EP 517: Balancing AI Productivity and Human Intelligence in Everyday Work
33:51 Share

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. I've been struggling with this one for years. There's so many instances when I want to personally use generative AI in large language models to maybe help me complete a certain task faster, maybe to go into more depth. But then I think,

okay, yes, this is going to save me some time. It's going to make me more productive. But am I just going to get dumb over time, right? The more I blindly hand off knowledge work to a large language model, right? So it's something I'm excited to talk about today, kind of this dichotomy of, okay, how can we still use the best technology that I think any of us have ever seen maybe in our lifetime, yet still, okay,

Keep our brains sharp, right? How can we find that balance of being productive, but not just getting dumber at the same time? This is going to be a fun one. I can't wait. What's going on, y'all? My name is Jordan Wilson, and welcome to Everyday AI. This is your daily live stream podcast and free daily newsletter, helping us all not just learn and learn,

AI, but how we can leverage it to grow our companies in our careers. So it starts here with this very podcast. I'm excited for our guests for today, but it continues in our newsletter. So if you haven't already, please make sure to go to youreverydayai.com, sign up for the free daily newsletter. We're going to be recapping some of the most important points and takeaways from this very conversation in

laying it all out for you, right? So you can actually take this info to grow your company and your career. All right. I'm excited for today's guest. So please help me welcome to the show, bringing him on. There we go. We have Sumit Gupta, who is the lead BI engineer at Notion. Sumit, thank you so much for joining the Everyday AI Show.

Yeah, no, thanks. Thanks a lot, Jordan, for the welcome and morning. All right. I'm excited for this one. Before we get into this topic, which I'm going to try hard to shut up and let you talk a lot because I've got a lot of feelings on this one, Sumit. But can you first tell us a little bit about what you do in your role at Notion? I'm sure most people know Notion, but maybe just for those that don't as well, maybe if you could describe that as well.

Yeah, sounds good. So I lead the BI engineering at Notion. I support sales and marketing team here. If you've never heard of Notion, so Forbes like to call Notion as AI everything app. You can think of Notion as a blank canvas and you can do pretty much everything.

Uh, you can, you can do notes, you can do database automation. You can, you know, create, uh, your own Gantt chart and project management notion can pretty much do everything nowadays. Right. Uh, and I, I come from a place in India. I was born in Bombay. I've been in us for 10 years and, you know, um, and have been using AI like

in professional life for a couple of years now. And I'm pretty excited to talk about the topic. I'm actually like, as much as I'm excited, I'm also like, it's a topic which is very close to me. And I think it's the same with you also. So let's get going. All right, let's get to it. Is using AI making us dumb?

Yes. The short answer is yes, but allow me a couple of minutes to explain that part. Right. In my role, right at Notion, day in day out, my job is to write SQL query. My job is to create dashboards and reports in Tableau.

Before AI, let's say if I wanted to create a Tableau report, it would take me three to four weeks. Generally, there's a lot of 2 or 4 involved with stakeholders and I would go and write a lot of calculated things. I'll do a lot of joins. I'll iterate over charts.

But nowadays, whenever I get a request from my stakeholders, I'm like, okay, I strategize of how the dashboard might look like. What do I need? What calculate will I need? And then I feed that data into AI. You wouldn't believe I have premium subscription of pretty much anything, every AI that you can imagine. GPT, Publicity, Cloud, like, you know...

all the AIs have their own strengths. So in my case, I would go to, let's say, Claude and be like, you know what, create a calculated field for financial year and comparing financial year metrics. Claude will split out a 200-year calculated field. If I was to do that previously, if I was to do that, it would take me at least four to five hours to even get it right. With Claude,

It does not even take four to five minutes. It probably takes like less than a minute. Right. So,

It makes me productive 100%. Instead of spending two days writing calculated field, I'm finalizing my calculated fields and my metrics in two hours. But the counter side counter of that is it's made me lazy. Let's put it that way. And when you're lazy, you're not thinking really hard. When you're not thinking hard, you're dumber.

