The Revenue-Focused Marketer

More From Our Content Repertoire:

• Don’t Let AI Break Your Brand: What Every CMO Should Know
• How To Become An AI-Ready Organization
• Breaking Down Barriers: AI Tools and the Democratization of Data Analysis in B2B Marketing

Transcript

0:11:
Hello and welcome to another episode of the revenue focused marketer where we discuss anything and everything related to marketing as well as data.

0:21:
I’m your usual host, Hershey, and today we’re diving into a really meaningful conversation at the intersection of performance, inclusion, and innovation.

0:31:
Joining me is Abhishek.

0:33:
Abhishek, welcome to the show.

0:36:
Hi, Hushi.

0:36:
Thank you so much for the intro.

0:38:
, it’s been a pleasure, and I’m looking forward to talking with you.

0:42:
Yeah, just a little bit more about Abhishek.

0:45:
He’s a software engineer at Microsoft, who’s impacted a lot of people through his work on performance and accessibility features, specifically in Word Web.

0:56:
He’s been a driving force behind inclusive tools, and he’s also passionate about mentoring and helping shape the next generation of tech.

1:05:
So, , we’re really excited to kind of dive into this conversation with him today.

1:12:
Abhishek, let’s start with the big idea.

1:15:
You’ve kind of, you know, worked on tools that have really impacted and served many people, making technology more accessible and faster from that lens.

1:27:
How do you see AI not as a replacer but as a powerful enabler?

1:34:
Yeah, sure.

1:35:
So, , like I can dive a little bit into, , the kind of accessibility tools that I have worked on like, , Microsoft’s immersive reader was, , is a tool that exists, and, , it is like integrated in various, , Microsoft’s tool as well as it’s kind of open source, so a lot of third party websites and other sites also.

1:55:
have it as a part of their, , tools and it is an incredibly helpful tool for both the people with dyslexia or disabilities as well as, , like young children or people, , from not so educated backgrounds to easily understand, , documents and content that’s in the website.

2:16:
So.

2:17:
This has been a great source of impact that I’ve been able to deliver in the front of accessibility and inclusion to people, , enabling content to be easily, , like absorbed by people.

2:31:
And in this front, I, , seen like the newer, , AI enabled innovations have been an incredibly powerful tool, , at scaling this at a larger scale.

2:42:
, with, , co-pilot and Chad GPT, these AI tools, having the ability to, , naturally converse in any language, including various local as well as colloquial tongues, has, , like completely changed the scale at which, , these tools are able to reach.

3:02:
, like earlier when we started at a very earlier stage, I remember like the very first draft, immersive reader was supposed to be for English and a few Latin languages only.

3:13:
And then like since then it has increased over various like Hindi and other languages also, , Indian as well as Philippines and, , Latvia, other like South American languages, etc.

3:27:
, however, even in those aspects, it, it used to be just the most popular languages in those countries.

3:35:
Those, , like zones, but with, , how good LLMs have become today, , we are able to, , like scale this up to almost any local language that a person understands, and I think that’s an incredibly powerful thing that has happened that’s wonderful, yeah, and.

3:56:
or, you know, as an engineer, , what do you wish non-technical folks would better understand about how AI actually works?

4:05:
Because a lot of people are, you know, I think not having the right information also makes it makes AI kind of scary.

4:13:
So what are things that, you know, you wish that they would actually know better about how AI works?

4:21:
Yeah, , so I think if you, , go into house, let’s say LLMs were built.

4:28:
So they were built by training on existing human, , outlet text as well as things that humans have written or expressed in various modes.

4:38:
So AIs try to emulate what humans have been generating over the years.

4:45:
So I think A lot of people today when they think of AI, they think it’s like some high, high technology and like it is a technology, but you don’t have to be like daunted or like scared about how to adapt it or how to use it because AI you can think of as just another companion friend that you have.

5:05:
And because it is built on human language and human characteristics, you will be able to communicate with it in the same manner as you would communicate with a human.

5:16:
So I think whenever you’re trying to think if you can add an EI try to think can a human with like an average skill set to be able to do it.

5:27:
If they can do it, then there is some way we can make EI do it as well.

