How To Analyse If you Can Use AI/ML?

Hi !!

I've been in the testing industry for nearly 20 years now.

Let me tell you this.

The things have changed. Changes a lot.

When I started, the word ‘Artificial Intelligence’ sounded like science fiction.

Now?

It is a reality. It is everywhere, including in our testing process.

So let's discuss about how you can actually use ML/AI in your day-to-day work.

No fancy jargon, no buzzwords - just practical advice from someone who's been in the industry for a while.

That is Bharath Hemachandran.

The following is the notes from his master class.

First things first: Do you even need AI/ML?

That is great question.

It is very challenging to decide that.

But, how to do?

Here is a framework suggested by Bharath Hemachandran.

Framework To Decide If You Can Use ML/AI For The Use Case

Here's what you can do:

1. Breakdown the use case in to smaller component process.

For example, in the test process, Test data generation can be a sub process/components.

2. In those sub components, identify the specific tasks.

Example is test data generation or maintenance is a crucial task.

3. Analyse these tasks. And see if your problem requires:

- Finding patterns (like spotting weird behavior in test results)

- Predicting stuff (like estimating how long tests will take to run)

- Understanding text (like analyzing bug reports)

- Working with images or speech (like UI testing)

- Making decisions based on tons of data (like choosing which tests to run)

4. Will this actually help the project? Or are you just playing around with AI?

And once you have broken the process in to smaller component processes and analysed, you can ask these questions to yourself.

How To Decide If You Can Use AI/ML In Your Use Case?

These questions are very crucial.

Ask them to yourself.

They should be asked and answered in sequence.

That is if you can answer to question 1, then only you can go to 2nd.

I hope you got it.

Please note that, answer to any of these questions suggests you if use case is the good candidate for ML implementation.

1. Can I do this manually or using existing tools easily ?

Sometimes, a simple script beats a complex AI model.

And some times, a simple Excel formula can be the solution.

2. What's the real benefit here?

Don't use AI just to say you're using AI.

Because AI implementation is costly affair. There needs to be lot of investment-time,money and resources.

3. Do we have the required data?

AI needs data. Huge data for training the models.

AI needs lot of data. In terms of TB or PB. That is huge.

Collecting such valid data is challenging and costly.

4. Can our infrastructure handle this?

You can not train model with huge data on your laptop.

Training and running ML model needs infrastructure like GPU, etc.

So decide accordingly.

5. Usage of AI needs permission.

You can use AI if it is ethical. Of if it is allowed as per your company policies.

For example you can not upload your company confidential data to public LLMs.

It violates the policies. You will be penalised for that act.

Are the higher-ups okay with this?

Trust me, you don't want to find out they're not after you've done all the work.

6. Are there skilled personnel to use AI in your team?

Do we have people who can actually make this happen?

AI isn't magic - you need skills to make it work.

Hiring and paying the skilled personnel is challenging.

7. Is the outcome testable ?

How will we know if it's working?

If you can't test it, don't build it.

If you're not sure about any of these, just apply the brakes.

Stop !

It's okay to say "not yet" to AI.

Example Of Determining If Use Case Is Best For AI Implementation

Here I will give you an example.

The use case is : Determine the bug concentration about a feature of product.

By identifying the bug concentration, you can determine which feature is more buggy based on the bugs reported.

I used the above framework.

These determines if I can use the Artificial Intelligence in this use case.

This I have documented in the form of mind map.

This is self-explanatory.

If if need explanation on this mind map, please let me know in comments.

I will explain more on this.

What You Can Do Next?

Now comes the fun part.

Here's how you actually start using AI/ML in your testing:

1. Know your project inside out.

You can't improve what you don't understand.

2. Look for the boring stuff.

What tasks make you want to poke your eyes out?

Those are prime candidates for automation with AI.

3. Start with something small and non-critical.

Your first AI project shouldn't be mission-critical. Learn on the small stuff.

Break the process in to smaller units.

Analyse if you can do anything about it.

4. Make friends with the nerds.

Yes !

Seriously. Networking is very important. Important to learn and implement.

You're going to need people’s help.

5. Keep learning.

This field moves fast. Set aside time each week to read up on new developments.

Have consistent learning.

Be updated with the latest developments in the field of AI.

6. Don't be evil.

Always think about the ethical implications of what you're doing.

AI can amplify biases if you're not careful.

And violation of any policy / law is punishable offence.

7. Share what you learn.

Now, this is super important.

According to me, teaching is best form of learning.

Share what you have learned. Share within your team. Share with your friends. Share in social media platforms like Linkedin,YouTube,X, etc.

Trust me, people will appreciate it.

You will be motivated to learn more. To implement more.

What To Expect With AI?

Please make a note.

And I can't stress this enough - AI is not here to replace you.

It's here to make your job more interesting. AI can boost your productivity. It can boost your growth.

Sure, AI can crunch numbers faster than any of us.

But it can't understand the context of a bug like you can. It can't have that "gut feeling" that something is off.

That's where you come in.

Your job is evolving.

You are not just finding bugs anymore. You're teaching these AI models what "good" looks like. You are the bridge between the tech and the real world.

As you grow in your testing career, you will develop a sense for when AI can help and when good old-fashioned testing is the way to go.

Never stop learning.

The moment you think you know it all is the moment you start falling behind. Make the learning as continuous process.And consistent process.

So, embrace this new world of AI and ML.

Let it make you a better tester. But never forget the value of your human insight. That is something no AI can replace.

Remember, the only constant is change.

That's all for now !

-Jayateerth