It's so true, right? And I'm curious, Sumit, through your personal experience, because I remember the early days back in 2020, right? Before ChatGPT came out, there's all these essentially writing tools, right? And I was a journalist. And so I was using this technology early on and I'm like, okay, this is good. And then it started to get to a certain point for me, even before ChatGPT came out, that I'm like, wait,

I'm starting to turn off my brain in some instances, right? And you have to start keeping that in check. I'm curious, what was the point for you, right? When maybe the power of certain LLMs started to really get really good and you're like, wait, do I have to sit back and look at how I'm using these? Did you have a certain point where you were like, whoa, I have to make sure I'm still using my brain?

Yeah, yeah. I remember the time that you're talking about, because I think before GPT came out, there were like article rewriter services, right? Like Copy AI and those kind of companies, Word AI, etc. And those were great. But GPT came along and then you're like, wait a minute.

Is this real? Right. And I suppose like specific instance of when I felt this is like life changing. Right. When I was at Snowflake, I was tasked to build like a streamlit report and then I'm not a Python wizard. Right.

So I started using GPT to debug my code, part of my trying to debug my code. Because previously you would go on Stack Overflow, you would research, you would spend a couple of hours and you try multiple different coding solutions to see if it's what's working. And then with GPT, I was able to do that in less than 30 minutes. And this is like

early to 2022 late 2022 right where models were still not like great right it was it was great at the time but now when you think about it how cloud works right with the coding output etc and with cursor ai etc right in 2022 the fact that i i did not need a mentor or a coding buddy to actually review my code and get the code working that was a mind-boggling moment for me so

What's your advice? Maybe for people who are struggling, and I'll even say one of my personal examples, I said I'm a journalist. I do a lot of writing. I write a daily newsletter. And there's some instances where I use AI, and then there's some instances where I don't. Because at least I feel if I start handing off one of my most polished skills that I've been getting paid for two decades to do, if I just keep handing that off,

I'm going to lose that skill, right? How can business professionals out there, Sumit, start to, how can they find that balance? Yeah. I think in my case, I'm not to a point where I'm handing my life off to AI yet. And I don't think so I will ever do that because when you're in tech, if you're not learning, if you're not critical thinking, you'll be out of job in less than six months.

and I'm completely aware of that and to be honest I only use AI for things that I am already great at or you know you need some support. I wouldn't go about and be like oh I'm an AI researcher, writer, technic, write a research paper of 30 pages. That's BS, that's not the best use of my time. Even if AI could write it

that's not who I am, that's not what I want to do. So in short, I want AI to do the repetitive tasks that I generally used to do previously and in the area that I am already expert in. I'm an expert in Google search but nowadays with the cloud just launched web search, little to the itself but before that, I would ask public city to be like you know what, like literally the early last week I've been looking to like

I do a lot more speaking gigs this year. I went to public city. I'm like public city is a problem. Go about and search like data conferences or analytics conferences which are still open for speaker. They have called for speakers. Right. And public city gave me like 25 different links. If I was to do that, right, it would take me a couple of hours, if not more to even do that. So things that I'm expert in and which has a repetitive element to it, that's where AI shines.

Yeah, that's a great point. Yeah. Anything with research, repetition, right? Things that don't like, you know, yeah, if I'm researching something manually and I have 30 tabs open, like I think maybe that process is making me dumber, right? Not by skipping it is probably making me smarter, right? Not getting stuck down all these rabbit holes. But Samit, I'm wondering, is there anything

Any downside, right? Like specifically like cost, right? Because sometimes, you know, as an example with vibe coding, right? You know, because I think this is huge, right? When it comes to, you know, coding, software development, everyone's just, you know, sometimes maybe turning off their brain too soon. Is there a downside even to elements like that?

Oh, I think we lost your audio there, Sumit. There we go. Yeah, yeah, yeah. So I think the white coding moment is here. I think a couple of weeks back, YC, right, Y Combinator published a post where 90% of the YC-backed startups are white coding. Like, they're apps completely based off AI coding, right?

