Feature
The Gist
- Key insights. AI data analysis tools like Bard and ChatGPT can streamline initial steps in data exploration.
- Generative push. Generative AI models have evolved to offer some level of programming-related assistance.
- Marketing advantage. AI in marketing analytics offers quick forecasting and data exploration, though caution is needed.
So many business tips regarding AI are aimed at content creation. But is Bard or ChatGPT also useful for deeper AI data analysis?
From an analytics perspective, it’s a complicated yes, made uncomplicated if you take the right steps.
Features like ChatGPT ADA are enhancing generative AI. They boost generative AI's ability to explore data accurately, despite the machine learning models deployed to generate prompt responses.
Still, you can use artificial intelligence to conduct a data guestimate — handy when you just want a working idea of what data insights are possible. AI tools provide relevant indicators suitable for brainstorming initial analysis steps and anticipating what to expect from known data details.
Let’s take a look at AI data analysis.
Where Would an AI Guestimate Be Helpful?
For example, suppose you have data on the number of website conversions per day for your website. You want to know if a trend of conversions is sustainable — in short, can you expect a similar number in the future. How likely is the possibility to have the same volume of conversion in the next week, month and so on.
To get an answer, you would need to treat the data as a time series, making a time series forecast appropriate.
Some analytics platforms offer forecasting features effective for short-term predictions. Google Analytics has a "Predictive Audiences" report that indicates which users are likely to convert within the next seven days. It also offers modeled conversions designed to fill in gaps in conversion data for a more comprehensive view of conversion behavior. Adobe Customer Journey Analytics provides a similar feature through its AI Assistant.
However, built-in features can only make predictions within the assumed context of the analytics platform. This approach limits the ability to accurately predict the sustainability of a data trend. The issue is exacerbated if there are too few data observations. Most built-in forecasts will not be reliable in such cases.
This is where a data model created using R Programming or Python comes into play. These programming languages have libraries that contain functions and allow for supporting parameters for further customization.
However, not everyone is familiar with programming syntax or advanced statistics. The absence of someone skilled for a quick analysis can impede data exploration.
Related Article: How AI and Data Analytics Drive Personalization Strategies
What Happens When Bard, ChatGPT, or Any AI Tool Analyzes?
So, how can AI data analysis benefit marketing analysts and managers?
Generative AI models are increasingly used for programming-related prompts, showcasing Python, R, and other programming languages in their response examples. Therefore, an AI tool can offer some initial data insights.
Typically, generative AI models perform text analysis and consolidate information — the models respond with phrases from their underlying dataset that are "likely" the correct prompt response. For example, if I prompt Propensity or Bard to "Add 2 + 2," the AI model "interprets" that 2 + 2 should equal 4, rather than specifying the number type for 2 + 2 and calculating a value of 4 (which, by the way, is what a given programming language would do). This approach limits precision, such as when calculating a vector of numbers with decimals. The one-shot example in my prompt engineering article demonstrates such accuracy limitations, even just two places past the decimal point.
However, a generative AI response can still suggest where an analysis could begin and what is feasible given the prompt and data details. To achieve this, we must craft prompt examples to train the generative AI model in interpreting and reasoning with the data, much like an analyst would typically do.
Related Article: Prompt Engineering and Its Vital Role in AI-Driven Marketing
How to Best Move Forward With an AI Data Analysis
To conduct a time series forecast, you would first supply the data you wish to forecast — for instance, the number of daily conversions in my example. This data should be in a tabular format, with each observation representing a value at a specific time point. When describing it, you should specify the column name and indicate the time series frequency.
You would then need to specify the forecasting method you want to use, such as simple average, moving average, ARIMA, or exponential smoothing. You can also state which programming language to use in your prompt. Additionally, you should define parameters or analysis conditions. For example, if you aim to forecast the number of daily conversions for the next week, you would instruct the model to forecast seven observations into the future.
Once provided with this information, the AI data analysis model can outline a forecast and present associated code for generating a forecast estimate.
Here's an example that generated an R programming script for creating a time series based on a sample of data. The data sample consists of random observations extrapolated from website values, intended to represent 30 days of data.
ChatGPT generated code that I could execute in RStudio, an IDE. While the code was functional, the syntax for the time series visualization did not display 30 days on the x-axis. This was somewhat acceptable, as analysts often adjust dates and time periods. However, this is something that should have been corrected in ChatGPT's response. One of the key advantages of AI is that it automatically sets up the basic parameters in any provided code.
I also had ChatGPT generate a visualization of the Autocorrelation Function (ACF), a visualization plot that displays autocorrelation to help analysts determine whether a time series is stationary. A stationary time series pattern indicates that the pattern is independent of time, making it suitable for advanced forecasting analysis.