5:33:
So.

5:34:
That’s kind of the approach I take when I’m trying to think of how to use AI.

5:38:
Yeah, and let’s kind of dive more into this thinking.

5:42:
How can engineers make ethical thinking into system design, especially when building with AI?

5:48:
What are your thoughts on that?

5:50:
AI is like a huge field today that is not totally explored and is being like constantly explored as we are like taking AI more and more into everyday applications.

6:03:
And, , there are lots of like, , interesting nuances that has to be kept in mind.

6:08:
Like, , for example, the data on which, , AI has been trained highly determines its biases and, , tendencies that you see.

6:17:
So it is extremely important for the like owners or whoever is building these models like OpenA JGBD or Google for Gemini and all of the other models that are popular or not so much as well.

6:32:
We must make sure that the training set, , , like post training that they’re doing to fine tune their models are neutral in terms of, , the tendencies that they’re building.

6:45:
And this has been, I think, foremost in these big companies today.

6:50:
However, as more and more smaller companies are trying to fine tune their models, , they need to make sure that they are keeping this in mind and not just, , fine tuning their model to a specific group of people.

7:02:
That can like create bad biases or tendencies here.

7:07:
So that’s like one aspect of what I think like ethical things come, but then there is the other aspect where what are you using for your training and like.

7:19:
The, like recently, if I, , think about the Ghibli image trend that just started, so it’s like, , Hayao Miyazaki, who is the author of Ghibli Studio, like founder of, he’s a great animator, and he has this unique style that he brought and won the hearts of millions of people over the years in the anime industry.

7:39:
And share DVD just trained on these videos and these images that were found on the internet and now we can like generate the years of effort that higher music is spent on developing this technique and generate these images in like a second or so.

7:54:
, and that brings you to question whether is this should this even be allowed like that GPD got I think around 50 or 200 million new subscribers just from this trend itself and none of that benefit goes to the original author.

8:12:
So this, this opens up a whole set of like discussions that we can like dive deep into and it like can go forever for this podcast.

8:23:
No, definitely.

8:24:
That’s such an, no, definitely, that’s such an interesting aspect.

8:28:
Like, you know, this is where, , because AI being relatively newer as technology also doesn’t have too many constraints that are being applied, right?

8:37:
Even when it comes to data sharing, , like a lot of companies are also concerned about the privacy aspect in addition to the ethics.

8:48:
What are your thoughts on, you know, using data responsibility?

8:54:
Yeah, though that is, , foremost concern we have like at Microsoft Microsoft takes like data privacy and security is a very important thing.

9:03:
And as an employee, you feel it when you work there there’s constantly trainings and meals that you get trying to nudge you to make sure we are holding that privacy and security standard and that’s something I’ve seen like while working, we constantly have to keep in mind that when we are collecting data.

9:23:
From users are using it for creating tools, how we are, , first of all storing it and even when we are using it, where does the impact of those tools go?

9:34:
So like Europe is one country where they have like GDPR laws which are very like strict in terms of what Europe EU data you can use and how it should always remain within those boundaries.

9:46:
So these are things that we have to keep in mind when we are building tools, , whether it’s.

9:52:
tools or any other tools, , and comply with these things and I think, , that is very important.

9:59:
, so one is users privacy perspective, the other is the company’s own, , private like IP information or other stuff, , private information.

10:09:
So both of these need to be, , like kept, , in a way that it does not, , get leaked outside.

10:17:
And this is a very interesting challenge when you are building EI.

10:21:
, like, like if I go a little bit into the technical aspects of it, so you think like this chat GPT which can where you can use to search the internet.

10:31:
Now the easy thing about searching the internet is everything is public, so you just type something and chat GPT can just openly search the net and whatever it finds for you.

10:41:
But think of it in an end on.

10:43:
Enterprise setting you search for you are an employee, but you don’t have access to like the CEO level documents, right?

10:49:
But you just type a keyword to search an enterprise document and the internal LLM, if it doesn’t understand what who has access to what and has like indexed everything behind the scenes, it might just give you that confidential document to you.

11:04:
Definitely.