I mean, that's great. But I think I have a couple of examples that I would want to talk about. One of them is like, I read a post on Twitter, as we used to call, someone wrote a code and, you know, they used AWS access key, etc. And then the backend wasn't very secure, right? And then hackers were able to access backend and the founder saw like a million dollar bill.

right because when you're white coding you you're not you're not thinking of backing you're like oh yeah the website looks pretty that website looks great you know i'm ready to go ready to get that funding uh so that's one instance another instance is recently uh one of our non-technical stakeholders yeah white code and you wrote a sql query the sql query with the joins and everything else should have taken less than like

50 dollars to execute on our platform like on snowflake but it took over ten thousand dollars because when you're querying like trillions of rows if you're not careful if you're not in if you're not efficient in writing your query the select star statement can cost you thousands of dollars and that's what happened right

So that's the downside of using AI without really like when I think of AI, I think of myself as a guardrail for AI. AI can do what I ask them to do, but it's my job to make sure there are guardrails around it and make sure that it's working, but it's not costing me dollars than it should.

Yeah, that's a good point. And so for those that aren't familiar, vibe coding, that's essentially when either non-technical people or people who know how to code, instead of just going in there and writing it by hand, they're just kind of passing it off to maybe a tool like cursor or something like that. But if you're using an API and you're paying the usage,

Yeah, those bills can skyrocket, right? You know, I'm curious, Sumit, is there anything that you're doing, you know, actively, you know, to ensure that you're still staying sharp, right? That you're not getting dumber, right?

right? I think this is something I'm asking you, right? Because I need to find these things myself because I sometimes find myself over reliant on large language models, right? What are you doing in your day-to-day to make sure that you don't hand off too much and, you know, finding that balance? Yeah. I think my way of thinking, as I alluded to it, right? My way of thinking of

Best use case of AI is a repetitive task. Anything, let's say, as I mentioned, especially in my case, when I'm SQL querying or when I'm writing or building a dashboard, the first iteration is always mine. I'll go ahead and build the dashboards because I don't want to lose that touch.

But once that first iteration is done, generally, after a point of time, the way you have created a writer's block,

engineers and data visualizers like me can have that block too. So I'm like, okay, what do I do next? How can I improve? That's why AI comes into picture because think of AI. What's LLM? LLMs are being fed with a million visualizations, right? And they know what to look, what's good, what's great, what's bad, right? So you can take that suggestion. And that's what I do. I take the suggestion. I'm like, okay, how do I improve this chart? How do I improve this query? Can I make this query more efficient? And that

I have that cut off right without that cut off I would be even more number I'm I mean I know I keep using number as a very loosely but not that I'm dumb like I work for notion right I mean obviously we are all smart here but the fact that

that cut off is really important for me i i am hoping and hoping and praying that that cutoff does not get lower and lower and lower as as the day is fire and it should ideally get higher and higher and that is why i wanted to talk about this topic because it's an emotional topic for me itself where i'm like it's an it's an inflection point where i'm like i don't want to get this cut off lower i want to keep this keep keep a high bar up you know

making sure that I stay on top of my feet, I critically think and I be in the job. The market is already rough. You don't want to make it easy for folks to not hire you.

Yeah. And now's a good time to point out, live stream audience, if you do have any questions for Sumit, make sure to get them in now. There was a recent study, we covered this one in our newsletter. It was from Microsoft. It was a great, great study. It was called the Impact of Generative AI on Critical Thinking. And it found that for knowledge recall, 72% of participants

reported using much less effort or less effort compared to not using Gen AI. So to me, like I think, right. And you know, maybe this isn't a stretch, right. But I think of the brain as a muscle, right. The more you use it, the smarter you get, the less you use it, the more you turn it off, right. The dumber you get, right.

