus on that particular data to answer and not go to the uh generic uh.
Avinash Tiwari
Uh data points that llm has uh this really uh made a huge shift in terms of how anyone can create Enterprise applications focused on their own needs or the organization needs um.
Avinash Tiwari
So rag was a big shift uh which happened towards end of the uh 2023 now in 2024 uh we have started seeing use of multimodel AI.
Avinash Tiwari
So so far what we have spoken for example you can build an application on top of chat GPT or you can build an application on top of Lama.
Avinash Tiwari
Now uh developers have started experiment with multimodel AI there are some uh llms which are good in a particular area for example code generation there are some who are very good with reasoning there.
Avinash Tiwari
Are some who are very good with images how can we combine uh these different models to create a end one workflow.
Avinash Tiwari
So that's where the rise of multimodel AI has started happening and that also got pushed because of the new framework that has come in the market which is called agentic workflows which which means.
Avinash Tiwari
You can create different agents interacting with a with a particular task and those agents combin uh to create a end to end workflows.
Avinash Tiwari
Right and if you have heard about Davon uh the code generation code generation uh engineer which was uh released the the the teaser was released few I think few months back uh that kind.
Avinash Tiwari
Of follows this whole agentic workflows approach and it's becoming much more popular but the the latest one which is happening which is quite significant for people like us who are looking to build Enterprise.
Avinash Tiwari
Specific uh uh applications or to solve Enterprise specific uh use cases uh is the arrival arrival of smaller models like Mistral.
Avinash Tiwari
Now uh as I was talking about GPT has 1.7 trillion parameters now if you have to uh deploy that kind of model within your Enterprise the compute cost uh the hosting cost could go.
Avinash Tiwari
Into millions and it might not be worth the ROI and that is where some of the smaller models with million parameters which can just serve the need of your Enterprise you don't need those.
Avinash Tiwari
Many parameters to for your Enterprise use case uh are becoming very popular and I think it will help everyone uh build some of the Enterprise use cases much faster uh at a manageable cost.
Avinash Tiwari
So this is something which is happening as we speak uh and we in 2024 I think we'll see more advancement into these areas like multimodel agentic workflows or or more smaller models which will.
Avinash Tiwari
Help us build more practical use cases uh with a with much more realistic rois and the budgets for the AI projects.
Avinash Tiwari
So with that progress I think uh probably you you would have gotten sense of where do we stand right.
Avinash Tiwari
Now um uh we we are seeing so many applications and I'm sure you you would have heard about many of these.
Avinash Tiwari
So we have ai chatbots we have meeting summarizers we have code compilation platforms we have presentations content creation you name a field and there are uh these gen applications available in the market just.
Avinash Tiwari
To give you a a sample of what exists in the world this is just kind of a very very small summary of kind of use cases which can be uh served uh with J.
Avinash Tiwari
Now our focus is uh how we use this into the world of testing right and uh I'll talk about what we have done for Pcloudy that will give you a sense of uh.
Avinash Tiwari
How uh it can be used uh but what we are looking at is what are the AI testing strategies with respect to gen and in general uh the some of the previous AI approaches.
Avinash Tiwari
Uh can be used to really speed up your testing delivery and uh overall application life cycle itself so this is one of the diagram or the the graph presented by Gartner and it kind.
Avinash Tiwari
Of uh represents it well where do we stand so we have seen the testing uh Journey from waterfall to Agile to devops uh to AI augmented which is the the latest phenomena in the.
Avinash Tiwari
Testing right and when we say a augmentation which means means uh we are talking about improving the testing efficacy uh while reducing the delivery cycle significantly.
Avinash Tiwari
So the top line that you see represents the the delivery uh cycle so it has reduced from it has gone down uh from years to.
Avinash Tiwari
Now within hours right while the testing complexity or the rate of change has increased and the testing efficacy expectations uh have increased significantly.
Avinash Tiwari
So how do you balance that uh is the question that we are trying to answer here so this is the phenomena where we stand.
Avinash Tiwari
Right now where expectation is to reduce the delivery cycle um on on significantly it could be depending on the organization it could be weekly it could be daily it could be hourly.
Avinash Tiwari
Right while the testing efficacy has to increase and that can only happen when we start using a augmented uh techniques.