11:06:
So it’s very tricky and challenging to make an AI retrieval tool that understands every user’s like scope of what they can access, what they have authorization to and respect the privacy while bringing all these AI capabilities upfront.

11:25:
Definitely.

11:26:
And we’ve in the industry, we’ve had some standards that have been regulating the space, right?

11:32:
We have so to audits that are really kind of responsible for a lot of, like, you know, ensuring privacy and security, and even confidentiality in a lot of spaces.

11:43:
So how do we kind of, you know, rely on these standards with AI?

11:49:
How do you think that space is going to change?

11:52:
So, , most of the AI providers that you have today like Open AI or Google, they are providing such enterprise training, , like tools which respect these guidelines, , where, , if you are trying to fine tune a model to your.

12:10:
My own data, it is not like going to go to the overall training set of ChatGD or OpenAI in general.

12:20:
So there are tools that are being made more and more on the enterprise level, keeping these like constraints in mind.

12:27:
, because at the end of the day, I think besides the consumers, a huge chunk of the like cus customers that this, , EI services of these companies are having are the enterprises, the small and medium enterprises as well as the big ones.

12:44:
So you.

12:45:
Have to design, , these AI models in a way that are able to follow these guidelines and, , like encapsulate these the company specific data policies and everything restricted within their, , control.

13:03:
And I see a lot of progress happening on this front on like Open AI as well as Google, and that gives me like more confidence that enterprises will be more and more comfortable in including AI in their, , everyday activities and workflows.

13:18:
Definitely.

13:20:
And where do you think AI is still untapped, not just in terms of market, but also in terms of engineering workflows, user experiences or product strategy, like, , where can we, you know, still utilize AI?

13:35:
I think AI AI hasn’t reached its maximum potential in any field.

13:40:
I feel like we are still like just touching the tip of the iceberg.

13:44:
Like I can totally see that a single person can start an enterprise, enterprise in like a couple of years or so or even like today with a little bit of stretch, a single person has all the tools to write an entire software application to write a blog.

14:04:
to reach out to, , marketing and all of these stuff there are AI tools that somebody can leverage with the right ideas and mindset to start like to manage an entire, , company.

14:17:
But yes, this is still in a very early stage and the quality is still like not that far to make this a sustainable thing, but, , as more and more improvements are happening in the AI industry, I think that.

14:33:
We are getting closer to that point where people will be more like managers of AI employees and AI will be, there will be specialized AIs in every field that you can utilize to get an entire process running, so.

14:52:
Yeah, like there’s no one answer I have because I feel like whatever field you choose, I can think of areas that we can further utilize AI.

15:02:
So AI is like AI has potential to be included in everything that you’re using today.

15:07:
Mhm.

15:08:
Definitely.

15:09:
And we’ve kind of talked about this earlier, right?

15:11:
There are tools like lovable, reply that are really doing things that people developers used to do, but doing it at a much faster pace also.

15:21:
So what are your thoughts on the tools and where they are at right now?

15:25:
Have you experimented with any of these yet?

15:28:
There for software development, there is also cursor and like there was Windsurf, which are some of the more AI friendly advanced dev tools for developers and there’s this common term that is there everywhere now the white coding so.

15:44:
People are generating working prototypes of software just by talking to AI, which is really, really cool, and things are getting to the point where people are able to productionize some of these as well.

15:57:
So that’s like incredible achievement I think we have already reached in terms of AI.

16:03:
Like when I, when I just look back at like 2020.

16:07:
To November, I think that’s when AGBD dropped.

16:10:
Like I couldn’t imagine like in 2.5 years, this is where technology would be like an entire application that is being built just by AI itself working over minutes like this concept was so unimaginable just 2.5 years ago.

16:28:
Definitely.

16:29:
If I just see this trajectory at the rate we are improving and usually these rates tend to be exponential.

16:36:
, so I think like incredibly soon, like in a one in 2027, I can imagine total, , independent software engineers being able to walk and get hired, which are totally AIs.

16:52:
Definitely.

16:53:
Yeah.

16:55:
So are getting so out of hand.

16:58:
And so when we talk about this, there is the fear of AI kind of really replacing these jobs, right?

