(To start the demo, click on the snapshot. Your feedback is very welcome!🙂)
To the many subscribers who provided feedback via DM, thank you for your contribution. This article is a consequence of your feedback.
The main feedback was that article #98 was a bit technical and difficult for non-IT professionals to understand. You suggested showing an example of how GPT can deliver complex responses by building agents and making these agents work together under a central entity acting as a project manager.
We are getting complex to follow. Well, let’s give you an example.
( I published this one on Sunday, so many of you can watch it with a cup of coffee. It’s about 10 minutes long. You will not regret it…)
Most probably, all readers have a basic to advanced understanding of Excel. You can filter data, create pivot tables, or generate graphs. To achieve this, you likely invested some learning time, ranging from a few minutes to several hours.
Using ChatGPT as an Analytic Tool
What if I told you, or better yet, showed you how to do the same in a conversational mode using ChatGPT?
In the video, you will see how to create a ChatGPT application, including loading an Excel file, creating prompts (in my case, text prompts, but voice commands are also possible), and getting tables, pivot tables, and graphs related to the Excel file.
The file I loaded is the 2023 Fortune 500 Company list, which is already available. From that, I will ask GPT to look at the data and create tables and graphs using natural language. Chat GPT will create new “entities” (a team of other resources capable of analyzing the data and displaying the information as requested).
Is this the end of Excel?
Not quite. The current stage of ChatGPT cannot handle the large amounts of data that Excel or formal analytical tools can, especially on an enterprise level. However, it clearly shows the potential that is just around the corner.
Imagine loading multiple files, connecting those files using some index, and making live queries to support decision-making processes.
What’s the value of this use case?
Can you imagine the simplification for an individual contributor getting insight into complex data with little or no code?
Picture this to fulfill your personal needs: financial, travel planning, playing with your hobbies, music, and movie catalog, looking for car parts to do your maintenance, navigating blueprints, and selecting parts using voice commands, to name a few.
Can you imagine using this as an Enterprise-level analytic solution in which any business executive can get accurate information without coding and at their request?
Did you know that the US Air Force announced that it digitized all aircraft technical manuals and enabled an AI-based solution to assist technical support teams WW?
Now, all maintenance units (WW) navigate any manual using AI. This can scale to global scenarios and is in compliance with DoD cybersecurity controls.
- Hmm, I got you, Jose. Is that around the corner?
- Maybe, maybe not. But we can start dreaming about it.
My taking on this is that many AI thought leaders are talking about this AI-agent approach, which provides an exponential benefit to any use case. So, tuning the coding and use cases is a matter of time and effort. At the same time, it is an essential message for individual contributors and enterprises, as many jobs and businesses will be redefined due to the new leveraging of digital transformation with this solution approach.
I’d love to get your comments or hear from you via DM if you want to have a deeper conversation on potential use cases.
- Good enough?
One more comment: As we discussed in article #98, we typically use foundational models (ChatGPT, Claude, Gemini) to analyze large volumes of text and get responses with evidence. OpenAI and Google recently announced new capabilities to respond to user prompts in multiple information channels (text, voice, images) similarly. However, as of today, most of these interactions are based on input, processing, and output, the so-called Level 1 AI agents.
The demo I’m showing today uses ChatGPT as a Project Manager (an AI-Agent Level 2). This Project Manager analyzes the request and creates other AI agents (Level 1 agents) with specific roles to work together and deliver a complex response. Instead of relying on a single GPT, you’ll see a team of GPTs managed by a primary interface with project management capabilities.
There is a programming approach to building these solutions on an enterprise level called prompt nesting. In this approach, you can create multiple agents as stations, with the primary agent orchestrating and interacting with all of them. Think of it like a Formula One pit stop, where the team leader, positioned at the front jack, oversees all the pit crew members. This leader ensures that each team member completes their task efficiently and only releases the front jack when everything is in order, allowing the car to race off seamlessly.
In this context, the front jack station is a project manager who monitors all pit members and ensures quality assurance. This demo showcases how AI can coordinate complex tasks through effective management and collaboration.
Maybe some of you would like to revisit Article #98. Here is the Link.
Thank you for reading this episode of the Digital Acceleration series.
To start the demo, click on the snapshot.
Your feedback is invaluable. Please “Like” this episode if you found it insightful, and share it with others who may benefit from it. As always, feel free to DM me or comment below. Get ready, and let’s make it happen!
Share this post