Avinash Tiwari
Now this is another uh representation of uh how uh AI augmentation is helping uh in the overall uh death Cy I mean overall software development life cycle itself um and and you can see.
Avinash Tiwari
Huge number of uh applications of uh AI augumentation across different phases for example in coding I mean obvious automated code generation but within the testing space automated unit test case generation automated UI test.
Avinash Tiwari
Cases generate generation automated API test case Generation all that uh is very obvious in the in the coding phase.
Avinash Tiwari
Then on the planning phase a lot of organizations are using it to create uh their their test strategies um uh test cases from user stories uh using it for General productivity gains like summarizing.
Avinash Tiwari
Minutes of meetings um in the in the uh operate or the deployment phase we are talking about um predicting change failures.
Avinash Tiwari
Right uh there is a huge area coming up in the in the space is test observability how do you observe the system when the the uh system is in production.
Avinash Tiwari
Right and there there's lot of scope to use AI to build anomal detection uh and and find out where could the potential failures be and if there are failures pointing out the real uh.
Avinash Tiwari
Reasons of the failures immediately without kind of developers looking at it so there are a lot of areas uh in the operate and measure phase where uh anomaly detection pattern recognition uh use cases.
Avinash Tiwari
Can be used so you as what I want to highlight I mean you can you can go through each of the test case each of the use case here what I want to highlight.
Avinash Tiwari
Is that at each stage of the life cycle we have use cases which can be built using uh Ai and that's where real AI augmentation happens.
Avinash Tiwari
So you have to start looking at uh in your uh life cycle what are the areas that you can start uh planning for.
Avinash Tiwari
Now coming to specifically into the space of testing right uh because that's our core Focus um this is again uh projected by Gartner on the.
Avinash Tiwari
Right side the the chart that you see is a survey that we did with our customers uh in terms of what are the areas where they have started using Ai and that gives you.
Avinash Tiwari
A sense of uh where most of the organizations have started using Ai and getting a good benefit of AI augmented AI augmented uh techniques.
Avinash Tiwari
So Gartner clearly uh five areas uh test plan and prioritization we talked about it sdlc as well uh test creation and maintenance U.
Avinash Tiwari
So from user story test creation uh NLP based test generation uh third test data generation this is something which where I have seen a lot of organization adopting AI.
Avinash Tiwari
Because synthetic test generation has been a key Challenge and I think with uh the AI models uh this problem can be solved much quickly visual testing and this has been there in the market.
Avinash Tiwari
For quite some time because it it is it is using the computer vision model nothing to do with Gen as such visual testing uh is one of the largest use case we see in.
Avinash Tiwari
The industry in terms of AI augmentation and then test and defect analysis this this is where a lot of those ml uh models I was talking about in the beginning were being used even.
Avinash Tiwari
Five years back uh that how do we predict the defects right uh we have the defect data can we draw some pattern out of it.
Avinash Tiwari
So test and defect analysis is something which has been there in the market but it's it's growing because of the ease of use of the models that exist in the current scenario.
Avinash Tiwari
Now look at the chart here again it kind of gives the same sense visual testing manual test case generation impact analysis um automated test uh script generation object identification cell feeling script maintenance using.
Avinash Tiwari
Cell Feeling Again a big area in the testing where AI augmentation is helping a lot um and again test data creation.
Avinash Tiwari
So you can see the chart it kind of maps to what gutner has been projected which means these are the areas where customers have been using AI uh augmentation for quite some time and.
Avinash Tiwari
And getting some results positively now again uh this is slightly elaborate uh uh result of what we s just saw in the chart uh these are the use cases which were given by our.
Avinash Tiwari
Customers in the survey that we did so synthetic test data generation visual testing bug classification uh Predictive Analytics uh test case generation uh auto trigger in cic based on um some of the parameters.
Avinash Tiwari
That exists uh server configuration planning sell feeling are some of the areas that you see listed here uh have been used by our customers and I think if you have not started thinking into.
Avinash Tiwari
Those areas you can uh take a look at that and then start it start planning in your uh organization.
Avinash Tiwari
Also um so good list of practical use cases in in the testing space and uh uh we give you uh food for thought if you have not started still.
Avinash Tiwari
Now um with with some generic customer use cases that we have seen in or the Gartner predictions that we have seen how we have implemented uh some AI stack within PR clouding.