17:04:
Software engineers are concerned, even designers are concerned because there are tools like Stitch that are literally creating things that would take a lot of time, but doing it so much faster.

17:14:
So what are your thoughts on, like, you know, that aspect?

17:18:
Yeah, so that is a concern.

17:20:
I see a lot of in in my industry and elsewhere as well, and, , it is a fair concern, I think, as a lot of the easy tasks, , that, , software developers or designers or, , be, be the field that you think of we’re doing, , are like being easily done by AI today and, , like the more and more reliable.

17:40:
of these AIs are increasing.

17:42:
, these, these easy straightforward jobs will definitely get replaced.

17:47:
I don’t think there’s a doubt I have in that part, but where I think, , humans are still invaluable is on the creativity aspect.

17:57:
Like how do you prompt this AI or tell this AI to do and to do what?

18:03:
Like at the end of the day, EIs are still instruction following robots kind of you can think of that.

18:09:
Like they are really good at doing a lot of stuff and they will keep getting better, but at the end of the day they’ll be following instructions.

18:17:
So how good are you at giving detailed instructions?

18:21:
, and what you’re trying to exactly get out of it.

18:24:
These are aspects that humans will have unique, always, and the more you get better in your, , both critical thinking as well as creative thinking, I think you will be higher up in this hierarchy of like the unable to replace you type of thing.

18:41:
So, , that’s where I think people should try to upskill themselves, .

18:47:
Overall, I don’t think jobs will reduce.

18:50:
There’ll be newer jobs all the time, , new roles that haven’t been created like prompt engineering wasn’t a term, , 3 years back.

18:57:
, it is now a proper role in a lot of, , like it’s a skill that you should have expect everybody’s expecting.

19:05:
So yeah, there’ll be new roles, , getting created all the time.

19:09:
You need to keep up a little bit as technology is, , like improving at an ever increasing pace.

19:16:
So that definitely is a challenge.

19:18:
But overall, I think, , the number of jobs, the number of opportunities to create value in this world will remain constant.

19:26:
So, , like I’m not worried at least.

19:30:
Yeah, definitely.

19:32:
And prompt engineering is, you know, you talked about it.

19:35:
It is so essential, yet often misunderstood, , especially like, you know, you kind of having that technical background.

19:45:
Maybe you can help us understand what’s the biggest mistake people make when trying to use LLM models like Chad GPT in production systems and how does prom design play into that?

19:58:
Yeah, so, , I think prompt engineering, even though it’s like a fancy term that has been coined recently, is essentially just how well you can list down the instructions to follow.

20:11:
It’s mostly that and this is a skill that most people don’t like realize because in school we are more trained on following instructions rather than giving instructions like you know our curriculums are based on similar lines like we are given a question and you just have to answer it.

20:33:
But what if you’re given a like elaborate answer.

20:36:
And told you what are the different ways someone can question things in this.

20:40:
That’s not something we are taught like how to think critically and how to think in steps, how we achieve certain things.

20:48:
And because our mind is like over through evolution, it has, , kind of been trained in trying to create short, , like connections between things, things, and because of this, we tend to forget how we arrived at something like.

21:05:
, how should I put it?

21:06:
Like if you think reflects actions that we have gained over through evolution, it is a way to bypass the actual thinking process.

21:16:
So I’ll give a very crude example here, but like if you touch a hot surface, your hand moves away, right?

21:22:
That’s a reflex action.

21:24:
You don’t have to think about it because it’s something that has developed over us over the years.

21:29:
But if you are training an AI or telling an AI how it should behave.

21:34:
You like fundamentally cannot expect I will know suddenly that if it touches the hot surface it should move its hand or something like that.

21:42:
So you have to be self-aware about the intricate details of, , the problem and how the solution is being approached and able to write that is a very crucial skill.

21:56:
I think prompt the prompt engineers are in general, I think people should have to be able to solve problems more effectively.

22:04:
So this is one tip that I always like kind of suggest people when they’re trying to write prompts that think of the solution in as minor minute steps as you can before going ahead with it.

22:17:
No, that’s a great perspective.

22:19:
And also I really like that you kind of shed light on, like, you know, our education system also and how it needs to be more ready for sort of the changing environment that we’re in because everything is changing and upskilling really means now being able to learn what is important in the era that you’re living in.

