Dinakar
Hello everyone and good evening.
Dinakar
Uh we will thank you for joining the webinar.
Dinakar
We will be uh starting the webinar in a few minutes.
Dinakar
U we'd like to see more participants joining in and we want to be we want them to be there from the beginning.
Dinakar
So just give us a couple more minutes and then we'll get right with you.
Dinakar
All right, great.
Dinakar
Uh, thank you for joining in.
Dinakar
Um, very excited to see so many of you are joining in for the webinar.
Dinakar
U, so hello everyone and welcome to Pcloudy's webinar on automate tests before your build.
Dinakar
Be test ready and achieve true shift left automation with AI agents.
Dinakar
My name is Danaka and I'll be the host for today.
Dinakar
And I would like to go over a few housekeeping rules.
Dinakar
You may have noticed that your line is currently muted.
Dinakar
And if you have any questions, you can submit them during the webinar using the Q&A option that is available.
Dinakar
Uh we will answer your questions as part of the Q&A section at the end of the webinar.
Dinakar
But please do keep your questions coming.
Dinakar
We would love to see what you have to say.
Dinakar
Uh also note that this webinar is being recorded and we'll send the link of the recording to share with your colleagues or watch it again later.
Dinakar
With that said, let me introduce the speakers for today.
Dinakar
Uh we have Avanash Tari and Jessica Selvin.
Dinakar
Avin is a renowned thought leader and a recognized keynote speaker and the co-founder of Pcloudy.
Dinakar
Uh he has around 20 plus years of experience in product development and testing.
Dinakar
uh he brings with him a passion for emerging technology and quick adoption uh both of which which has solidified his reputation as a leader in the AI space as well as the app testing space.
Dinakar
Uh we also have Jessica who will be demonstrating the product for us today.
Dinakar
Uh Jessica is a pre sales specialist at Pcloudy.
Dinakar
uh she has around four plus years of experience in the IT industry and prior to her role here at Pcloudy she has worked with LTI Mry and is as a automation solution designer.
Dinakar
She has a keen interest in data analytics and various automation tools.
Dinakar
So without further ado, let me hand it to Avin to kick off this exciting webinar.
Dinakar
Over to you Avinach.
Avinash
Um hello everyone again uh from my side and thanks Dakar for the introduction.
Avinash
Um really excited to kickstart this webinar and I can see um that so many of you have joined from different time zones.
Avinash
Uh some of you are from Singapore must be very late for all of you touching midnight.
Avinash
So really appreciate you finding time for this webinar.
Avinash
Um everyone good night I good afternoon good morning and uh good evening depending on your time zones.
Avinash
Um again welcome warm welcome to all of you.
Avinash
Grateful for you to spend uh your time with us for this webinar.
Avinash
Um as Takar said u our topic for this session is going to be how do we achieve shift lift testing um using the AI agents uh in in current world that we are in.
Avinash
So today we'll focus more on show than uh talking more on theory part.
Avinash
So we'll not bore you much on the uh intro and I mean some of the general stuff.
Avinash
We'll jump right into the showcase part.
Avinash
That's the whole objective of this webinar.
Avinash
But I will take few minutes to kind of set the context and uh we'll jump right in.
Avinash
Uh so in my journey over two decades into this space of product development testing I have seen huge transformations uh many times u starting from early manual testing days to early days of test automation with tools like windrunner early days of selenium appm uh but one of the pivotal moment was devops and the continuous testing practices that came in that fundamentally changed the way teams had to operate.
Avinash
Silos got broken.
Avinash
The role of quality engineers really evolved and then test automation became one of the critical pieces as far as the quality is concerned.
Avinash
And in last couple of years with the rise of AI we are again seeing a fundamental shift in the way uh we all need to operate.
Avinash
Uh forget about testing uh in general anything that we do right now within application development quality engineering deployment every aspect of our work is getting fundamentally changed.
Avinash
Uh so focus today from our side would be to kind of uh talk about how this new era and what we have done uh using some of the newer emerging technologies is helping our customers achieve u what we call it true shift left and uh I know that there are many skeptics um and uh we have been hearing this from many people that AI I is not capable to handle the complex enterprise scenarios.
Avinash
But in last 12 months of journey that I have taken building AI agents and uh what seemed like science fiction 12 months back is already being rolled into production.
Avinash
Um so my sense and my u uh what I can foresee is AI agents are going to be here.
Avinash
uh that's the new normal and considering that considering the way technology and the AI world is changing at breakneck speed um things are going to change further right so I'm not just saying this you can hear it from the OG himself um 12 months back I was also a bit skeptic when we started building this but over a period of time what we have seen it's it's just so exciting and so amazing to see the results that we we have been able to show to our customers.
Avinash
Uh so that being said I think uh like I said we'll not get into too much of theory and we'll focus more on show than tell but I want to take a moment to again emphasize the role of shift left.
Avinash
It has been a pipe dream from many organization.
Avinash
This concept is very simple uh has been there uh in the industry for quite some time but it's been very very difficult to achieve for the organizations and in a very simple terms you see the two graphs traditional quality methods where most of the testing happens very late in the cycle once your build is ready um and then testing happens and then you go back and and fix fix your bugs and then come back and test again.
Avinash
So that's the fe that's the loop we have been in.
Avinash
So most of the testing happens late in the cycle and the organizations have been trying uh and there have been tools which has been evolving methodologies like TDD BDD came into picture from the quality side of perspective to shift left the testing itself.
Avinash
The idea is that can we do testing as early as possible so that the bugs that we see late can be found out earlier because we all have been seen uh we all have seen what's the impact of a bug which are found very late in the cycle.