ChatGPT also offered an explanation of the ACF and guidelines on how to generally interpret its details.
Bard was also capable of generating a code snippet and explaining the details of an ACF, but it did not produce an ACF chart. It did mention another statistical visualization, a PACF, which is also relevant for time series analysis.
You can apply this AI data analysis approach to visits, conversions, or any other daily event data. However, while AI provides distinct capabilities for quick exploratory data analysis, you still need a good understanding of your data to recognize if something is missing from the results. As illustrated by the example with the ACF, it's also important to remember that different AI tools may have varying responses to information that is considered universal.
Webinar
Sep
25
CMS Briefing: A Live Look at What’s Next in AI-Driven Platforms
Learn how leading organizations are using AI‑driven tools to publish faster, personalize smarter and stay secure.
Register
Webinar
On demand
Ready or Not: How Data-First Organizations Are Unlocking Agentforce Potential
Learn how to cut through the noise, activate Agentforce and build a Salesforce AI strategy that actually delivers.
Watch Now
Webinar
On demand
AI in Customer Service: Faster Resolutions, Happier Customers
Don’t let rising demand burn out your team. See how to build a smarter, more resilient support org.
Watch Now
Webinar
On demand
From Hype to High-Impact CX Strategies That Actually Scale
Turn buzzworthy AI and outsourcing trends into measurable CX wins with fresh 2025 data.
Watch Now
Webinar
On demand
Insights to Action Rethinking the Contact Center for Real Business Impact
Join our exclusive webinar to hear CX executives share their innovative strategies for transforming service delivery.
Watch Now
Webinar
On demand
Accelerating Healthcare Ops with AI
Sign up to see how your peers are putting AI to work — and making it stick.
Watch Now
Related Article: AI in Data Analysis and the Evolving Role of the CMO
AI Tools Alleviate Forecasting Woes — But Heed the Label Warnings First
It's crucial to view the output as not entirely scientific. Models provide consolidated information, so you'll need to verify the sources that influenced an answer and assess how repeatable the results are.
You can assess the forecast's performance on another dataset by asking follow-up prompts for comparison. This will give you a clearer understanding of how robust the forecast is.
You can also test the generated code in an Integrated Development Environment (IDE) to confirm its functionality and fine-tune the details. Your familiarity with the data is essential to ensure that the syntax is useful and will yield a reasonable estimate.
The programming should account for other statistical factors, such as the data's randomness and the likelihood of any underlying relationships within the data. You will likely still explore other forecasting methods to compare the accuracy of your results.
Also, keep the downstream impacts of your preliminary analysis in mind. For example, if you're examining conversion data to improve conversion rates, you must also consider other factors that influence conversions. This could involve scrutinizing your analytics reporting tactics, such as a checkout journey report in Google Analytics or an A/B test on a landing page.
Related Article: 5 AI Analytics Trends for CX Personalization
Generative AI Features Are Becoming More Insightful
In fact, you may want to stay updated on new features of your preferred AI platform. You can do this by checking the release notes, a page where platforms usually announce feature updates.
Just over a year ago, Bard and ChatGPT lacked critical features suitable for robust data analysis. Today, new features are being added weekly, quickly altering best practices for prompts and programming workflow. The recent addition of voice and image capabilities in ChatGPT could potentially expedite how analysts iterate on analysis examples.
The changes will also necessitate greater scrutiny when comparing AI platforms against each other. ChatGPT Plus, through its ADA plugin, holds a significant advantage over Bard and other AI tools. It allows file uploads up to 500 MB in size, while other platforms require users to paste data into a text window. This difference may seem inconsequential for small dataset trials but can quickly become cumbersome for data tables with a large number of columns or rows.
However, Bard also offers beneficial features, such as disclosing the online sources behind a prompt response — useful for discovering and documenting relevant industry articles and white papers. It also displays two drafts side by side, allowing you to select a preferred explanation. Google is utilizing this feature to gather feedback and enhance Bard's contributions.
Related Article: 3 Ways AI-Powered Predictive Analytics Are Transforming Ecommerce
Use AI Data Analysis to Explore With Care
Despite some limitations, an AI tool can still serve as a valuable forecast indicator for brainstorming initial analysis steps. This will help you better envision what to expect from your data exploration.
ChatGPT, Bard and other AI tools can quickly elevate the quality of logic in data and assumptions, but analysts and managers must identify the appropriate next steps when acquainting themselves with AI data analysis capabilities. While a GenAI model may have limitations in its findings, it can offer the right breadcrumbs to prevent analysts and managers from rapidly becoming lost in a vast "forest" of analytic options.
