For me GenAI is good for compiling documents from a large datasets. But i still tell people that GenAI should act as a Copilot and not the Pilot itself, so as humans will still need to think out of the box while GenAI helps to summarize and tidy up loss end.
By leveraging high-quality internal data, we successfully cut early-stage ideation and prototyping times by nearly 40%, allowing the team to uncover unexpected creative angles and connect previously siloed data trends. However, this venture wasn't without its challenges; we encountered significant hurdles regarding legacy data cleanliness and the constant need for human-in-the-loop oversight to mitigate model hallucinations. Ultimately, the realized benefits far outweighed these obstacles, freeing our team to focus on high-value strategic problem-solving while reinforcing the absolute necessity of robust data governance.
I’ve seen some unexpected successes with Generative AI in my projects. It not only helps with compiling and analyzing data, but also goes further by suggesting ideas and perspectives I wouldn’t have thought of myself. In some cases, it actually changed how I looked at a problem or led to new, practical solutions. The biggest benefit is that GenAI doesn’t just process information — it helps generate new ideas and ways of thinking. Saving Changes...
Yes – I’ve had a really powerful (and slightly unexpected) success using Generative AI for data extraction and cleaning. In one project, I had to turn a large amount of raw, messy data from different files into a clean, copy‑paste‑ready Excel format. Doing this manually would have taken me several days. My goal was to extract only specific fields and put them into a consistent structure so I could analyze them further in Excel. The main challenge was getting the AI to understand exactly what I wanted. My early prompts were too vague, so the model kept:
Including extra fields I didn’t need
Changing the order of columns
Formatting things inconsistently
After a lot of trial and error, I changed my approach: I created a small “ideal” example table in Excel that showed exactly:
Which columns I wanted
The order of the columns
The formatting I needed
Then I uploaded this example to ChatGPT and wrote a very explicit prompt:
“Extract only the data fields that appear in this example and output them in exactly the same column order and format.”
And then it finally clicked – the model started producing clean, structured data that matched my example. I could process large amounts of raw data and turn them into Excel-ready tables in a few hours instead of a few days. How data and GenAI worked together:
The example Excel file acted almost like a “schema” or specification for the AI.
The raw files provided the content, and GenAI handled the pattern recognition, extraction, and transformation.
Benefits I realized:
Huge time savings (days of manual work reduced to a few hours).
Much less manual copy-pasting and cleaning, which also reduced errors.
I could focus more on analysis and decision-making instead of data wrangling.
This experience showed me that GenAI is extremely powerful for data preparation, but the quality of the output depends heavily on:
How clearly you define the desired structure (in my case, via the Excel example), and
How specific and constrained your prompt is.
Once I solved the prompt design problem, the productivity gain was dramatic. Saving Changes...
Anonymous
great content
Saving Changes...
Alexandra RIZZITELLIProgram operations manager| Medical researchMelbourne, Australia
I've had a great experience extracting data from large dataset. One example is extracting a list of clinical trials from the .gov website and asking AI to source each and every publication available that relate to each of those trial. Doing this manually would take me weeks or months! Still have to check the output, as some papers colated where not accurate = not about the clinical trial. Overall, a sure gain of time. Saving Changes...
We have achieved excellent results with several implemented AI models.
The most notable results involve reading the source code of older programs and generating flowcharts. This has facilitated understanding the functionality and business rules contained within the code.
We have also built agents using diverse knowledge sources and documentation to facilitate end-user support and incident management.
I have recently been using GenAI more on communicating task requirements and completion deliverables from higher external agencies to my team. It helps ensure clear understanding of purpose of task, how to get to the desired results, and maintaining alignment with initial request.
PMI Team Member
