Twice weekly, I share proven strategies to help data scientists get noticed, promoted, and valued. No theory — just practical steps to transform your technical expertise into business impact and the freedom to call your own shots.
Why AI and Excel Don't Mix
|
Although Python is my usual weapon of choice, I'm not averse to building the occasional model in Excel when the use case fits. Having originally trained as an actuary, there was a point in my life when I literally lived and breathed Excel. It still feels like second nature - which is probably why I've never felt tempted to use AI assistance when building Excel models, despite using it when writing code. However, after speaking recently to a fellow data professional about how much AI struggles with Excel, I decided to give it a go. The first thing I noticed was how fast it was. In less than five minutes, it was done. The second thing I realised was what a nightmare I had just created for myself. Hundreds of formulae, each hiding in its own cell, each a potential source of error, each requiring me to check it manually. As astrophysicist Professor Matthew Bailes recently observed in the Australian Financial Review: "The challenge now is to know that what AI produced is actually correct because you can't use AI to check it." This is the part of the AI productivity story that tends to get missed. When I use AI to write code, validation is manageable. Code is designed to be read like text and loops mean hundreds of repetitive calculations can be reduced to a single line. If the logic is right, it's right everywhere. Excel is an entirely different beast. Every formula lives in its own interconnected cell. There are no loops, no way to verify once and move on. You have to check every individual instance manually. The AI saved me five minutes of production and handed me an hour of tedious validation work. After five minutes, I lost the will to live. It was like delegating work to a keen but unreliable graduate. The question isn't whether they're fast, it's whether supervising them costs more than doing the work yourself. Sometimes the answer is no, and AI assistance is clearly the right call. But sometimes, as in the case of Excel, the intelligence required to verify the output exceeds the intelligence required to produce it. In those cases, the productivity gain is an illusion. Every time you reach for AI assistance, you're making a choice about where to allocate intelligence. It's worth making that choice deliberately. Talk again soon, Dr Genevieve Hayes |
Data Science Impact Algorithm
Twice weekly, I share proven strategies to help data scientists get noticed, promoted, and valued. No theory — just practical steps to transform your technical expertise into business impact and the freedom to call your own shots.
A data scientist friend of mine once took a role in a team that never stopped complaining about their difficulties in recruiting technical staff. She had solid data skills, but not the specific organisational knowledge the role needed. Yet, she was willing to learn, and management seemed willing to train her. So, she asked for training. And she asked again. Her boss promised to show her “as soon as an easy project came up.” The work was too important for her to make any mistakes. In the...
When the AI wave first hit, it was all about chatbots, and tech CEOs promised a future where humans and AI would work together to deliver better outcomes than humans could manage alone. Then, as AI progressively became better, businesses started to question whether they needed human workers at all, and the first wave of AI-driven redundancies hit. The tech CEOs suddenly started changing their tune and began warning of the impending white-collar job-pocalypse, especially following the advent...
Data science is obsessed with scalability. Can your model handle bigger data received at a higher velocity than ever before? So, it's ironic that some of the best things you can do for your data science career don't actually scale. Genuinely understanding the problems of a single stakeholder - doesn't scale. Building a reputation as an expert in your domain - doesn't scale. Delivering recommendations that actually get acted on - doesn't scale. AI has made it easier than ever to produce...