Generative modeling uses learned patterns to create new content, such as text, code, images, audio, or video. It is a subset of artificial intelligence, and its output is typically guided by a user’s prompt. Many generative systems use foundation models pretrained on large datasets; large language models are designed to interpret and generate text across many tasks without training for each specific problem. section 12.4.1 section 12.5
For analysts, a generative assistant can help produce an initial draft or respond to questions about a dataset. Analytics vendors are also adding assistants to visualization and reporting tools. These capabilities may speed up work, but they do not replace the judgment analysts contribute when deciding what to ask, how to assess an answer, and what stakeholders need to know. section 12.4.1 section 12.4
A practical way to use a generative assistant is to:
- State the task and available data in a prompt, then refine the instructions when the response is too broad or misses the intended question. section 12.4.1 section 12.5
- Ask for the steps, code, and appropriate performance metrics behind an analysis so its output can be compared with an existing approach. section 12.4.4
- Edit recommendations and presentation for stakeholders, and assess where the assistant succeeds or falls short of an analyst’s work. section 12.4.4
From a broad request to a useful metric
Asked simply to analyze a customer-login dataset, a chatbot described its columns and requested a more specific question. A follow-up asking for mean and median login days, plus mean login days by region, produced those figures. The change illustrates how a precise business question can make generated output more useful. section 12.4.1
Table 12.4 Updated prompt and LLM response using ChatGPT 4.0
|
Prompt
|
Response
|
|---|---|
|
Provide the mean and median for login_days, and mean login_days values by region.
|
Here are the statistics for login_days:
Mean Login Days (Overall): 15.73 days
Median Login Days (Overall): 16.0 days
Mean Login Days by Region:
Asia: 21.37 days
Europe: 14.49 days
North America: 17.53 days
South America: 9.51 days
These metrics highlight how often customers log in per month, both overall and within different regions. If there’s more you’d like to explore or another aspect of the data you’re interested in, just let me know!
|
The updated prompt and response show the requested overall and regional login metrics after the initial request proved too broad. section 12.4.1
Generated content also carries risks. Models can reproduce bias present in data and produce misinformation that appears truthful; their wider adoption raises employment and resource concerns. Treat plausible output as something to evaluate, not as a substitute for checking an analysis or exercising judgment about its use. section 12.4.2 section 12.4.4