Avinash
It can have uh huge impact in the timelines the the and the cost of the project itself.
Avinash
So there has been many tools that came into the world and within the AI agent space as well there are many tools which are doing test planning requirement analysis but our focus has been one of the most complex problems which is how do we bring in sprint automation that's been like a holy grail of automation that everyone has been trying to achieve uh that while my developers are building the application can I make my automation uh scripts or can I keep my automation scripts ready so that as soon as the feature is there I should be able to run it and I should be able to uh perform the testing in a fully automated way.
Avinash
That's been the I mean that's the problem statement we are trying to uh address and uh I mean in automation methodology as far as the unit testing is concerned the API testing is concerned it's been easier to uh shift some of those testing uh to left but the the UI side of it uh the end toend integration side of it the typical user journey testing is always very late right so how do we shift that into the I mean as far as left as far as left is concerned is where we are trying to solve the problem right so we will talk more about how we are doing it with the AI agent that we have developed and like I said we'll focus more on the uh showcasing it so Jessica who has been our champion product specialist for Qpilot she lives and breathes Qpilot has been uh in multiple P implementations across different customers.
Avinash
So she'll be taking it forward.
Avinash
The approach that we have taken for the demo is first we will talk about how Qpilot works and how it automates typical I mean uh automation how it's used in a typical automation cycle.
Avinash
Your application is there and you automate it, generate scripts, run it.
Avinash
But then we will switch to how do we use it to move to shift left.
Avinash
How do you create scripts when your application is not there?
Avinash
That's the second part of the demo that you will see.
Avinash
U so with that set uh I mean with that context setting Jessica over to you.
Jessica
I will jump in uh in the second phase of the demonstration where we'll talk more about how we are using Qpilot in shiplift.
Jessica
Thank you so much uh Ain and hello everyone again um and good evening and good morning depending on your time zones.
Jessica
So let me just quickly go ahead and share my screen and let's get started.
Jessica
Yeah.
Jessica
So with that context right I just want to introduce Qpilot and uh the kind of u transformation that it would bring to your testing in terms of your end to end autonomous testing right so that is the first thing I want to emphasize with Qpilot is that it's an end toend uh AI agent for you right right from your building your test scripts and now um we'll even be seeing even before your build is ready uh right up until executing your scripts right how Qpilot will be that um uh you know end to-end uh agent for you all right and secondly it is a cross-platform agent right uh so you will be able to test for your web applications mobile applications and soon to be API testing as well and also this opens up a whole range of specialized functionalities be it uh testing for flutter applications which has been a challenge or for multi- language uh use cases right so pilot has enabled us to open doors for several uh uh you know specialized features.
Jessica
So that is a bit on Qpilot.
Jessica
Uh let me now go into just a quick uh flow about how uh Q-pilot works.
Jessica
So firstly there's a from your end as a user you just give a natural language input right uh there is no need for specialized coding knowledge in this case uh right and you will just be giving your natural language description of your test scenarios post which uh the AI agents take over uh it will interpret and reason uh right uh identify what are those elements and objects that you're referring to and it will generate a workflow uh and the workflow generation and execution is where the feedback loop happens where there will be multiple times where the generation happens it will execute see if it's working and if not give a feedback and fix it till it gets it right so that is the key differentiator of Pcloudy cube pilot where you will have the accuracy of the script is very high right you will have a high a ready to run script that you don't need to waste those hours debugging right so that is the flow of cq pilot and uh the pet behind behind it is that you know uh the AI agents all communicating with each other.
Jessica
First we have the expert locator agent uh which will be uh identifying going through the entire DOM structure uh getting those fetching those object properties uh right those experts and locators generate the code uh and execute the step.
Jessica
So that's where the feedback will be happening and in case if there's any error also present the error analysis.
Jessica
So all these agents are uh working seamlessly in the back end and um again I want to point out uh this is this flow and this technology is what enables it to be really accurate with the feedback happening.
Jessica
So with that let's just see first how Qpilot works.
Jessica
So firstly I'll be taking you through a generic flow of Qpilot and then we will be moving into uh the specialized features that Ainach was mentioning.
Jessica
So this is the landing page of our platform of Pcloudy platform.
Jessica
It's a SASbased application and you will be able to access both our devices and browsers.
Jessica
So we host um we have 5,000 plus device browser combinations or real devices that are hosted in our data center.
Jessica
uh right uh and these are all real devices in the sense that they're not simulators or emulators and we have multiple data center locations um across the US Europe uh sorry UAE and uh India and Singapore right so these are all the real devices that you're seeing here uh let me just quickly go ahead and connect to a device so once I click on connect I will have complete access of the device uh with me so this will enable you to do your um standard functional testing where you'll be able to install your application and do your functional validations where we also support you know several specialized features like enabling your logs or testing for biometrics uh QR code scanning and so on.
Jessica
So these are all uh you know those features.
Jessica
If you're interested in uh you know our normal functional features you can uh reach out to us.
Jessica
But as this is focused on Qpilot I will straight up go to the Qpilot mode.
Jessica
Let me click on the Qpilot mode over here.
Jessica
And now we are in the Qpilot mode.
Jessica
So the only prompt or the only thing you need to do from your end here is to give your test description in natural language.
Jessica
Right.
Jessica
So I'm selecting my build uh right and I'm giving the app package and the app activity name uh a feature name that I'm testing for and I'm going to give uh my test description in plain and natural language.
Jessica
So let's just go ahead and do that.
Jessica
So I'm just giving a natural language description over here.