Avinash Tiwari
So as I said we have been working on uh implementing AI for quite some time and over a period of time we have built uh AI stack on top of our core platform.
Avinash Tiwari
So just to H give you a Clarity of uh what the AI stack is and how it it fits in the overall ecosystem for Pcloudy uh.
Avinash Tiwari
So poud CL its score has three pieces one is a codeless automation where you can create scripts and if you're not using codeless automation you can use open source Frameworks like AUM slum.
Avinash Tiwari
So creation of scripts either through our codeless platform or through uh your existing uh open source ecosystem then the second piece is execution where we provide a com comprehensive test infrastructure Cloud uh with.
Avinash Tiwari
Huge number of mobile devices and browsers on our Cloud both uh in form of public or private instances available for customers.
Avinash Tiwari
Right so you create your scripts and execute on a large infrastructure without any any challenges in a very seamless way uh without writing any other piece of code uh when you build the scripts.
Avinash Tiwari
And third piece is the analyzing where we generate so much insights because the scripts are running our on our infrastructure and we are getting.
Avinash Tiwari
So much data so we have we provide very interesting insights in terms of how your application is performing when uh the test cases are executing on the cloud U how much pass fail uh.
Avinash Tiwari
Aggregation in a very uh interesting way across different bills uh so there are third pieces analysis of the uh results which are going on and you can do it in a cicd mode uh.
Avinash Tiwari
Without any manual intervention and again it it has integration to various tools which are required within your Enterprises so from jira to slack to Microsoft teams to different project management tools different cicd systems.
Avinash Tiwari
More than 50 plus Integrations have in place so that's the core of our platform now what we have done is we have built a AI stack and we have done it in a way.
Avinash Tiwari
That not only it can be used by Pcloudy we are we can provide access to you to start using it within your Frameworks.
Avinash Tiwari
So there are IML models that we have built uh for example for visual a uh and cell feeling and a interactive AI power assistant for code generation and there are it can do lot.
Avinash Tiwari
Of other tasks as well and on top of it we have built some AI bonds uh which are used for speeding up your testing for example sync test where you can run your tests.
Avinash Tiwari
Across different devices uh while you're performing or man tests and then a certify a bot test where you can run your test you can just upload an application and it will be run on.
Avinash Tiwari
Different device combinations without any uh line of code that you need to write uh it does some kind of exploratory test on your application and.
Avinash Tiwari
Then you get the result uh on a within a span of uh minutes so these are the AI Stacks that we have developed and we are adding more to this AI stack.
Avinash Tiwari
Now one of the interesting thing that we have done is for example at least the a IML models that we talking about can be integrated within your framework.
Avinash Tiwari
So for example if you have developed a selenium or apium based framework or any other tool that you are using uh you can call our uh apis and.
Avinash Tiwari
Then use the visual I uh component within your tools U if it's a appium selenium based framework it's just about adding few lines of capability code and.
Avinash Tiwari
Then visually is integrated so we want to uh the approach that we taking is to democratize the access of uh AI models that we have built for our customers um they should be able.
Avinash Tiwari
To use what we have built rather than building it themselves uh so visual AI cell feeling or the code generation assistant which will launch very soon for external customers while Visual and sell feeling.
Avinash Tiwari
Is already there if you want to integrate with your framework so this AI layer is already inbuilt on our Cloud.
Avinash Tiwari
So if you're running it on Cloud by default to get the access of these but if you want to integrate uh within your tools uh feel free to reach out to us and.
Avinash Tiwari
Then we'll be able to help you provide access of those uh models and then you can start integrating it just just give you a slightly more deeper sense of what we're talking about.
Avinash Tiwari
So visually I I mean some of you might be knowing it's about uh identifying differences on the application when the test execution is happening.
Avinash Tiwari
So it I didn't it compares the uh runtime execution with the with a base image that you provide and automatically detects the differences on the application as your test functional tests are going through.
Avinash Tiwari
Um again a huge gain uh moving from just the functional testing to integrating visual AI within functional test by just integrating uh or inserting few lines of few lines of code within your framework.
Avinash Tiwari
Um sell feeling right for example it's a big challenge for automation scripts where as the application change one of the big reason of fa failures are uh application object changes.
Avinash Tiwari
Right it could be ID change it could be expass changes what we have done is we have built a model U where it identifies the changes as you run through and it automatically heals.