22:38:
, I do want to talk about, like, you know, with the forces moving more towards agentic AI lately.

22:46:
, a lot of, you know, specifically in marketing, also, a lot of companies are really focusing on that.

22:54:
, so when does, you know, having human in the loop become critical, and when can we confidently trust AI to take the wheels, in your opinion?

23:06:
So, , I think, , like confidence in the system builds with like, , seeing more success from it.

23:14:
Like it’s very normal.

23:17:
Like I, I feel this personally.

23:18:
I’ve used more chat GBD and it’s like highest model of 03, and I naturally trust it way more than any other model that I’m using later.

23:27:
Like if I’m using, even if I’m using like Gemini’s latest 2.5.

23:30:
Pro or something like that or one of the deep or one or any of those, no matter what my trust in 03 becomes so much more because I’ve used it so long and I’ve seen it being very, very successful in a lot of complex tasks that I rely on it even in like important, like legal, sometimes legal scenarios, sometimes like bank financial scenarios.

23:52:
I Relying on it and there is like I of course I’m cross checking it eventually and I recommend that to everybody not to take AI advice blindly, but this trust kind of builds with seeing the board, the AI model succeed more and more in complicated tasks, and that comes from more and more, , usage, as you can see.

24:15:
So, , yeah, these agency AIs will be more and more adopted as we see people build successful things from it, and this success will encourage more and more people to try it, and that’s how I think adoption of this will increase more and more.

24:33:
Definitely.

24:34:
And as systems get more independent, I kind of want to talk here about like, you know, how should engineers really rethink software architectures, like what should they, you know, kind of be changing now?

24:49:
, yeah, so when I think of the future of software engineering, I think that the engineers need to go broader.

24:59:
Earlier people would be very skilled in, say, one language, and they would be very confident they can build whatever they need to using that language.

25:07:
But now I think language is no longer something you need to be very expert in.

25:13:
Tools AI, everybody can like write.

25:15:
, there are like very difficult languages like rust which are used for system level design and AI can write very good rust today and that is very, , like very, very impressive and that like a lot of software engineers I know including friends who are into like system level jobs where like.

25:35:
, learning rush from my college days thinking it will be very valuable in the future and it’s still valuable, but you have someone else write it for you now.

25:44:
So the more and more I think software engineers need to think at a broader level now, more on a design and architectural level, and the details of how to implement things will be more and more dedicated to agents and AI models as they are getting better and better over time.

26:03:
So, , the entry level software engineering roles today that are entry level, they may not be as valuable tomorrow.

26:12:
, the higher senior level or the architect level roles are going to be valuable.

26:19:
And the reason I think we will always have this one top layer of software engineers is at the end comes down to accountability like if you are a CEO and the software does not work you cannot blame a board.

26:35:
You have to have someone to hold accountable and.

26:39:
That’s why I think there will be always one senior role who might be owning a huge response like ownership now as people are becoming 100% more productive than before.

26:50:
So 100x more responsibilities than for financially.

26:54:
, but yeah, those 100 people that would have existed under you would, would be more like agents, , in the future.

27:02:
Definitely.

27:04:
Thank you so much for kind of, you know, sharing that.

27:07:
I do want to move into what we call the lightning round now.

27:11:
So this is just going to be short, quick, and, you know, really fast responses from your end.

27:17:
So, , if you could describe AI in one word, and it can’t be intelligent, what would it be?

27:26:
One word.

27:28:
, I would say a two word like a simulated human because I think it’s just stimulating us.

27:34:
I don’t think it is like actually intelligent.

27:38:
I, I don’t think it is thinking or has consciousness, so it’s trying to simulate our behavior.

27:44:
So if I can use a hyphenated word like a simulated human, then that’s what I’ll go with it.

27:49:
That’s a great one.

27:51:
what’s one AI capability you think is under use today?

27:57:
Mm.

28:00:
That’s a good one.

28:04:
Should I, it’s a difficult one, I.

28:11:
I would go with use AI to like generate ideas for how to use AI.

28:18:
-huh, that’s a good one.