Jessica
So I don't want to input static data for my password.
Jessica
In that case, you can just add your test data in the JSON format here, which I'm going to go ahead and do.
Jessica
And I'm adding the test data.
Jessica
I'm going to save my test case under a particular test suite and give a test case name.
Jessica
And now I will simply generate my test case.
Jessica
So firstly what's going to happen is that based on the application details it is going to install the application into the device uh post which the AI agents will kick in and uh start doing its magic right so you will uh it will be able to launch um the application and generate each step.
Jessica
So yes, the application has been launched and now um the AI is going to start interpreting the uh you know the step and then get into and identifying what are those objects it's referring to do the particular action and execute it step by step.
Jessica
So this does take its time.
Jessica
So uh we have prioritized accuracy over speed.
Jessica
So um instead of the time that you would uh you know spend debugging a script or having to re regenerate multiple times right we have prioritized accuracy here and so it will take its time maybe a minute or so for a script to get generated.
Jessica
Uh but post which you will have a ready to run script.
Jessica
You do not need to worry about debugging it.
Jessica
Um it will simply be ready to run and you can trigger it across uh you know different devices.
Jessica
So yeah uh for the for example here the first step is to uh enter the username in the username field.
Jessica
So that is completed and progressively it will keep adding and executing each and every step.
Jessica
So this is what brings about that accuracy.
Jessica
So I was talking about the the feedback that happens between uh you know the workflow generation and the execution.
Jessica
So that's what's happening here where for every step there is a generation and an execution and that feedback happens to ensure step accuracy.
Jessica
So uh the login has happened now and it will click on transfer and then the script will be ready.
Jessica
So yeah, it has clicked on transfer and now we're ready to generate go to the cube pilot dashboard to view our script.
Jessica
Of course I've given a fairly small test case but of course you will there's no limit to the number of test steps you can give here.
Jessica
You can add validation steps and assertion steps as well.
Jessica
Yeah.
Jessica
So here is the description and here is our ready to run script.
Jessica
It's available in multiple languages.
Jessica
So, JavaScript, Python, Java, right?
Jessica
And uh even we are soon bringing about a template feature where you can even have it in a template of your framework as well.
Jessica
So, you can run this script directly uh individually, right?
Jessica
But if you want to, how we typically do testing would be is to add it into our collection and uh run it as a suite.
Jessica
Right?
Jessica
So I'm going to go to the create collection um option here and I'm just going to add a few test cases.
Jessica
So here is where you can achieve true crossplatform testing.
Jessica
You can select um test cases across your web your iOS based test cases.
Jessica
I'm going to insert a couple of iOS based test cases and a few from our Android based test cases.
Jessica
And I'm going to give a name to my collection and then select the devices that I would want it to run on.
Jessica
So um here for my Android I can just choose the OEMs right or which specific device you want it to run on.
Jessica
Um and you can give a version range uh the OS version range and the number of devices you would want it to run on.
Jessica
Similarly, you can do the same for iOS.
Jessica
So, I'm just giving a range from 15 to 18 here.
Jessica
Uh the number of devices and clicking on save.
Jessica
So once this is done, I can simply run it uh and trigger all the test cases with a single click.
Jessica
And you can also trigger this from your CI/CD as well.
Jessica
So yeah, I've triggered my collection.
Jessica
Now I can go to the live session to see the live view or I can head straight to the reports as well.
Jessica
So here in the my active session so you will be able to see all the live views that are coming in as well as the reports.
Jessica
So we have a progressive report.
Jessica
So as and when test cases are getting executed, you will be able to see the live reporting of the same.
Jessica
So yeah, as you can see, we have all these automation runs that are being triggered in our live view, right?
Jessica
And if you want to see, you can even go into the live view and you'll be able to see what actions are exactly taking place on your device.
Jessica
else you can just wait simply wait till your execution run is completed and you can go into your um report page simply click on your build or your collection name and you'll be able to see the status of all the test cases how many were completed how many failed right and into each and every test case you will be able to see uh you know the text logs the device logs um your APM logs as well and app profiling details with your CPU memory details as well as uh the video of the entire session.
Jessica
So this will be available for every automation trigger uh that you do.
Jessica
So that is a bit on collections.
Jessica
So I have just taken you through a very basic and a simple flow of how you are Qpilot is enabling you to just trigger your generation right from our end we have simply given a natural language input.
Jessica
we have not reviewed or debugged spent time debugging our script and we have executed our collection and we're able to see the um results of our test suite.
Jessica
So this is how Qpilot really helps you and enables you to achieve that autonomous testing that I was talking to you about.
Jessica
So with that let me just hand it over to Ainach to take you through the specialized use cases which we will now be walking.
Jessica
Yeah.
Avinash
Um, thanks Jessica.
Avinash
It was a great demonstration and I hope everyone kind of understood how Qilot works.
Avinash
uh ju just to set the context what you saw was a typical use of Qpilot in uh helping you create automation scripts in a completely autonomous way um when your application builds are there and that's how any uh UI automation across web mobile works that you get the uh bill ready and then you create your automation scripts you add that your regression so it's sanity so it's etc and can go ahead and run it.
Avinash
Um the collection feature that you saw while Jessica added few devices but it can you you can scale it up to any number of devices depending on the licenses that you have.
Avinash
So you can run it on maybe 50 concurrent instances, 100 concurren concurrent instances.
Avinash
So that's the level of scalability you can achieve.
Avinash
So you create your script autonomously, you execute it on a large infrastructure all that with very few clicks.
Avinash
That's the power of Qpilot.