Avinash Tiwari
The scripts and your execution uh continues uh it also provides the feedback in terms of where the changes happened.
Avinash Tiwari
So later you can analyze whether those changes were acceptable or or you want to log it as a bug uh but your execution will not get stopped and you have a choice uh to.
Avinash Tiwari
Either switch on switch off again just few lines of code and your framework will be uh or your your scripts will be sell feeling enabled uh third one is the AI power assistant I.
Avinash Tiwari
Was talking about so it not only does lot of Pcloudy tasks so for example uh you can just say to a assistant that show me all the devices of a particular version list.
Avinash Tiwari
Uh so you can just ask the question in a natural language and it will uh respond and from there itself you can start taking action.
Avinash Tiwari
So it makes the uh usage of platform much more easier plus it has its own code generation engine as.
Avinash Tiwari
Well U which uh you can access within few days as we launch the next version of it and then I was talking about some of the Bots uh.
Avinash Tiwari
So this screenshot talks about lfie which we call uh has the Cod stion engine I was talking about and.
Avinash Tiwari
Then we have some of the Bots like sync test where you can run your test across different devices in the manual testing mode itself.
Avinash Tiwari
So one way is to write automation scripts but if you're just doing some quick testing um uh you made some changes to your UI or you want to validate across different devices you just.
Avinash Tiwari
Come to our platform open multiple devices together and then perform your uh testing and see the result on different devices at the same time.
Avinash Tiwari
So huge productivity gain for anyone uh uh in in such scenarios and here there is a bit of learning that we have implemented from the ml side where as you uh use the application.
Avinash Tiwari
It becomes much more intelligent to identify the actions and replicates it easily and lastly the bot test I was talking about where it does a lot of exploratory testing uh it's it's a pattern.
Avinash Tiwari
Recognition kind of uh uh algorithm that we have built where it goes through the application um activities and and finds out the flows and executes it on its own with different sets of data.
Avinash Tiwari
So some kind of monkey testing follow with some some real flow tests as well however I mean for very complex application it has its own limitation in terms of the number of flows that.
Avinash Tiwari
It can find out but it's becoming better better as uh we uh use it more and more U so these are some of the uh uh AI I these are part of our AI.
Avinash Tiwari
Stack and as I was saying we are adding more and our objective is to give access to our customers in a very uh easy way we don't want to kind of uh tie up.
Avinash Tiwari
With newer licensing and additional cost for our customers uh so in that sense I will encourage all of you to go and check and and if you want to integrate feel free to drop.
Avinash Tiwari
Us a note now we have talked about um using AI uh a augmented testing and there are so many existing tool that exist in the market the point is if what if you want.
Avinash Tiwari
To build it on your own right uh there'll be a lot of use cases that you you think of building on your own there are a lot of use cases for which you can.
Avinash Tiwari
Use existing tools that's much more easier because you can just access it uh with with apis or with few lines of code.
Avinash Tiwari
However if there are situations where you want to build uh your internal use cases you need to follow a framework to figure out what are the.
Avinash Tiwari
Right cases for AI and this these are the four steps again very simplified way of looking at it but these are four steps to figure out whether you really have a use case for.
Avinash Tiwari
AI now the first and foremost is any AI model works on lot of data so do you have a necessary data uh for the use case that you're looking at how big the data.
Avinash Tiwari
Is right do you have uh do you have the data which resembles the real world problem that you're trying to solve.
Avinash Tiwari
So for example if you're building a bug prediction model until you have enough set of bug data across different component of your application the bug prediction model will not work.
Avinash Tiwari
Right so you have to see what kind of data you have is it enough is it covering and a breadth of the problem that you are solving.
Avinash Tiwari
Then only it makes sense to look at implementing AI use case on top of that the second part is about the problem that you're trying to solve.
Avinash Tiwari
Now with the initial excitement of chat GPT or the Gen that came into picture uh people have been using it for a lot of purposes where probably it's not even required.
Avinash Tiwari
Right uh building applications is one part of it but then deploying maintaining has huge cost so you have to recognize is the problem that you're trying to solve complex if it can be done.
Avinash Tiwari
With few rules some statical statistical analysis uh please go ahead and do it that we don't Implement AI on top of it as I said it's complex to maintain and and the ROI will.
