Unlocking the True Power of ChatGPT: Beyond a Know-It-All Assistant

I bet many of us are still looking at ChatGPT as a know-it-all assistant that has knowledge about practically everything you can find on the web. Or in a business context, as a smart assistant that can retrieve information from various knowledge bases when combined with GPT and search.

However, is that its true power in a business context? Definitely not.

Whatever way you look at it, it becomes clear that every (!) knowledge process within organizations is up for revision - not next year, not tomorrow, but today.

GPT is not merely a passive assistant; it is an intelligent agent and a reasoning machine capable of comprehending human input and accomplishing complex tasks. Recent advancements, such as Agent GPT and Auto GPT, have revolutionized the potential of GPT by equipping it with tools (APIs) that enable task execution. As a result, GPT can not only understand instructions but also autonomously or semi-autonomously perform tasks on our behalf. This is no less than a paradigm shift. Whatever way you look at it, it becomes clear that every (!) knowledge process within organizations is up for revision - not next year, not tomorrow, but today.

ChatGPT as a Multifaceted Assistant

In simple terms, we can equip GPT with basically any type of API and teach it how to use those APIs. Consequently, once it understands what syntax to use to consult the APIs, it just needs human-like instructions and can perform the corresponding API calls autonomously. This opens up a whole new world of automation and smart decision-making. I will discuss two popular frameworks in this context.

In the framework of Agent GPT, we instruct GPT that it has several 'tools' at its disposal. These tools can be web searches, API calls, and others. We then instruct GPT that, once a human gives it a certain goal or multi-layered problem, it should strategize about a stepwise approach to accomplish said goal. In each step, it can use the tools on behalf of the human to perform the necessary activities. For example, you could ask it to buy a washing machine, and it will break down that goal into substeps such as searching and comparing washing machines on the web, making a final choice - possibly by prompting you for your preference - visiting a webshop, and putting the machine in a basket for you.

Auto-GPT arguably goes one step further. You can prompt it to create predefined workflows or processes that GPT can follow without constant human intervention. It is a fully automated agent that, once it receives an input or trigger, autonomously reasons, processes information, and executes the necessary actions or tasks. This automation saves time and streamlines operations by reducing the need for manual intervention at every step.

Examples

In the customer support domain, GPT connected to customer support APIs can generate appropriate responses to common inquiries and frequently asked questions. The agent can provide return policies and general customer service inquiries, as well as information about product availability and shipping details, by consuming real-time ERP APIs. For example, if a customer asks about the estimated delivery time for a specific product, GPT can automatically generate a pre-filled response for the customer support agent with the estimated delivery timeframe based on the customer's location and the current shipping status. Clearly, GPT could also respond automatically, but human oversight at key moments should remain an important cornerstone in all that we do with AI.

In a logistics setting, GPT connected to logistics APIs can start analyzing the incoming shipping requests. It reads and understands the details provided by the customers, such as the nature of the goods, the destination, and the desired delivery timeline. Based on this analysis, GPT can determine the most suitable transportation method for each request. It takes into account factors like the size and weight of the goods, the shipping regulations for the destination country, and the availability of different transportation options (e.g., air, sea, road). It can even trigger the shipping order – ideally after validation

Do you see it?

Admittedly, the logistics example is already quite far-fetched! But that is not the point. The point I would like to make is that the potential of GPT as an agent and reasoning machine goes far beyond its initial role as a know-it-all chatbot. It can understand complex tasks, strategize about tasks, and connect to relevant tools and APIs to execute them. It can perform those tasks on demand, with supervision, or even fully automated (although one should be careful with full automation).

Once you see this, there is only one correct conclusion: it's essential for any business and management to reevaluate every main knowledge process in the organization and identify opportunities for enhancement and automation. Not next year, not tomorrow, but today. By harnessing the intelligence and capabilities of GPT, organizations can streamline operations, improve decision-making, and drive growth.

This article is written with the help of our dear assistant ChatGPT.

A couple of relevant sources:

• https://bernardmarr.com/what-is-auto-gpt-a-next-level-ai-tool-surpassing-chatgpt/
• https://medium.com/codingthesmartway-com-blog/auto-gpt-agentgpt-the-groundbreaking-ai-innovations-you-need-to-know-cc7820463d4b
• An interesting TED talk by OpenAI co-founder where you can see such agents in action: The Inside Story of ChatGPT’s Astonishing Potential | Greg Brockman | TED