Avinash
Now, um what we're talking today is about shift left, right?
Avinash
How do we uh shift to uh or move to a state where I can create my scripts when my application build is not there, right?
Avinash
And typically we have seen this approach with TDD BDD where we create our um we keep our uh scenarios ready and um as soon as the build is there you try to generate the script but again you have a limitation there that you need the application build uh where you create the script verify and then add to your collection or suites.
Avinash
We have gone one step further and there are two approaches that we have taken.
Avinash
I'll explain those approaches uh in a bit and then Jessica will demonstrate how those approaches work.
Jessica
So um Jessica so the approach one um is very typical scenario where let's say you have already created a regression suite um there are test cases already automated uh let's say take the example of Qpilot you have already automated some of the functionalities which are there in your current build it could be let's say 100 test cases 200 test cases now there is a new uh build coming in let's say next two weeks for your current sprint and there there are some new features getting added.
Jessica
Now, how do you ensure or how do you achieve that state where the day your build is ready, your automation script should be ready?
Jessica
That's the problem statement.
Jessica
Um, so basically what we are saying is that you have certain scripts ready and you are planning to add new scripts for your new features which will be released in the sprint.
Jessica
Right?
Jessica
That's the uh situation and that's the use case.
Jessica
Uh so basically QA teams are supposed to add new test scripts for upcoming features and run regression along with the new uh feature scripts.
Jessica
That's typical in sprint automation that we want to achieve.
Jessica
Now here what we have done is um since you already have some scripts created we have historical data.
Jessica
AI is very smart to understand the historical data.
Jessica
We're using that historical data of previously generated scripts and then have added something called think and act approach for the new steps that you need to add.
Jessica
Right?
Jessica
Um and that think and act steps are only for the uh features which will be upcoming in the uh in the sprint.
Jessica
rest of the the steps will be taken from the historical uh uh steps that has been already generated.
Jessica
Now what this enables us is to create a pre-generated script even before your application is ready.
Jessica
And what we call is some scripts will be the real script because some of the steps has already been generated.
Jessica
this the steps where we have added a think and act uh tag for those are very real lifelike placeholders are getting added.
Jessica
Again we are adding the placeholders because we know how your for example uh app how your developers are creating the name of your object label etc.
Jessica
So very real life like u placeholders are getting added and there's no shorty that those placeholders will be really working when the build is ready but that's when when the build is available and when you run it as part of collection our feedback loop that Jessica explained earlier that there'll be a step execution and if it fails it modifies it.
Jessica
So in a way itself feels on its own during the execution time and your scripts gets generated and executed as the build as soon as the build is ready.
Jessica
So that's the first approach and that's the most used approach in many customer scenarios that we have tried with.
Jessica
Uh the second approach is is uh it takes it one step further.
Jessica
However, this is less likely scenarios in in most cases where let's say you are developing your application for the first time.
Jessica
you have no historical test case uh or you are creating an automation project for the first time you don't have your application built you don't have your previous test script so there's no historical context that we can use now how do we achieve u the automation in that state this becomes much more resource uh uh intensive for us but in this case you just give your natural language descriptions and we will generate the scripts totally at the runtime during your suite execution.
Jessica
So this is this can be used when you're when you you don't have any historical test case data.
Jessica
You don't have the application built at all to create any historical data.
Jessica
Right?
Jessica
So let's say I'm I'm creating a new product which is going to be launched in 1 month from now.
Jessica
Can I create some of my automation scripts for the product which will be launched in next two weeks or so?
Jessica
That's the state we are talking about here.
Jessica
So these two approaches we will show.
Jessica
However, approach is one.
Jessica
Approach one is where we uh we see the most usage in the real world conditions but we'll show both and we are also kind of talking to many customers experimenting which worksh.
Jessica
Um we love your feedback as well as to what do you think which one would work in your scenarios.
Jessica
Um so let's see it let's see it live working rather than me talking about it.
Avinash
So over to you Jessica again.
Jessica
Yeah.
Jessica
Yeah.
Jessica
Thanks Ainach.
Jessica
So uh let me just take you through the example that we'd be using for both the approaches that Ainach just um explained.
Jessica
So uh we have a build here.
Jessica
It's a sample uh banking app where we have this is the build one where we these are the features that are available and we have a build two where a couple of features have been added that is the add card feature.
Jessica
Right.
Jessica
So um we have the scripts already generated for our build one.
Jessica
Uh but we have to test this new feature that is coming in our build 2.
Jessica
But our build 2 is not yet available with us.
Jessica
But I still want to begin my testing.
Jessica
Uh say you just have a Figma of your of the new features.
Jessica
Uh right.
Jessica
Um so this is the first uh use case that I will be uh demonstrating.
Jessica
So with that let me just go back to the platform.
Jessica
Let me just log in back to the platform and I'm getting into the Qpilot dashboard here and seeing all my test cases.
Jessica
Now, uh just for the purpose of this demonstration, let's assume that these are all the test cases that I've developed for my first build.
Jessica
Uh right and this is the second build that's coming in and I want to create uh a test case testing for the new uh add card feature but I still do not have the build ready with me.
Jessica
All right.
Jessica
So, let me just I will go ahead and click on create new test case.
Jessica
And in this case, it's for a mobile.
Jessica
So, I'm just clicking on mobile Android.
Jessica
Selecting my suite under build 2.
Jessica
And I'm giving a test case name.
Jessica
Right.
Jessica
So, let me just give a name to this.
Jessica
Yeah.
Jessica
Application selection.
