Wow , the article was really good , it was very insightful , thanks for sharing this @Andres Vourakis.
But i had a very important concern / doubt regarding automating data analytics tasks.
So i feel things which are very redundant in nature like we should use these agentic ai systems to just detect problems in our data like missing data , outliers etc... treating those would be little scary.
Coz i feel like still AI is not that intelligent , like my real example , currently i am doing a data analysis project , the level of deep critical thinking which i did for treating missing data i am very sure no AI model would do that as currently they are not that smart enough.
So i feel like yes we should definately embrace AI Automation but carefully integrate them as always Human Intervention would be needed to get the most accurate results , am i right ?
Great reflection, and you are right, AI still can't fully match the level of critical thinking we have to deal with on a day-by-day basis, BUT this gap narrows when AI systems are given our business context and a clear set of rules to follow (I'm speaking from experience here, having built my own AI agents for data science work).
In my view, you can get surprisingly far by focusing on robust systems: human-in-the-loop, feedback loops, guardrails, evals, etc.
Regarding the topic of agentic AI, this really resonates. Your breakdown of the shift to agentic analitics is very insightful. What if this paradigm fundamentally alters the iteration speed of new AI models, making traditional software development cycles obsolete? It's exciting to consider the implications.
Happy to hear you found it insightful. I’m going to be honest, it’s exciting but also a bit scary, I think it will mainly remove the friction from data to insights, but it will also take away some “manual” work that some of us found fun to do.
Sir, I read your post and it was amazing that agentic ai is going to be a hot topic in 2026, but as a fresher what should we focus on , i am currently in pre final year of engineering.
If you are a student, you main goal should be to build a strong portfolio to land your first role. That’s it, don’t worry about Agents and other more advanced topics right now. Focus on proof of skill and demonstrating business readiness. I have lots of articles on that topic.
When you say, "Conversational dashboards for fast insights
Think dashboards you can talk to, ask questions in plain English, and get instant summaries or visualizations."
Are you thinking the entire dashboard is a conversational model with an AI Agent helping?
My experience with users is they often have a hard time articulating the thing they are trying to solve for. So a dashboard is designed not as much to give ansers, but to describe what has occurred and spark questions.
I use a customer portfolio approach with retail(ecommerce) data and 6 customer portfolios: new, reactivated, stable, growth, declining, defected and then quintile those portfolios so you see top performer and bottom performers.
The goal is NOT to answer every question, but to illustrate things like revenue concentration in customer portfolios. And to get the user to ask questions like "What sets apart the top quintile customer from the third quintile customer and can they be directed to products or categories to get them to the first quintile?"
is that what you mean by conversational dashboards? And are the SLM's better for that purpose or does it really matter which model you choose?
Very curious about the conversational dashboard bc it is on my roadmap.
Wow , the article was really good , it was very insightful , thanks for sharing this @Andres Vourakis.
But i had a very important concern / doubt regarding automating data analytics tasks.
So i feel things which are very redundant in nature like we should use these agentic ai systems to just detect problems in our data like missing data , outliers etc... treating those would be little scary.
Coz i feel like still AI is not that intelligent , like my real example , currently i am doing a data analysis project , the level of deep critical thinking which i did for treating missing data i am very sure no AI model would do that as currently they are not that smart enough.
So i feel like yes we should definately embrace AI Automation but carefully integrate them as always Human Intervention would be needed to get the most accurate results , am i right ?
Great reflection, and you are right, AI still can't fully match the level of critical thinking we have to deal with on a day-by-day basis, BUT this gap narrows when AI systems are given our business context and a clear set of rules to follow (I'm speaking from experience here, having built my own AI agents for data science work).
In my view, you can get surprisingly far by focusing on robust systems: human-in-the-loop, feedback loops, guardrails, evals, etc.
Could you share your linkedin , i wanna connect with you sir.
Here it is: https://www.linkedin.com/in/andresvourakis/
Thank you Andres for the useful information
My pleasure Preeth. Glad you found it insightful
Regarding the topic of agentic AI, this really resonates. Your breakdown of the shift to agentic analitics is very insightful. What if this paradigm fundamentally alters the iteration speed of new AI models, making traditional software development cycles obsolete? It's exciting to consider the implications.
Happy to hear you found it insightful. I’m going to be honest, it’s exciting but also a bit scary, I think it will mainly remove the friction from data to insights, but it will also take away some “manual” work that some of us found fun to do.
Sir, I read your post and it was amazing that agentic ai is going to be a hot topic in 2026, but as a fresher what should we focus on , i am currently in pre final year of engineering.
If you are a student, you main goal should be to build a strong portfolio to land your first role. That’s it, don’t worry about Agents and other more advanced topics right now. Focus on proof of skill and demonstrating business readiness. I have lots of articles on that topic.
Best of luck!
When you say, "Conversational dashboards for fast insights
Think dashboards you can talk to, ask questions in plain English, and get instant summaries or visualizations."
Are you thinking the entire dashboard is a conversational model with an AI Agent helping?
My experience with users is they often have a hard time articulating the thing they are trying to solve for. So a dashboard is designed not as much to give ansers, but to describe what has occurred and spark questions.
I use a customer portfolio approach with retail(ecommerce) data and 6 customer portfolios: new, reactivated, stable, growth, declining, defected and then quintile those portfolios so you see top performer and bottom performers.
The goal is NOT to answer every question, but to illustrate things like revenue concentration in customer portfolios. And to get the user to ask questions like "What sets apart the top quintile customer from the third quintile customer and can they be directed to products or categories to get them to the first quintile?"
is that what you mean by conversational dashboards? And are the SLM's better for that purpose or does it really matter which model you choose?
Very curious about the conversational dashboard bc it is on my roadmap.
The trending part of ai hype everywhere is like it will take a jobs but no one wants to utilise it better just AI hype 🤖