AI has become a buzzword, even as machine learning (ML) and data science push toward automation. We’ve all seen a million articles about how AI will revolutionize this or that, from agriculture to video games. The hype is enough to make anyone skeptical. But in terms of real impact, few areas have really maximized the potential of AI as have data science and ML.
The truth is that AI is not some quantum leap in computer intellect — it’s simply a tool. And like most tools, it’s only effective when it’s deployed and used properly. In ML, automation can be a touchy subject. On one hand, automated ML (autoML) tools are increasingly popular and commonplace. It would be tough to do a lot of the legwork required in building models, parsing millions of data points, and building databases without it.
On the other, data scientists are sometimes wary of using more autoML as it might seem either superfluous or simply not as good as them. Even worse, the endless conversation surrounding AI and autoML can sour even the most ardent automation advocates after a while.
Even so, there’s value in organizations embracing AI tools, especially those that help you automate key aspects of your organization’s funnel. However, for the skeptical data scientist, this may be a tough sell. The responses may range from doubts that a tool could actually be effective, to attempting to show why it wouldn’t, and simply refusing to automate parts of their job due to a fear that it might devalue their manual work.
For data leaders, then, it’s all about how you frame the AI revolution. Surprisingly, it’s not what you think it is.
Before we go further, it’s worth noting that AI is not going to put data scientists out of a job. While AI is many things, it’s not necessarily smart enough to outthink a highly trained data scientist, or even apply the creative thought required to solve problems in many cases.
Even so, there are reasons why data scientists might be hesitant, or simply unwilling to embrace AI. The resistance is not entirely unwarranted. After all, if you spend even five minutes looking up AI automation, you’ll find countless articles that talk about how soon, AI will do EVERYTHING from scraping data to building models to making us coffee while it does so. Such inflated hype would create backlash even in the best of cases from experts who’ve spent years honing their skills.
Moreover, it’s hard to see what simply adding an automation tool would do in real terms. This is where data leaders come into play. Instead of simply handing down an edict to make it work, it’s up to leaders to show the real value for data scientists in embracing the AI revolution and automation tools in general.
It’s also true that most data scientists aren’t diametrically opposed to automation tools — in this day and age, it would be like an engineer opposed to electricity. The problem is that, often, they may not really see the value in adding more tools to what’s already a pretty complex ecosystem. Why add more automation if I can just handle it myself — or so the thinking goes.
It’s not a matter of fear, but inconvenience. Adding more tools makes jobs more complex, and means having more moving pieces to manage when a data scientist can just do it on their own. Here’s where you can start changing the narrative:
A lot has been made of how Ai is democratizing technology and making it accessible to the masses, and in a way that’s true. However, it’s not black and white. Think about it this way — it’s actually a good thing you can democratize some of the more tedious tasks data scientists have to go through.
Your time is already limited by the complex tasks you’re actually specialized to handle, so why not delegate some of the less demanding tasks to citizen data scientists in your organization? The value of AI is that it makes this possible without requiring tons of startup on your data scientists’ part.
Sure, hunting for the right dataset and testing it to find that it works is its own kind of rush, but do you really have three weeks to find out? Also, you just spent three weeks vetting a single dataset, when you said you needed at least four — at this rate, you’ll be done by Christmas. Instead, you could simply connect to a data ecosystem featuring thousands of already cleaned, vetted, and collected data sources you simply need to click on to access. In the time you could manually find and test a single dataset, you can already have hundreds integrated with an AI-powered platform.
Sure, an AI-powered data science platform helps automate the often tedious stuff and remove the need for mundane tasks, but it can’t do the heavy lifting you still know how to do. Tools are only as effective as the person using them, so you shouldn’t think of a data science platform as replacing you, but simply making your job easier, and giving you better results. When it comes to the tougher tasks — choosing the right model, building nuance into your algorithms, understanding, and interpreting the results — AI will only go so far. Your AI tools still need you to make sense of them.
Perhaps most importantly, the AI revolution — and specifically automation — means that suddenly, you have more time on your hands. You don’t need to spend weeks hunting down the right dataset, or testing and retesting your models. You can skip a lot of the more repetitive, but still important tasks and simply automate them. Instead, you can start thinking creatively about the problems your organization has, and start finding new ways to tackle them.
Think about your unknown unknowns — the blind spots in your organization. When you’re stuck doing tedious tasks, you can’t really take a step back to consider them. Automation suddenly frees you up to start abstracting yourself from the trees to see the whole forest. This is likely the biggest benefit you might get. You can start thinking of new ways to drive value, improve your data science infrastructure, and solve new problems without having to worry about how long it will take you.
In the end, embracing automation comes down to how you present it to your teams. It’s not a matter of replacing them, devaluing their work, or overcomplicating their stack. It’s more about liberating them to do the work you hired them to do. The AI revolution is not about replacing data scientists, it’s about giving them the tools they need to take their work to the next level, unbound from the tedium of having to do menial tasks to get any results from their existing machine learning models. It’s about finding better data in seconds, and not days. But most importantly, it’s about building data science infrastructure that lets them think of more ways to do great work and improve your organization.