Jessica
And then I'm going to give the app package and the app activity name, the feature that I'm testing for and then the description.
Jessica
So let me just take you through the description.
Jessica
I think Avin had already elaborated.
Jessica
So for example here the login functionality is already there in my build one uh right so that will be as uh you have done in your previous testing scenarios.
Jessica
Um when I have add this hashtag act right below my u before or prior to my description um this notifies the AI agent that this is a step that would be generated at the runtime.
Jessica
All right.
Jessica
So right now in my build I do not have these features but I still want to proactively create my script before the build comes and I just have say the Figma office.
Jessica
All right.
Jessica
So in that case for all those steps that involve the new feature I will go ahead and add the hashtag act and then continue to give my test description exactly like how I would for a normal Q-pilot test case.
Jessica
All right.
Jessica
So, and let me also add the test data for this for the username and password.
Jessica
Yeah.
Jessica
And I'm adding the test data and I am saving my test data.
Jessica
Right.
Jessica
And now once uh you know now I will go ahead go to the feature that script that I've created.
Jessica
Right?
Jessica
And as you can see here is my description and I will simply click on create script.
Jessica
Now when I click on create script what will happen is uh I have the actual script generated for my first three steps for the where we already have the build and for the rest I have placeholder scripts that is based on the prior intelligence that is already there because we have developed scripts for build one uh right as a was explaining for approach two.
Jessica
So I have a placeholder script uh for the rest of the steps.
Jessica
So for example, you can see here the AI is saying placeholder action for runtime interpretation.
Jessica
So this uh we are not sure whether this uh you know script will work at the runtime.
Jessica
So there will be that feedback loop that will happen at the runtime to ensure that the uh script is accurate and your test cases are getting executed.
Jessica
So once this is done you can simply go to uh create collection add your new feature into the collection right and you can give a name to the collection I'm just adding a build name in this case build two and selecting the devices that I wanted to run on.
Jessica
You can also combine this with all your uh previous existing test cases for build one as well that would be applicable for build two and go ahead and save your test case.
Jessica
And now I can simply go ahead and run this collection and I will be able to see the report for the same.
Jessica
So let's just go to the live session.
Jessica
So in the live session what will happen is runtime generation will be happening and then post which the execution will happen.
Jessica
So this ensures so you do not need to spend that time generating your script at only after the build is ready right so you have just created your scripts and now you're simply as and when the build becomes ready you can just trigger your test case or your collection and it will generate at runtime and then execute it.
Jessica
Let's just go to the reports.
Jessica
Uh just one point uh that this yes this collection you will execute it only when the build two is available right so it can be done using CI/CD so as soon as your build two is uh pushed to your git or any other uh platform that you're using this can be triggered and your automation scripts can be uh created and executed and you can just focus on going and checking the reports right so uh once you can see the live view or you can just directly go ahead to your report section and once the execution is done simply view the reports so here is the build two it's currently in progress so I can also see um for example another sample report for the same right so here you'll be able to see just like how I showed you the text log device logs uh your APM logs and the video of the entire test So this will be uh available again for each and every uh test run that you do.
Jessica
Uh so that I hope uh the approach one has been clear.
Jessica
So we have we've had a new feature um and we've tested it.
Jessica
We've created the test cases without the build and at the time when the build becomes ready I'm just simply triggering uh my collection and getting all these reports.
Jessica
Right.
Jessica
Uh next I will move to uh the our second use case.
Jessica
So this is when no version of your application is available.
Jessica
All right.
Jessica
So sorry.
Jessica
Yeah.
Jessica
One more point that I wanted to add add here was when the build to was available and we reran the script the placeholder will get replaced with the actual script.
Jessica
So we go go back and see the uh script again in your cuppilot dashboard.
Jessica
you will see the placeholders getting removed and the actual script actual DOM element the the actual action of the the object will be displayed here um and the even the description act can be uh will be removed so that's the I mean so that the the feature which was missing becomes the feature which is available and you can now focus on the other features which are upcoming so those will be part of the act and think rest is part of the normal Q-pilot mode.
Jessica
So it's it's that intelligence is already built.
Avinash
Go ahead Jessica.
Jessica
Thanks.
Jessica
Yes.
Jessica
So the second use case is uh when no version of the application is available and uh again as a mentioned this would be resource in in uh you know intensive because you do not have any history uh or a previous builds uh right for that intelligence and you're creating so you can create your entire test descriptions uh right uh without the build uh so here we have used our own p-cloudy application as an example and assuming that this is a new build and we're going to be testing for uh a particular feature called my bookings.
Jessica
All right.
Jessica
So that's um what we're going to do here.
Jessica
So again I will go to my Qpilot dashboard, go to all tests and say um you know I have a new feature this feature that I want to test.
Jessica
All right.
Jessica
And I'm simply going to go ahead and create new test case and uh in this case it would be browser.
Jessica
Right.
Jessica
I would select uh which suite that I would want it to be a part of.
Jessica
Right.
Jessica
And I'm going to give a feature name and so on.
Jessica
So I'm going to give the URL of the application here and then a description.
Jessica
So in this case the description need not be marked with a hash act because it's going to be the entire test case is going to be generated at the runtime.
Jessica
Uh right.
Jessica
So I'm going to save my test case and I'm simply going to go to my collection.
Jessica
Uh and I'm going to add the booking feature collection.
Jessica
I can also add uh you know multiple test cases here.
Jessica
I'm just going to add a few more.
Jessica
Uh yeah give a name to my collection.
Jessica
collection details needed.
Jessica
You can add that and simply you know give a name to your build.