And, you know, we hear constantly about how, you know, these LLMs and the benchmarks and, you know, these pseudo IQ tests and, you know, like,

It's getting to the point where, you know, as we quote unquote approach AGI or ASI or whatever you want to say, right. That any large language model is going to know more than any human. So like, how should we be addressing this? Right. Because part of me is like, okay, maybe it's good to just give more and more to these large language models because they're smarter and they know it. And I don't right. Versus, okay, well, I'm not then developing new skills, right.

Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Gen AI. Hey, this is Jordan Wilson, host of this very podcast.

Companies like Adobe, Microsoft, and NVIDIA have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for chat GPT training for thousands,

or just need help building your front-end AI strategy, you can partner with us too, just like some of the biggest companies in the world do. Go to youreverydayai.com slash partner to get in contact with our team, or you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on Gen AI.

Yeah. So that's exactly where you should have your own, or like, you know, cut offs, or, you know, baseline of when you bring AI in your life, you cannot be bringing AI in your life or everything. I mean, you might be tempted to, right. But

it's like the way I think of it as like you're getting short-term results for long-term losses right that's how that's how I view AI if I start using AI for everything in my life let's say if you're a student right I think I saw a post like last week from a Berkeley professor and he's been teaching like database and database for like 10 years

he noticed that people have stopped asking a lot of questions in the classes now like students but then he also realized that he's like oh maybe I'm I I've just become so great that students don't have any questions but then during tests right midterms either the average score was the lowest it has been in last 10 years right because

Because students actually without even trying to understand or learn, they're using AI to write assignments. But then when you have a midterms in the class, you don't have access to, let's say, GPT. You are not going to recall what was thought because you just weren't focusing or you just did not stretch your muscle enough in your assignments to retain or recall that information.

So that's a very good use case of where AI is actually impacting even younger generation. Because at least for us, we come from a place where we still have built that muscle. When we were growing up, we had to do it because GPT did not exist. I wrote a book, The Tableau Workshop. It took me a year and a half to write that book.

book. Nowadays, if I was to do the same thing again, it would take me less than three months because I know for a fact a lot of things would be GPT driven. Right. So the fact that, you know, having that guardrail, having that cut off for when to use AI and when not to use AI is going to be as critical as the fact that your prompts are supposed to be great.

Yeah, I think you bring up, yeah, a lot of good points there. And I'm sure there's going to be, you know, many more of some of those types of stats, right? That just show, hey, you know, scores across the board on certain standardized testing are probably going to continue to go down, especially when you can't use AI, right? Because I think that people are maybe leaning a little too heavily on it.

So even for myself, right, I think the productivity piece maybe has that downside, right? Because I've never felt so productive in my life. I've never felt that I've learned productivity.

so much in my life, right? Since AI came out, you know, you mentioned, you know, being able to use something like a, you know, quick search tool, perplexity, deep research, grok, deep research, open AI, deep research. So I feel I'm able to bring in all this knowledge, but then I just feel I forget it, right? Like, like, do you ever find yourself? Yes. Now I'm more productive and more knowledgeable, but I'm still not retaining anything. Like, do you ever feel that way? Yeah.

Yeah, I can give a specific example. I think everyone who's listening in will understand. So when you type something in Google, let's say, right? And if you make a mistake, then Google says, did you mean, let's say, a word, XYZ, right? And you do not even try to correct it. You just click on the correct spelling of that word.

And you know, Google searches for you. Imagine your ability to recall. When you're typing, you almost know that the spelling is going to be wrong, but you hope that Google is going to correct you. That happens far too often. And then to your question again, I have started noticing that when I used to write SQL query myself,

I would know how to write case when statement, I would know how to write, let's say, a special function in SQL. But now if I don't have GPD, I have to like, okay, how do I write even case statement, right? Because case statement can be complicated. So when that inflection point comes in, I'm like, I don't think so this is working for me, right? AI is making product. But then if the fact that I can't retain information anymore, I am becoming more of an AI than a human.

