top of page

A look ahead: Getting productivity out of AI in 2024

We are quickly approaching the end of 2023 – a year that saw the tech landscape shift rapidly with the emergence of AI tools. There is a lot of excitement and trepidation about what these tools will eventually be able to do. However, looking ahead to 2024, I want to highlight some ways that you can start interacting with AI to become more sophisticated and more efficient in your data work. Advances in AI will make a meaningful impact in the coming year, and if your 2024 to-do list includes figuring out what that will mean for your business, I would love to be part of the discussion.

Here are my do’s and don’ts for using AI as we head into 2024. I am specifically focusing on Large Language Model (LLM) chatbots and the experiences I’ve had using ChatGPT (OpenAI) and BARD (Google AI) to assist in data analysis and technology deployment projects. Other use-cases and tools are outside of the scope of this article:

Don’t:  rely on AI to complete tasks or give you final answers to anything (yet)

Do: Use AI chatbots as a personal assistant to help you brainstorm a project, organize your thoughts, and to give you a general overview when you are learning something new.

Spin: Although tools like ChatGPT and BARD have made extraordinary advances in the past year, they are still not ready to take the reins from humans for most tasks. However, they can provide considerable assistance. Think of AI as a copilot and not a captain. You can get solid guidance and a great companion for brainstorming out of your AI tools as long as you are ultimately the one with the master plan.


Don’t: Use unedited output from AI chatbots.

Do: Use AI to generate first drafts. This applies to a number of different areas:

  • Presentations

  • Budgets

  • Web Pages

  • Contracts

  • Code

  • Statistical Models

Spin: This is where AI can provide a lot of efficiency gains. Let your AI tool produce the scaffolding and boiler-plate material for your project so you can dedicate a higher percentage of your time to the details and polish. Once you get used to working this way, you can save hours on the front end of your project and come away with a much more professional deliverable.


Don’t: Use AI chatbots as a direct source of factual information

Do: Use AI to point you to good sources of factual information that you can check out for yourself.

Spin: LLMs like do an amazing job at producing coherent, appropriately structured language. That is what they are built to do. But there is no requirement that the language it produces needs to be factually reliable (pretty much the same thing can be said for human language). We always need to verify the sources of our information, and it is no different with an LLM. What an LLM can do, much better than a human, is to point you to a bunch of possible sources of information that you can pursue on your own. So, once again, AI is great for helping you get started.

Don’t: Learn new skills by starting general tutorials from scratch and working start to finish (unless you are working toward a certification)

Do: Ask an AI chatbot to provide step by step instructions for completing the specific task you are trying to learn. Also – ask the chatbot for explanation and detail as you go if you don’t completely understand something.

Spin: I’ve been leaning on this more and more lately. If I’m using a modeling technique that I’m unfamiliar with, or creating a complex data pipeline in the cloud, I’ll ask ChatGPT or BARD to walk me through the steps needed to accomplish my goal. If I get stuck somewhere, I’ll ask for more details. If my code gives me an error, I’ll ask for guidance on how to resolve it. This isn’t altogether different from what you can accomplish by searching through online documentation of statistical packages and posting questions to forums like StackExchange, but it keeps the whole process in one place and it maintains a record of your chat as you work through the process so at any point, you can go back and reference what you’ve already done.

Additional Tips:

  • Ask AI for its last data update. This can be a mixed bag. ChatGPT will give you an exact date (as of this writing, ChatGPT 4 indicates its last data update was in April 2023) and will caution you that all of its information is based on what was available up until that time. BARD, on the other hand, will tell you that it doesn’t have a single, fixed date for its last update, and it is continuously learning and improving. Just be aware that, because training LLMs requires a lot of time and processing power, they can’t be continuously updated with new information – this has to happen at scheduled intervals. Currently, AI does not have the ability to perform an internet search in real-time to find and consolidate new information either. Therefore, it is essential to engage with AI for tasks that do not require real-time information.

  • Use AI for editing. I actually did exactly that on my first draft of this article. I’ll provide two positive and one negative take-away from this process:

    • Positive: I uploaded the draft to ChatGPT and asked it to check spelling and grammar. This request helped me catch some missing commas, found one or two places where my subject-verb agreement was off, and offered some suggestions for re-wording that were helpful.

    • Positive: ChatGPT suggested adding an infographic depicting AI chatbot use cases, indicating that this should be a visual representation of different scenarios where AI can assist in brainstorming, organizing projects, or drafting materials. I thought this was a helpful suggestion, so I asked ChatGPT to create an applicable infographic….

    • Negative: This is the infographic that ChatGPT created:


  • Use AI to generate synthetic data. This one is a bit of a special case, but I’ve found that at times, I could really use a synthetic dataset with specific properties in order to test out some assumptions I have about analysis. Alternatively, I may need some synthetic data to supplement material I’m teaching for a class. In general, I know how to do this with standard libraries in Python and R, but it takes a fraction of the time to just ask an AI tool to do this, using (very specific) instructions in plain language. ChatGPT will generate a downloadable dataset whereas BARD will generate code (Python / R) that can be copied and pasted to create the dataset that you request. Please be aware that these tools may not be able to generate datasets with complex specifications. For example, I generally have no problem if I state the necessary properties of my columns independently, but if I pair that with requirements around relationships between the variables represented in the columns, sometimes the request fails.



Where does that leave us? AI figures to be a dominant force in the technology world once again in 2024 and it will likely have evolved considerably in its complexity and applicability by this time next year. However, AI tools can prove to be very useful today as long as you have the right perspective when you use them. Chatbots such as ChatGPT and BARD can be very powerful general-purpose assistants. It is worth the time to familiarize yourself with them now if you think it is likely that you’ll gain even a little efficiency in the near-term. Remember, not only will this help you in your current tasks, but you’ll be ahead of the curve as these tools evolve over time.

I plan to continue posting on developments associated with AI tools in the coming year, as these technologies are instigating important shifts in the technology and data landscape.

All the best in 2024!!!

34 views0 comments


bottom of page