Jessica
Right?
Jessica
So you can see that the my booking feature one where uh you know where we have um just create given the description the toggle for runtime generation is enabled here.
Jessica
So what this means is firstly it will uh you know go ahead and do an execution a pilot generation like how I showed you uh in the beginning right it will do that generation and get the script post which it will then trigger the run for all these three test cases.
Jessica
So whenever you select uh you know a test case at the runtime generation enabled what it'll first execute the entire script and generate the script and then it will be able to uh you know execute all your test cases as a collection.
Jessica
This is in the case where you know um we are not uh creating the script like or adding a particular feature functionality testing but it's for the entire description uh of your use case being uh you know uh generated at the runtime.
Jessica
So I'm just going to again select uh the browser that I wanted to run on and go ahead and save my collection and I'm going to go ahead and run my collection.
Jessica
So the collection executed has started.
Jessica
I can simply go to the live view session here and you know see the generation taking place.
Jessica
Uh else I can again simply go to my reports and go ahead and view that.
Jessica
So here is a sample report.
Jessica
Uh this is the report where we have generated just uh executed a while back.
Jessica
So this is the new uh my booking feature along with all the other functionalities as well.
Jessica
So for example, if I go here again, I will have the logs of my entire test case uh with the video of the entire uh session that's happening.
Jessica
So you can see that the execution the script has gotten generated and executed from scratch.
Jessica
Uh right so I have not spent that time generating my script waiting for my build.
Jessica
Uh right and it's going and clicking on my bookings and the execution is completed.
Jessica
So what has happened here is firstly the script has gotten generated and executed in a single click.
Jessica
So uh that is the explanation of our second feature where you do not have a version of your application or you do not have history uh or any historical data from previous builds.
Jessica
Uh you would still be able to do that shift left and bring about your automation um at the time of development.
Jessica
So with that uh I hope those two use cases were uh gave you some insights into the kind of breadth of uh you know uh test cases and use cases that are possible with Qpilot uh and the kind of u you know testing and shift left that you would be able to do here.
Jessica
So with that I'll again hand it over to Aash uh to conclude on this.
Avinash
Thanks Jessica.
Avinash
Thanks a lot for explaining this so well and showing the the real time execution and creation of scripts uh for both the cases.
Avinash
Uh on the second case I want to add one more uh scenario where you can use it.
Avinash
So if if your application is changing very frequently right I mean in in especially when your application is in the initial stages of development there are many changes which are happening.
Avinash
So if you create the script and keep it and use that u it fails many time because the UI has changed elements have changed and then you need to recreed or self use self-filling which are again not useful when there are so many changes right self-healing can work if there are some locator based changes and some elements have changed but if there are many changes in your application it fails and that is where this whole runtime generation where we are not dependent on any previously generated script.
Avinash
We take the context of the application at the runtime.
Avinash
Generate the script and then go ahead and execute it.
Avinash
Right.
Avinash
U like I said it would it would be little resource intensive.
Avinash
uh the execution will take a bit more I mean a bit of time compared to the normal execution but you are ensuring that your uh scenarios are getting tested even you even when your applications are changing right so that's the power of this whole runtime generation that we have been able to build um so depending on your situation you can choose approach one and approach two if your application is already stable and you are focusing on um testing new features which are upcoming in the next sprint the approach one where think and act works well.
Avinash
But if your application is very dynamic going through a lot of changes then you can opt to use runtime generation till your application is stable and as soon as your application is stable you can always switch back to your uh standard generation and where your pre-generated scripts will be used and your executions will be much faster.
Avinash
So that's the power Qpilot provides with so much flexibility for you to choose across different approaches that you can use across I mean within your organization and depending on your situation.
Avinash
So with that being said I just wanted to uh conclude this webinar.
Avinash
Thank you so much for joining and I hope you have seen what AI agents can do.
Avinash
Like I said, this whole AI agent space is evolving very fast.
Avinash
We are seeing changes at breakneck speed from the LLM perspective.
Avinash
LMS are becoming much more smarter.
Avinash
So all the initial challenges that anyone encounters with AI agents are getting solved as we speak.
Avinash
So I foresee that AI agents are going to be the primary way of I mean testing automating uh the scripts and the the QA and uh the humans can focus on much more complex tasks.
Avinash
That's the uh scenario we are looking at going forward.
Avinash
Um do let us know your feedback.
Avinash
If you wish to try uh do reach out to uh any one of us or our support or contact email ids we will be more than happy to talk to you and address your specific use cases.
Avinash
Uh thank you so much.
Avinash
Um I loved talking about it and hope it it's been helpful.
Avinash
With that we'll open up for Q&A.
Avinash
Uh I can already see many questions in the chat window.
Avinash
We'll try to address u uh as many as we can uh in the limited time u that remains and if we are not able to do that we will uh send a email answering your question.
Avinash
Um so the first question is around u would this be able to recognize regular UI element that are clickable but don't visually appear as buttons.
Avinash
Yes, I mean uh we have been able to support most of the I mean UI elements that exist in generic applications.
Avinash
However, we do find uh new objects, new uh controls, custom controls.
Avinash
For that, we might have to improve the training of the model.
Avinash
Uh but once the training is done, I think uh uh the the model understands and the next time you generate the script, uh it works fine.
Avinash
So there can be a bit of iteration in making sure that your specific controls have to be added to our model but we have tested across most of the generic application.
Avinash
We have implemented in customer scenarios.
Avinash
We have done I mean many customer pilots successfully.
Avinash
So this like I said our focus was to build accuracy and we have been able to achieve that pretty successfully.