Yeah, that's a good point. You know, as, as, as weird as this sounds, I like, I find in, in this is for me why this topic really does hit home. Like I'm not exaggerating. I'll look up things on, you know, chat GPT search perplexity, you know, Google search, whatever on a certain topic that I'm trying to learn. And in so many cases, and I, I kid you not in so many cases,

a lot of the first sources are myself, right? And that, and like, and that, you know, cause I covered something six months ago or I had a guest on, or I did a tutorial, right? And for me, that's that it almost hurts because I'm like, wait,

I'm learning from myself and I forgot this and this was three months ago, six months ago. How did I forget this? Right. Is, is like, is that going to become, do you think more and more common, just our ability to retain information for a longer period? Right. Like I find myself struggling with this often. Yeah, that's going to happen. If we, if we don't,

See, as I mentioned with our pre, let's say, 90s born folks, right? Pre 80s or 90s born.

at least in our case, we grew up with that critical thinking ability, right? We were supposed to do our own job. So that's why like, we both are able to talk about AI and how it's making us number, right? Talking about cutoffs and guardrails for folks younger than us. Like, if you're 15 or 16, right now, you have like, the whole computer in your palms, right? And you are like using AI for everything in your life.

right uh so uh the fact that you do not go through that muscle bending exercise for like a decade or two right uh you you don't have that retain ability to recall or retain that much your muscles not built right uh your brain muscles just not built uh but i hope that

as much as AI is a boom, right? It's helping folks. It does not turn into a course in five to 10 years where, you know, there's let's say now there's 10% of world population who can get, let's say job in IT because of, you know, their critical thinking ability. But in 10 years, I hope there's not a

That number still stays at ten or above, not goes down to, let's say, one or two percentage because the top one or two percentage who know how to use AI in the job are going to be millennials, right? Are going to be in the market in the demand for years to come. But that top one to two person are not going to run the world. You need more people, right?

right you need critical thinking ability you need people who can retain and recall and you know use ai to like as an assistant not replace a real human with the ai ai is your assistant not not like your personal full-fledged human who can do everything that you can imagine it should be complementary it should be complementary ai should always be complementary to you not the other way around

Yeah. So, you know, anyone that's listened to the show, like you all know, I hate when, you know, people come on here and pitch their products, but, you know, Michelle is asking, and I, like, I feel Samit, you have to take this one because, you know, asking, tell us more about Notion. Can it help with this retention? Like, I got to give you that one because it's a softball because it's great for that.

No, 100%. So if you've never heard of Notion, I highly recommend you. We have a free trial. I'm not representing Notion here, right? It's all personal opinion. But go ahead, go to notion.com, sign up for a free trial. You would be surprised with what Notion can do for you. Obviously, it's a note taking app, but Notion now also has like Assistant AI, which basically if you connect your, let's say Slack, you connect your other tools, right?

and if you ask question it's like a knowledge ai now where it will search through your post messages on slack it will if you if you were to use notion in your professional setting it will search through your docs and give you like relevant information right gpt works great but it does not have access to your personal information right or you know your company information or knowledge base in the company and that's where notion ai shines the fact that you can i i i have had so many instances with notion here and i'm like

Did someone talk about, let's say a metric, right? Because I've been at Notion for less than a year. So there have been metrics that the company has been using for multiple years, right? But okay, when was the first time someone mentioned, let's say, month over month retention for product users? So Notion here actually goes through the whole Slack history that we have and gives me like exact

link and conversation context of okay this is when let's say director of demand gen was talking about this right so the fact that notion ai does that for you is amazing right obviously

That's one of the features of Notion AI. The fact that you can create nodes, you can create automations. If you are new to Notion AI, I would highly recommend you just Google Notion templates. People have their whole brain on Notion. There's the whole concept. Jordan, you would know like there's a whole concept of like second brain on Notion. Just Google that term second brain on Notion and you would be amazed with what people can create.

