Amazing article! How does one bridge the gap between someone who now uses AI for the hard part of technical work (writing scripts, running a correlation analysis) learn or read about to raise their bar to be a better analyst as well
Great question. I think the bar shifts from “can you do the technical work?” to “can you judge whether the answer is actually good?” That means deliberately building skills in problem framing, statistics, experimentation/causal inference, data quality, and business context. Also, get lots of reps challenging assumptions and explaining why a result does or doesn’t make sense. AI can accelerate the mechanics, but you still have to learn the judgment.
Makes sense! I think the biggest challenge I have faced challenging assumptions is also ensuring keeping personal bias while evaluating the results/data, sometimes it might feel counterintuitive to me but I have to remember the consumer we are serving is not equal to me.
It can be very disrupting to receive a report that eventually is wrong and needs to be improved by analytics team by yesterday :)
From my experience, it is not that easy for analytics teams to share only data that are forgiving in case of a wrong analysis, however the part with the managers makes perfect sense.
If an end user utilises data to create a misleading analysis it is their responsibility and their manager should make sure they understand it.
How can the analytics team properly communicate it to management though… It is usually more convenient for the data team to take the blame.
That’s exactly the tension! In the moment, it’s often easier for Analytics to step in, fix the work, and absorb the blame, but that can reinforce the wrong operating model.
The better approach is to make accountability clear upfront: if a team chooses to self-serve an analysis, they and their manager own the quality of that work.
At the same time, Analytics has a huge opportunity to add leverage by building the foundations that make self-service more reliable in the first place: trusted data, clear metric definitions, and a strong semantic layer that gives people consistent ways to ask and answer questions.
Analytics can enable the system without becoming accountable for every analysis produced from it.
Another awesome one! thank you!
Amazing article! How does one bridge the gap between someone who now uses AI for the hard part of technical work (writing scripts, running a correlation analysis) learn or read about to raise their bar to be a better analyst as well
Great question. I think the bar shifts from “can you do the technical work?” to “can you judge whether the answer is actually good?” That means deliberately building skills in problem framing, statistics, experimentation/causal inference, data quality, and business context. Also, get lots of reps challenging assumptions and explaining why a result does or doesn’t make sense. AI can accelerate the mechanics, but you still have to learn the judgment.
Makes sense! I think the biggest challenge I have faced challenging assumptions is also ensuring keeping personal bias while evaluating the results/data, sometimes it might feel counterintuitive to me but I have to remember the consumer we are serving is not equal to me.
Thank you for another interesting post.
It can be very disrupting to receive a report that eventually is wrong and needs to be improved by analytics team by yesterday :)
From my experience, it is not that easy for analytics teams to share only data that are forgiving in case of a wrong analysis, however the part with the managers makes perfect sense.
If an end user utilises data to create a misleading analysis it is their responsibility and their manager should make sure they understand it.
How can the analytics team properly communicate it to management though… It is usually more convenient for the data team to take the blame.
That’s exactly the tension! In the moment, it’s often easier for Analytics to step in, fix the work, and absorb the blame, but that can reinforce the wrong operating model.
The better approach is to make accountability clear upfront: if a team chooses to self-serve an analysis, they and their manager own the quality of that work.
At the same time, Analytics has a huge opportunity to add leverage by building the foundations that make self-service more reliable in the first place: trusted data, clear metric definitions, and a strong semantic layer that gives people consistent ways to ask and answer questions.
Analytics can enable the system without becoming accountable for every analysis produced from it.