Avinash Tiwari
Not be there so look at problem which is really complex to solve for example in case of testing building a self AI selfhealing is is a complex problem and if you're able to solve.
Avinash Tiwari
It will help you in Long Run uh and and for different kind of application so that's a complex problem to solve in the testing side uh or within your ecosystem developing let's say from.
Avinash Tiwari
User story to tesc generation so if you're able to solve it for your Enterprise specific use cases uh it can work for a long period of time and can give huge productivity gain but.
Avinash Tiwari
For example maybe the bug prediction can be solved through some of the um existing models pattern recognitions Etc you don't have to build sophisticated AI for that the third one is is it a.
Avinash Tiwari
Static problem or is the problem changing over a period of time um because for for the static problem there are a lot of rule-based algorithms which are in place which can solve the problem.
Avinash Tiwari
You don't have to really look at using uh complex uh gen modeles or LMS or uh other other ways to do it and last one uh and the last one is do you see.
Avinash Tiwari
Scale for your solution so I think I covered that in the second point that is it a problem that you're solving which is there for a long term you see lot of uh uh.
Avinash Tiwari
Data getting generated over a period of time which will be again used uh for your model then it makes sense if it's about a small problem where uh you have some existing data but.
Avinash Tiwari
You don't see a lot of addition to that data in future doesn't make sense right so you you need to build uh a solution only when you see a scale for resolution a significant.
Avinash Tiwari
Scale for resolution otherwise you'll not kind of get the ROI and again uh just to kind of add on to this framework um if you have found AI problem I think this is the.
Avinash Tiwari
Four step to kind of build one first is of course defining the problem which uh which is about finding the use case uh scalability of the solution we were talking about once that is.
Avinash Tiwari
Done then you choose a data strategy right you have enough data how do you segregate the data between training and validation if there is some missing set of data can you generate that data.
Avinash Tiwari
To uh add uh that missing piece uh what will happen to the Future data generation how will you include that as part of your data strategy.
Avinash Tiwari
So a lot of questions to to answer there once the that part is done then the third phase comes where you.
Avinash Tiwari
Look at Tech or the infrastructure strategy and the tech means which model right lot of times we go reverse there is llm and we want to build something on top of it that is.
Avinash Tiwari
Not the right way the tech or the the the infrastructure strategy comes later once you have identified the problem and the uh the data strategy.
Avinash Tiwari
So Tech is all about uh deciding which ml models suits me what would be the hosting strategy do I need an application on top of the the model.
Avinash Tiwari
So that my users can access the the model easily all that becomes part of your Tech and infrastructure infrastructure strategy and.
Avinash Tiwari
Then finally finding the right skill uh for someone to do that so do you have the right skill to kind of develop or or build on top of the models not only that host.
Avinash Tiwari
Manage deploy Etc so there are a lot of phases involved in this process so these are the four steps that exist U uh today I mean not very exhaustive ones but it will give.
Avinash Tiwari
You a sense of what it takes to build your AI strategy if you're looking to build some of the AI use cases internally and with that I I'm reaching my conclusion here um with.
Avinash Tiwari
AI what we have seen is uh there are there is a complete spectrum of respon responses uh we have seen over a period of time especially in last few years the.
Avinash Tiwari
First Spectrum on the left side is a denial right um that AI cannot do my Z AI cannnot help in my use case.
Avinash Tiwari
Right and we don't have time to look at AI so I'm happy with uh how things are going on and I'm confident that AI will not impact my U current use case on the.
Avinash Tiwari
Right side uh of the Spectrum which is The Other Extreme is the Panic that AI is going to steal my job.
Avinash Tiwari
Right um it's going to replace my use case um by Bots by different applications but we are somewhere in the middle with a if you apply the positive mindset where it's proven with whatever.
Avinash Tiwari
We have seen in last two years that AI can be insanely productive for all of us right if you use it uh.
Avinash Tiwari
Right Way AI can be insanely productive and that will help not just individually but as an organization as a as a team to become productive.
Avinash Tiwari
So what we have to look at is this Middle where start looking at implementation of AI with a positive mindset that we want to increase the productivity um there's no need to be Deni.
Avinash Tiwari
Is it's not good to be in denial it's not good to be in the Panic State the middle state is what we need and with that conclusion uh I think I'll end the session.
Avinash Tiwari
And um we can take question and answer uh we would love to understand any new use case that you have tried which we not spoken about um any any other thoughts suggestions uh I.