Yeah, yeah, I do agree. And I was like, hey, perfect, perfect chance there to answer a question super directly. But, you know, this one might be controversial. So, Sandra, thanks for this. You know, she's asking, could this be the end of the knowledge worker, right? You know, for decades, we've gotten, you know, we've gotten jobs, we've gotten promoted, right? Companies have grown into unicorns, trillion dollar companies based on what its people know.

So me, could, could this, could we be not now, right? But could we be seeing the end of the knowledge worker anytime soon? I, so I know folks talk about, talk a lot about, Oh, is AI going to take my job or is AI going to replace me? I think that's partially true for jobs, which has a lot of repetitive aspect to it. Right.

And I believe knowledge workers fall in that category. But the counter argument to that is, so if you're a knowledge worker and if, let's say, you were doing, let's say, ten tasks a week with AI, if you upgrade yourself, if you know how to use AI to complement your job instead of doing ten jobs, if you're doing 40 jobs now, because AI can hallucinate.

we all know that right uh ai is not perfect like any any time i see an ai output for a metric or something i generally go about and you know verify that myself right and as a knowledge worker

if your job wouldn't be now to research a lot, but to validate what the research is saying. So instead of doing ten tasks, if you upgrade yourself to do, let's say, 30 tasks, 40 tasks, you're always going to be in the demand. But if you are one of those lazy workers where you're still stuck at ten and you ask AI to do 90% of your job and let's say you do a report for a CD of your company and the CPU uses that number in a board deck.

And that number is absolutely weird and it does not make sense. You're not going to be in your job because your job is to validate and verify and obviously get that info. But if you don't validate and verify, you're going to be out of job. And that's not a good thing.

I think that's a great take there. Yeah, yeah. I'll have to save my hot takes on that one for another day. But Samit, so we've covered a ton in today's episode, you know, from some of your personal examples and how, you know, sometimes using AI might just be more expensive. It might be better, right? It might be making us dumber in the end.

But, you know, as we wrap up, what's your one most important takeaway for business leaders out there grappling with that same thing? You know, feeling, yeah, I'm becoming more productive, but I might be coming a little dumber. What's your what's your best takeaway for everyone?

Yeah, don't change the hype. I think the hype right now is MCP and model context protocol and white coding. Don't change it. Think of your specific use cases. Think of all the repetitive tasks that your company has or you are doing and think of

Can AI help me here? Don't go chasing, can AI create reels for me with one click? Yes, it can, but is it going to add value to your audience? Probably not. Think of repetitive tasks where

you know, you can use AI as a compliment, but don't think of replacing AI as your full-fledged assistant yet. And the way I like to believe is like, this is the worst AI will ever be.

That's how I like to think. Two years ago, Cloud AI, the coding output wasn't great. Now, Cloud AI can build your website in less than two minutes. Like a full-fledged gaming thing with Unity, etc. So, to business leaders, as I mentioned, think of all the repetitive tasks that you can automate.

Use any time, use make.com, automate your workflow, but don't chase the cloud and don't chase those clickbaity titles. You can go on YouTube and you can find people talking about how AI helped me structure my life. That's great. But those are just for recording purposes. You go talk to them. That's not a real use case.

Love it. I think some great takeaways and some wise words for those of us out there, you know, struggling to walk that tightrope between, you know, productivity, but still retaining all of that knowledge that makes us special, unique humans. So, Sumit, thank you so much for sharing your time and talents with the Everyday AI Show. We really appreciate your time. Yeah, thanks a lot, Jeremy. Thanks.

Thank you. All right. All right, everyone. There was a lot of great info there. If you missed anything, if you need to hear it again, if you need to know more, our newsletter is where that's going to happen. So if you haven't already, please go to youreverydayai.com. Sign up for the free daily newsletter. If this...

was helpful, tell someone about it. Right. So if you're listening, maybe here on LinkedIn or Twitter, just go ahead, repost, retweet, re-X, whatever it's called. Help someone out. Also, help us out. Tune in tomorrow and every day for more Everyday AI. Thanks, y'all. Yeah, thanks.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.