232: Why AI Hasn’t Replaced Data Analysts (According to an ex-Google Data Analyst)

Than about your business. And the truth is that at any given company, the context is actually spread across a dashboard, documentation, code, Slack, conversations, and the memories. What is it called? There is a term for it. Tribal knowledge. Yes, years of tribal knowledge that lives in people's head. Buying an AI tool does not fix that foundation. And if your company has recently adopted AI to do data analysis work, let me know how that is going. The second reason why AI has r- failed to replace data analyst is human judgment. Yes, AI can produce an explanation that sounds reasonable. You still need to know whether it makes sense. Simple as that. I once used AI to understand why clicks from Google search had dropped, right? Simple question. The AI actually suggested failing search demand, uh, because of a ranking change, broken tracking and whatnot, and mix of queries. Like, that was a mumbo jumbo thing, but that was actually not the right explanation. All of those theoretically made sense, but somebody like me who actually had the business context, I actually knew the real cause why the traffic went down was because Google ended something called continuous scroll. So instead of, like, continuous scrolling through the search page, you actually had now page numbers. So a user had to specifically click on the next page to land on basically the next search results. And when you add more clicks, you actually, uh, increase user fatigue, and people are less likely to do it, and that was the cause of the traffic drop, which AI was not able to catch. This is an exam- Example of business context and human judgment. In the real world, you actually needed to have the business context. You needed to have the history, product history, and have to know Interface itself to basically figure out why traffic was down. So now you must be wondering which data analyst should actually be worried about AI. Let me just be very transparent. If your value comes from pulling a number after someone else has defined the question, you're just writing basic SQL, a refreshing dashboard, and summarizing changes, that work is becoming easier with AI and more likely to get automated. The more- A durable analyst defines the metric, challenges the request, understands the domain, validates the output, and connects the analy- analysis to a decision. Remember decision? We actually haven't talked about that. Let me know if I should cover that in a different video. This could be a product analyst who understands user behavior, a growth analyst who understands acquisitions and retention, or an analytics engineer who builds foundation AI needs Or a data analyst who actually is focused on defining business problem and decision-making. The common thread in all of these role is ownership, context, and judgment The role is Ownership, context, and judgment. So has AI replacing data analyst officially failed? I'm happy to share that it has failed to eliminate the role. It has succeeded at changing it, but it has not replaced it And this is true for majority of the tech roles. The pro- Coding only version of the job is, yes, under pressure, but the analyst who understands the business, use AI well, and knows when to answer is wrong, answer is wrong, is becoming more important. If you work in data analytics field, let me know what has been your experience. Do you agree with this? Do you think AI has failed to replace a data analyst, or we're not there yet? Let me know in comments. With that, I hope you're having a beautiful day. I'll see you in the next video. Have a good one. Bye. Than about your business. And the truth is that at any given company, the context is actually spread across a dashboard, documentation, code, Slack, conversations, and the memories. What is it called? There is a term for it. Tribal knowledge. Yes, years of tribal knowledge that lives in people's head. Buying an AI tool does not fix that foundation. And if your company has recently adopted AI to do data analysis work, let me know how that is going. The second reason why AI has r- failed to replace data analyst is human judgment. Yes, AI can produce an explanation that sounds reasonable. You still need to know whether it makes sense. Simple as that. I once used AI to understand why clicks from Google search had dropped, right? Simple question. The AI actually suggested failing search demand, uh, because of a ranking change, broken tracking and whatnot, and mix of queries. Like, that was a mumbo jumbo thing, but that was actually not the right explanation. All of those theoretically made sense, but somebody like me who actually had the business context, I actually knew the real cause why the traffic went down was because Google ended something called continuous scroll. So instead of, like, continuous scrolling through the search page, you actually had now page numbers. So a user had to specifically click on the next page to land on basically the next search results. And when you add more clicks, you actually, uh, increase user fatigue, and people are less likely to do it, and that was the cause of the traffic drop, which AI was not able to catch. This is an exam- Example of business context and human judgment. In the real world, you actually needed to have the business context. You needed to have the history, product history, and have to know Interface itself to basically figure out why traffic was down. So now you must be wondering which data analyst should actually be worried about AI. Let me just be very transparent. If your value comes from pulling a number after someone else has defined the question, you're just writing basic SQL, a refreshing dashboard, and summarizing changes, that work is becoming easier with AI and more likely to get automated. The more- A durable analyst defines the metric, challenges the request, understands the domain, validates the output, and connects the analy- analysis to a decision. Remember decision? We actually haven't talked about that. Let me know if I should cover that in a different video. This could be a product analyst who understands user behavior, a growth analyst who understands acquisitions and retention, or an analytics engineer who builds foundation AI needs Or a data analyst who actually is focused on defining business problem and decision-making. The common thread in all of these role is ownership, context, and judgment The role is Ownership, context, and judgment. So has AI replacing data analyst officially failed? I'm happy to share that it has failed to eliminate the role. It has succeeded at changing it, but it has not replaced it And this is true for majority of the tech roles. The pro- Coding only version of the job is, yes, under pressure, but the analyst who understands the business, use AI well, and knows when to answer is wrong, answer is wrong, is becoming more important. If you work in data analytics field, let me know what has been your experience. Do you agree with this? Do you think AI has failed to replace a data analyst, or we're not there yet? Let me know in comments. With that, I hope you're having a beautiful day. I'll see you in the next video. Have a good one. Bye.