AI-Proof Your Data Science Skill Set by Embracing Four Timeless Concepts
With AI productivity tools like Microsoft Copilot, ChatGPT, and many others emerging, some technology professionals have drawn concerns around their skill sets becoming obsolete. Since AI is still in its infancy, it's impossible for anyone to predict exactly how the skill sets of data teams will evolve in the coming years. However, when you consider how Business Intelligence has transformed since the 1960s, you'll notice that there are timeless human elements that give analytics professionals strong job security for the foreseeable future. Let's dive into those below:
Domain Knowledge
When you think about it, Data Scientists have never been paid to code, build models, or create dashboards. They have been paid to solve problems. That isn't changing with AI, but the tools they are using to do it are.
A competitive edge for any candidate in a Data Science role is having the ability to clearly articulate a narrative to a business problem and offer a recommendation. For example, healthcare companies want candidates who can understand a hospital's relationship with insurance companies, what types of patient services are offered, and what their overall revenue cycle looks like. If patient visits decline, hospital managers will want to understand any emerging trends on why this is happening. You can't quite answer this question without some foundational knowledge of how the hospital system works.
Having hired for several Analytics roles, I would take a candidate that has strong domain knowledge but less of a development background over someone who is just a coding wiz. If a predictive model can't be explained easily, a simple spreadsheet with good business context will always be more impactful.
Understanding Source System Design
Most analytics teams are working out of OLAP (Online Analytics Processing) systems. These can include Cloud Data Warehouses like Amazon Redshift, Snowflake, Google BigQuery and many more. While these tools typically store all data points needed for an analysis, they are not where the data life cycle begins. Meaning that a data engineering team will take data from a source system and replicate it to the OLAP on a schedule.
Because of this inherent separation of systems, analytics teams should understand the processes around how data gets produced. This can give better context to an analysis, because it allows the analyst to intimately understand the reliability of certain data points. For example, an organization might use Salesforce as their CRM system, and replicate data daily to Amazon Redshift. If an analyst can understand how Salesforce opportunities are entered into the system, they will understand how to manipulate data to better answer a question.
Learning Technical Concepts, with Abstraction
When we look at how programming languages have evolved since the 1960's, every language that has succeeded the next has moved further away from machine code (0's and 1's) and closer to natural language. While we are at the very forefront of this, you can start to see that large language models are in some ways this next step in programming.
Now with the power of AI tools, we are beginning to abstract ourselves away from the technical aspects of the role. And to a certain extent, that's ok. Professionals should be using these tools in any way they can to bridge a technical gap (as long as it helps grow their understanding of the topic).
But just because a task can be outsourced doesn't mean that you don't need to know about what's going on behind the scenes. Similar to how we use calculators for convenience when solving large computations, it doesn't mean that we can forget how division or multiplication works. Analytics teams should be encouraged to outsource code with tools like Chat GPT when necessary to save time, but must be able to explain what the output means.
With modern AI tools, it may be tempting for analysts to brush over this skill set. But to be successful, every candidate must still be doing the following:
- Continuing to broaden your skill set on concepts like Statistics, Data Mining, and Data Visualization
- Refining how to communicate & present results of an analysis in the context of the business. Not in the context of what is technically going on behind the scenes.
Networking
Unfortunately, like any maturing job market, the data industry as a whole has seen a lot of saturation in the last 10 years. This means that there are more candidates with the same skill set and competing for the same number of jobs. It's becoming rarer to see data jobs on LinkedIn that don't already say "Over 100 applicants". It's also unrealistic to think that an 8 month Data Science bootcamp will land you a six-figure job with no experience anymore.
This means that data professionals must have a more creative approach when it comes to looking for work opportunities. Networking might look like:
- Attending Conferences in your area, and maybe getting your current employer to help sponsor you to go
- Attending industry happy hours if there are any in your area
- Building up your LinkedIn profile, and even posting about related data science concepts to gather attention
- Making an ongoing effort to maintain meaningful connections
Conclusion
As technology continues to progress, so must we as data professionals. However, the concepts that have proven to be timeless are also the ones that are most impactful in a hiring process. I encourage Data Scientists to be actively using LLMs like ChatGPT, with the caveat that it is used as a utility to outsource singularly focused tasks where you can understand the output. When backed by the proper technical and domain skills, AI tools supercharge productivity and help candidates stay competitive. However, they cant be a substitute for any of your core skills as a Data Scientist. With this rapid expansion of AI technology, data professionals should not fear change but embrace it, because the industry will continue to produce new lucrative job opportunities even as the tool stack changes.