28:21:
I think people are bending their mind trying to like think how to use AI, but AI is there for you as well.

28:28:
Like it’s just like another person you can think anything you can ask a friend or a person, you can actually ask AI and you can get some credible stuff from them.

28:38:
So I, I would definitely suggest that people should use it more and more.

28:43:
Definitely.

28:44:
What’s one startup that you think is doing AI the right way right now?

28:51:
Mm, one startup.

28:53:
So I used to think Perplexity was very good at using AI, but I recently I feel like, like as open AIGBT, everybody’s kind of using like has their features already, so Perplexity is losing its edge I feel, , other than that.

29:11:
.

29:12:
Like, I, I’m in a little bit biased towards software dev so I definitely think cursor is, , great, , to like startup that they are building a great product to make software development super easy through just ID.

29:27:
, so I’m really a huge fan of cursor, , and also like lovable also is doing it really well for people who are not devs.

29:36:
Like you don’t even have to be dev to build stuff.

29:39:
So lovable is there in my list as well.

29:42:
So, yeah, that’s on the top of my mind.

29:48:
No, that’s a really good one, and.

29:52:
You know, a system that you admire for balancing performance and ethics, so this doesn’t necessarily have to be a startup.

30:00:
So not a startup, but maybe, you know, another company.

30:05:
Oh, definitely Microsoft is there, , among the list.

30:07:
Microsoft puts a lot of emphasis on trying to be like following the ethics, the guidelines, and like it’s not just for the show on the outside, even as an employee, you feel that, , we constantly have trainings, we have, , security reviews and, , like lots of processes.

30:24:
While I feel that sometimes slows down the whole development process, but it’s extremely crucial when you are delivering products to millions of people around the world to make sure this, , this is followed and the standard is set.

30:37:
So I admire their commitment to, , ethics and privacy while delivering optimum performance.

30:44:
Definitely.

30:45:
And what, like, you know, which tool or shortcut do you use every day, but don’t necessarily get to talk about?

30:56:
Hm, .

30:59:
Shortcut.

31:00:
So like I use ChatGPD a lot of course, but one thing that I use it for like that I don’t maybe see a lot being talked about, maybe it is talk is, , I do some, , like self-reflection through the advanced voice mode.

31:15:
So at the end of the day I’ll just turn it on and talk about various.

31:19:
Things that happened over the day and ask it for the best insights from that and I think that really helps me like learn better both my patterns as well as what I could do better.

31:31:
And in general it creates a log for me also that I keep elsewhere that I later reflect back.

31:38:
So it’s a very handy tool.

31:39:
And I think the voice mode makes this process super quick because before I had to spend a lot of time trying to think everything and try to summarize insights from that manually.

31:50:
Now I could just like blabber for one minute straight and the chat you will just give you those insights in a second.

31:57:
So I love this process, yeah.

32:00:
Oh, that’s a great one.

32:02:
I love how, you know, you can kind of utilize AI not just for work, but also, you know, learning more and better about yourself and organizing yourself.

32:12:
So that’s a really great one.

32:16:
what do you, I guess I’m gonna ask this next.

32:19:
What’s one mistake founders make when trying to be AI first, according to you?

32:27:
, OK, so.

32:30:
While I can see something, but I don’t exactly have a solution yet in mind, so I think like there’s a lot of hype around creating wrappers whenever a new technology comes.

32:42:
This is also talked a lot about in a lot of other places.

32:46:
So when Chad GVD came, I remember there were like 100 different tools that came from as simple as some.

32:52:
Thing that helps you useGPD through PDFs, which is now a negative thing, but there were like companies, tens of tens of apps that came out just to be able to pass Excel sheets and PDFs using ChatGPT and people went nuts on building such like, , small, small wrappers for everything, and then all of those.

33:15:
We’re like swooped up in one go and JGBT introduced their Omni model and you can upload any form of documents or input and JGPT does it.

33:26:
So I think when you are building an AI product, think how your product stands out differently and is less likely to be.

33:38:
, like, , like a small tool that the big giants can just add on on their existing product.

33:44:
So, , that’s where I think, , the like the hype is a little too much on these sometimes whenever these new things come.