Avinash Tiwari
Would love to take up that and thank you so much for your patience and uh uh listening to the listening to this topic.
Avinash Tiwari
So we have uh few questions already um in the chat the question and answer window um the first question is uh can we integrate sell feeling from Pcloudy to my current framework.
Avinash Tiwari
Yeah absolutely I think I spoke about during the slide uh as well that we have built it in a way that you can integrate with any framework uh and if if you have ai.
Avinash Tiwari
Uh sorry if you have APM selenium based framework it's much easier to implement you can just insert some of the capabilities that we provide but if you have built your own Uh custom Frameworks.
Avinash Tiwari
Or if you have your own Tool uh you'll have to integrate our apis so uh the answer to this question is yes uh the method can change depending on the kind of framework or.
Avinash Tiwari
The tools that you're using but you can very well integrate um yeah the other question is around sharing the presentations of course this whole recording itself will be available to uh everyone uh and.
Avinash Tiwari
You can you can use this recording to um go back to the session or share with your friends colleagues um um.
Avinash Tiwari
Yeah it will be available after the session is over if there are more questions I'll be uh happy to take up uh if not.
Avinash Tiwari
Then we will uh conclude we are here for a few more minutes feel free to let your uh I feel free to drop in your questions to the Q in a box I think.
Avinash Tiwari
Uh we will uh we will take up one more questions and then we will conclude not sure if I can see all the questions oh sorry I was looking at the wrong window um.
Avinash Tiwari
How can big organizations start implementing AI should we start small uh or go all out uh it's a very interesting question if you have seen the journey it's the startups and the smaller organizations.
Avinash Tiwari
Which have taken the lead uh in implementing AI bigger organizations have lot of challenges with respect to data security uh sharing the data uh outside with some of the open LMS uh.
Avinash Tiwari
So so in that sense they have been a bit slow in adopting um the EI the AI use cases but uh the the.
Avinash Tiwari
Right ways to start prototyping uh as I was talking about there are so many open source llms available in the market you can download uh deploy it on uh your local environments and.
Avinash Tiwari
Then start experimenting with smaller use cases once you prototype then you will know what kind of rois are you getting.
Avinash Tiwari
So you have to identify some of the problem statement we talked about the four step strategy you identify some of the problem statements start doing the prototyping see uh the results and.
Avinash Tiwari
Then decide which is the right problem uh to implement at a larger scale so this is a general uh I think Trend that we are seeing a lot of big Enterprises are going through.
Avinash Tiwari
Those small experimentation and this is the right time because we going through that uh trough of uh disillusionment and uh we're.
Avinash Tiwari
Also seeing a lot of new techniques coming up in the market so I think within next 6 months to one years we'll see um much more evolved uh techniques uh but by the time.
Avinash Tiwari
If you have kind of understood the the problems or have done some prototyping uh you will know which problems to kind of go deeper uh with the evolved methods which will exist six months.
Avinash Tiwari
Down the line so start small do more prototyping try fast fail fast is what I can suggest uh there's another question uh how is AI used in visual testing and P clouding.
Avinash Tiwari
So this is as I was talking about is about using computer vision um it's it's more about just doing the screen comparison using the computer vision models and uh finding out the differences um.
Avinash Tiwari
Now there are variety of techniques where how minutely can you identify the difference that's where the real game lies and uh whoever has been able to kind of crack that problem um is ahead.
Avinash Tiwari
Of the game and that's where I think we have done quite a bit of work in terms of giving you differences uh as granular as possible.
Avinash Tiwari
So go ahead and try and share your thoughts there okay I think um uh we will conclude it for today uh if you have any questions uh any suggestions feel free to drop in.
Avinash Tiwari
Um to the host or directly to me uh I'll be more than happy to kind of answer those questions there are.
Avinash Tiwari
So many questions which came up just now so I what I'll do is I'll probably uh take up those questions and answer it uh and send it oneone um.
Avinash Tiwari
So feel free to just drop in more questions and we will come back to you with the answers uh post this session thank you.
Avinash Tiwari
So much for listening and I hope you've got some food for thought in terms of how to implement AI augmented testing within your Enterprise uh within your testing life cycle um happy implementing AI.
Avinash Tiwari
With that thank you so much and have a good day and good night.