Avinash
Uh one use case that any anupam is asking about tapping on video player to pause or resume.
Avinash
Yes, it's definitely possible.
Avinash
Interacting with a map to set a location or drag it around.
Avinash
So drag and drop is a bit of challenge.
Avinash
It works but how much drag do you want to do?
Avinash
There is there's no way to kind of specifically mention.
Avinash
So it depends on your use case.
Avinash
We can try and see.
Avinash
As I said, it can always be improved depending on the specific scenario that we come across.
Avinash
U um another question is around um I'm keen to see how the tool will support dynamically changing web portals.
Avinash
U this is specific use case for our organization.
Avinash
That's exactly what we are trying to address with approach to Giri.
Avinash
We would be more than happy to talk to you and um show you how it works.
Avinash
Um like I said, approach two generates the script at the runtime.
Avinash
It doesn't depend on the previous state of your application or previous state of the script which was generated.
Avinash
So it takes the current state of your application as the context and generates the script at the runtime.
Avinash
So even if your applications have changed as long as your scenario remains same, the the steps that you have mentioned remain the same, the validation that you want to do remains the same, it will generate a new script and that will get executed.
Avinash
Um, so that's that will solve the scenario that that you're talking about.
Avinash
Um, third question from Dave Dut.
Avinash
How does the feature uh platform with apps with flutter and PWS screen?
Avinash
I mean we are one of the first one to support flutter on real devices.
Avinash
There are other platforms who are supporting flutter on simulator emulators etc.
Avinash
We are the first one to support uh Flutter uh apps on real devices.
Avinash
Um so we welcome you to come and try and see and give your feedback.
Avinash
In fact we just released one video of uh Flutter apps with Qpilot.
Avinash
So we can share that video as well.
Avinash
Uh yes we are the question is around uh recording of the session.
Avinash
Yes, we are recording the session and we'll be sharing that uh with you uh once this webinar is over.
Avinash
I think you should uh get that recording maybe tomorrow um as part of your email.
Avinash
So please do check your email and uh access the recording link.
Avinash
If if in case you're not uh getting the link um do reach out to us and we'll be more than happy to share it individually.
Avinash
Uh question from Sashant Josie.
Avinash
Will it be able to handle pigeonation and verify value of table on UI to XLS?
Avinash
Yes, I mean we we have tried complex scenarios like validation in tables etc.
Avinash
So it's definitely possible.
Avinash
Um we can try on your case and and and show it to you.
Avinash
Uh can we provide test data via Excel sheet or DB access?
Avinash
I mean it's possible.
Avinash
Right now we have not enabled uh that feature but we are passing it on JSON uh data so we can just link your DB and Excel and convert that data to JSON and feed to our Qpilot.
Avinash
So theoretically uh is it's possible right now that feature is under development it will soon be available within the product.
Avinash
Um Abilash how secure these agents are absolutely I mean these are running in our uh uh secure environment.
Avinash
Um we have an option of uh hosting separate agent in a private environment as well.
Avinash
So if you're working uh in a regulatory uh environment within for example let's say BFSI or pharma kind of industry we can create a private uh instance for you which is absolutely dedicated secured air gapped for your organization.
Avinash
Um another question how will we how it will be handling the dynamic elements in a web page that's the that's the whole crux of how Qpilot works.
Avinash
we have been able to crack this algorithm of uh handling the dynamic object through the feedback mechanism that we have created.
Avinash
So there are multiple agents working together where we find uh depending on your natural step we find which element it is we try to find out the locator we we execute it at the runtime and then if it fails we try another locate.
Avinash
All that is happening uh behind the scene agents are performing those tasks passing feedback to each other and finally when the step is executed you see the result you see the script so I mean that's the power of uh AI that that gives us these kind of functionalities uh will it be able to perform API and load testing as well?
Avinash
Yes, I I mean we have not enabled load testing right now, we have enabled inclusion of API tests.
Avinash
So like you saw in the uh Q-pilot dashboard or when you're creating the test, you can add API step in your functional tests as well.
Avinash
Uh we we we have already kind of developed a independent API test suites.
Avinash
Uh but right now it's you can integrate your API test within your functional scripts.
Avinash
uh and then run it along with your functional mobile application web application.
Avinash
So what the objective that we have taken is that let's enable true end to end uh uh scenario generation for your customer for for our customers.
Avinash
So you can within a collection you can start with your web you can switch to API you can move to your mobile within mobile you can choose Android iOS in one collection running uh sequentially or parallelly depending on how your test cases have to be executed.
Avinash
So truly integrated end to end testing can be achieved.
Avinash
Uh one question is on from where we should get the package name and other fields detail which Jessica showed in the first step.
Avinash
So I mean every application especially in the mobile app has a package name either your developers can provide that or we have a small utility which you can run and get these details.
Avinash
Uh so as soon as you upload the APK or IPA on our platform, we call it my app data which is a repository where you store all your applications uh it automatically uh scans your application and gives that detail to you which you can use it while you are uh creating the test cases right and we we also have a feature which we call it collection level configuration.
Avinash
So even if you have given certain uh build name package name at your test level but if that has changed you don't need to go and change at the test level while you're running the collection you can run the collection on your different builds you can run it on a staging you can run it on your uh QA you can run it on your prod.
Avinash
So just change the configuration at the collection level and then run it across different environments.
Avinash
So that's uh that's also quite easy to do with your pallet.
Avinash
Um uh another question is how will it handle where multiple platforms are involved?
Avinash
One product generates some data.