33:52:
like recently there was MCP servers that came, like the protocol came and while it’s a good thing, it’s like it helps AI is getting more agenttic, but the hype was way more than the need was and I think like now the hype is dying.

34:08:
Down, so, , there’s more and more like, , like startups who try to launch something, but now they do not have customers, , to get it because AGPT is just integrating those features in there or Gemini is integrating them.

34:23:
So try to make sure that whenever you are using AI, , your use case is not, , very generalized to the point that this large, , companies can just put it as a small thing in their arsenal.

34:39:
Oh, that’s a great answer.

34:40:
Thank you so much.

34:41:
This was an exciting round with you and definitely got a chance to learn a lot.

34:47:
, I kinda wanna, like, you know, as we end this session today, where do you kind of see the future of tech and AI and what advice would you have for, you know, young engineers that are in the field right now?

35:02:
, that’s, that’s a lot of responsibility to guide all of the young engineers.

35:08:
So I’ll just be able to tell what I am following because I think I am also very young.

35:13:
I do not think I’m like an old wise man to guide everyone, but the thing that I’m trying to follow here is that, , , I, I, , spent some time thinking that up to what level can EI automate us.

35:29:
So.

35:30:
There are two things where I think AI will always like we will not let AI really do it replace us.

35:38:
One is where money is because ownership of money doesn’t make sense for AI to own money and we will not like giving that to EI itself.

35:48:
So that’s one site and the other is accountability.

35:50:
At the end of the day, there has to be accountability for every action and reaction.

35:55:
So.

35:56:
Investors and CEOs are things that I don’t think are gonna be replaced.

36:01:
That’s like this very, very TLDR Investors are where the money is coming from, and CEOs of companies are the final layer of responsibilities.

36:09:
So if you’re trying to , see, go higher and higher, , in your scope of things that you’re building.

36:16:
If you’re, , if you’re trying like the skill set that you’re trying to develop, don’t try to have a very narrow skill because it creates higher risk in, , getting, , like substituted with some tool.

36:30:
So the broader you, , make your skill.

36:33:
Said the better you are and try to get better at managing either people or tools but whatever with AI it’s more like managing tools now if you think of them as tools.

36:46:
So the ability to walk with a diverse set of tool like tools as I said is a good thing like so.

36:55:
That’s something I would want people to be more comfortable doing and not just picking their most like best skill and being comfort zone of being in that.

37:06:
So you got to move out of that comfort zone.

37:08:
You got to learn end to end of, , any.

37:12:
, like business or product, , and as I mentioned, like at the end of the day, the value comes from the revenue.

37:22:
So you have to think of how your like impact or contribution is generating the revenue.

37:29:
You no longer can be just an employee writing code and not be responsible for anything.

37:34:
So that layer is slowly, slowly diminishing.

37:37:
So that’s what I would like people to keep in mind and to the young children or students who are now developing their skill set.

37:46:
Yeah.

37:47:
Awesome.

37:48:
Thank you so much for sharing that Abhishek, and this has been such an incredible conversation.

37:54:
You know, , you’ve shown us that good engineering is invisible when it works, but unforgettable when it makes someone’s life better.

38:03:
And I love the fact that, you know, we were able to dive into so many different topics today.

38:08:
If you are an engineer, a product leader, or even just AI curious, I hope this conversation helped you see that, you know, AI isn’t just about output, it’s also about the impact.

38:20:
And Abhishek, thank you so much for your time and your insights today.

38:24:
It’s been an incredible pleasure, Hushi.

38:26:
Thank you so much for having me.

38:28:
I thoroughly enjoyed the conversation, the questions, and everything was very smooth, free flowing.

38:35:
So yeah, love the experience.

38:38:
I hope our listeners also learned a few things at least.

38:42:
And, , feel free to reach out to me or Hershe.

38:46:
I think we are all available on LinkedIn.

38:48:
, so, yeah.

38:49:
Mm.

38:50:
Yeah, definitely.

38:52:
I think if this episode gave you something to think about, then definitely share it with your fellow teammates and also let us know your thoughts.

39:00:
Thank you so much again for tuning in and we’ll see you next time.

podcast