Avinash
We have intermediate layer which we use to process this data based on some condition then transfer it to platform like Salesforce.
Avinash
I mean right now um uh that is something we need to kind of uh enable you we I mean what you are effectively asking for is passing the data from one test case to other which is very easy to to use because I mean it's just about creating parameters and then passing though that work is still going on uh we have that in one of our other tool which was record and playback portless engine we have to bring that in the Qpilot mode as well.
Avinash
So once that is done then you can achieve this uh state.
Avinash
Right now we are assuming that each uh scripts in itself are independent and then you can run it in a sequence or in parallel depending on the uh choice that depending on the case that you want to execute.
Avinash
Uh another question is can it understand the PR changer and take corre corrective action.
Avinash
I mean yes that that is definitely possible.
Avinash
U that's been proven by u I mean that's one of the first use cases AI agent started solving.
Avinash
Um so that's definitely possible um with our Qpilot as well.
Avinash
U but right now we have given the control to the QA team to either update it or keep it the same u but we can link it with your git repository and let it update fully automatically.
Avinash
Um more philosophical question what will be the future of QA role in many product companies?
Avinash
uh will developers do testing and QA role has come to extinct I mean that's the question everyone is grappling with but I don't think so that's the case uh what uh we definitely see is that everyone has to start uh learning AI and uh start integrating uh that in our daily work um so the famous saying is that AI will not replace you um if you don't learn AI then you'll get replaced.
Avinash
That's the much more practical or or foreseeable future statement that I can make.
Avinash
Uh QA will be doing much more complex task in terms of analyzing, creating much detailed scenarios.
Avinash
Right now, majority of the time of QA goes into creation of scripts, fixing the script and then we tend to miss some of the corner scenarios.
Avinash
A lot of exploratory testing doesn't happen.
Avinash
So QA has to focus on really understanding the user journeys creating complex scenarios while the AI agents can take up these repetitive task of generating the script fixing when the when there is a failure.
Avinash
So in that sense AI will become an assistant to you.
Avinash
I don't think there is case of that QA will become extinct.
Avinash
In fact, in this world, QA becoming much more uh important because the the change of the the pace of change has increased.
Avinash
Everyone wants to release faster.
Avinash
You can't do that in a completely autonomous way, right?
Avinash
I mean, you need people to monitor, you need people to intervene.
Avinash
Uh we have always believed that there will be there will always be a human in the loop.
Avinash
Uh but that human has to work with AI.
Avinash
I mean if if you if you think that I can do it the traditional way I think that is definitely going to change especially for example we we we come up with lot of queries in the automation space around record and playback codeless tools I mean I don't see a future for codeless uh tools at all going forward so that is definitely going to happen in next year or so most of the tools will have to switch to this whole EI agent methodology and we have been at the forefront of uh bringing um the the uh AI agent into practice from almost year now.
Avinash
We have reached a very high level of accuracy uh as far as the effectiveness is concerned uh fidelity is concerned.
Avinash
Uh does it require any coding knowledge?
Avinash
Can manual testers work on it without coding knowledge?
Avinash
That's another question.
Avinash
Absolutely.
Avinash
uh we are generating the script and showing it but that's uh uh that is not required to be changed I mean we have given the script uh right now let's say if if you have already built your framework outside Qpilot let's say you have appium framework already built and you don't want to tie yourself with something like Qpilot you can just generate the script copy it in your framework and and use it that's the only reason we are showing the script otherwise that script doesn't require any change that that that doesn't require any addition.
Avinash
Um so anyone who just can write the natural language uh test descriptions uh add the test data run it to add into collection should be able to run.
Avinash
So it's typically designed for uh anyone who is part of the testing uh business users product managers anyone can just write the scenarios and then get the automation scripts ready.
Avinash
Right.
Avinash
Um, so hope that answers your question.
Avinash
Okay, great.
Avinash
I think uh uh there are a few more unanswered question but in the interest of time we have already crossed almost an hour.
Avinash
Um, thank you so much again and uh thanks for uh asking so many questions being so interactive.
Avinash
Would love to um uh hear your feedbacks.
Avinash
Any more questions feel free to drop in.
Avinash
Um you can you can connect with us on LinkedIn.
Avinash
Um reach out to our support.
Avinash
We'd love to interact and uh hear from you.
Avinash
And uh if you like what you saw, go ahead and share.
Avinash
We'll really appreciate that.
Avinash
Thank you.
Avinash
Thank you so much.
Avinash
Uh have a good night or great day depending on where you are.
Avinash
Um I think we will conclude for the day.
Dinakar
Well, thank you uh thank you Avin and Jessica for sharing your wisdom in the field of you know AI powered digital experience testing.
Dinakar
You've done a great job at doing a deep dive into adopting AI agents and showing us how to automate tests even before the build.
Dinakar
So thank you so much.
Dinakar
Uh we also want to thank each one of our participants who are here and you know you've been a great bunch of participants who have been very engaging and I'm sure that some of these insights that were shared by Abinina and Jessica have sparked a few ideas to think through and implement.
Dinakar
My recommendation would be to you know mull over these ideas talk to your teams reach out to us to discuss further and keep the conversation going.
Dinakar
Uh so if you do and do feel free to reach out to us uh with any questions you can write to us at infopludy.com or just click on the chat option on our website and we'll directly reach out to you.
Dinakar
You can also reach out to Abinach and Jessica uh you know using their LinkedIn handles.
Dinakar
I'm sure they'll be more than happy to answer your questions or you know route it to the right person uh within the organization.
Dinakar
So, thank you once again for being a great engaging audience and happy testing.
