A series of applications are already being launched, utilizing AI from ChatGPT as a platform to respond to user questions or generate texts about specific domains. It is an interesting usage of the technology but with still limited impact. The game will change when AI starts to act as an agent with access to record transactions from the user interface in natural language. Join me in a little bit of dreaming about the future.
This is the first of a series of 3 articles where I intend to explore possible AI advancements in different tracks:
1. Natural language user interface
2. AI with access to business knowledge bases
3. Dynamic systems changing processes and rules
The goal of this series is not to present the current state of available technology but to design the next stages of evolution to guide research and development investments. It’s not about predicting the future but projecting it based on what we can see today.
The Evolution of User Interface
The way we interact with machines changes daily with technological advancements and the availability of increasingly refined products. However, some changes can be classified as major historical waves of transformation that revolutionized this interaction. Here are 3 of the most representative ones in computer usage:
- The creation of the Graphical User Interface (GUI) on personal computers in the 1980s.
- Hypertext navigation on the World Wide Web with the popularization of the Internet in the 1990s.
- Touchscreen interaction on mobile devices in the first decade of the 2000s.
The communication in natural language with AI models we are experiencing now represents a new revolution in the human-machine interface, possibly more impactful than the 3 previous ones. This revolution is just beginning.
Today: Chatbots using Natural Language
The new artificial intelligence based on language models, such as ChatGPT, Chat Bing, or Google Bard, are accessed in the form of a generic Chatbot. This means that the user must access a command box (prompt) and type an input instruction that will be processed by the AI to produce and return a response.
These AIs also have a certain ability to maintain the context of a conversation, so a response may not only consider the instructions from the last input but also the history of information exchanged in various past interactions. This makes their use much more conversational than simply “question and answer” like in traditional internet search engines like Google or Bing. It is possible to “iterate” in the message exchange process and work with the Chatbot as a partner in content creation, progressively refining the result.
Some applications are already available by training generic language models with an additional layer of proprietary knowledge about a specific domain, specialty, or business. Some examples of domains are the real estate market , management methodologies , and software development methodologies . With this additional and specialized layer in the domain, users can ask questions to a distinguished “virtual consultant.”
The next step in evolution is to transform what is currently a Chatbot application into an Agent to be used as a transactional interface by other applications and devices. Keep reading this article to understand.
Natural Language: The Pinnacle of the Interface
One of the main barriers to digital inclusion today is the human-computer interface. A user needs to learn how to use technology. It’s not enough to have expertise in the subject matter; the user must know the tool and understand its logic of menus, buttons, windows, shortcuts, and structure.
For example, a civil engineer who is very well-trained and experienced in creating plans using paper and ink is not immediately qualified to develop projects with a CAD Software tool. In reality, the complex interface of menus and commands has been a barrier for several older professionals to continue in the market.
A doctor, after years of training and practice in medicine and with a deep knowledge of the human body, still has to learn the commands and steps described in the manuals of diagnostic equipment and software to take advantage of technological advances.
An account manager at a bank needs to memorize a series of codes and be trained in dozens of applications to open accounts, negotiate investments, make loans, and attend to customer requests. And if they change jobs to work at another bank, they will have to learn everything again because the systems are different.
The Power of Natural Language UI
The availability of a natural language interface will allow the expansion of technology use to less digital-savvy users. The engineer will still need to know engineering, the doctor will need to know medicine, and the bank manager will need to know banking products, but they won’t need to waste time learning the specific commands, menus, and interfaces of a system.
The best way humanity has created to communicate is through natural language. Of course, human communication carries ambiguity and may lead to misunderstandings. There is a need for a shared vocabulary, the avoidance of environmental, physical, psychological, and semantic noises, and a shared channel for communication to succeed  as any other communication method. But it is still the easiest way for people to communicate with each other and even with themselves. Think quietly for a moment. Our thoughts are based on natural language.
What is on the horizon is systems capable of conducting transactions based on commands given in natural language, with an understanding of needs considering the context of user requests.
In addition to increasing productivity and efficiency, this type of interface will enable access for “digital illiterates” in the digital economy. It will only require mastery of the subject matter and the ability to communicate using words language to define objectives and indicate commands to be executed by an AI Agent capable of understanding what was requested within its context.
Tomorrow: AI Agents with Natural Language Interface
An Agent is an AI designed to think and act independently . Based on a goal provided by the user, the Agent follows an internal dialogue to establish the tasks it needs to perform and begins executing commands in interaction with the external environment, continually feeding back its understanding of the context to reassess and adjust the next steps until the goal is achieved.
In addition to the large language model (LLM) obtained through machine learning, the agent is capable of researching information online on the internet, interacting with other systems through command exchange via messages (API), and even exchanging messages with other users through emails or proprietary interfaces.
This is not conventional task automation because the tasks do not need to be predefined in a traditional programming model. The AI Agent can redefine and make necessary adjustments to achieve the user-defined goal autonomously and adapt during the process.
This changes the way we interact with systems. By conversing with your Agent as if it were a Personal Assistant, the user no longer needs to know the commands and interfaces of the systems that need to be accessed in a specific order to achieve their goal. All this complexity becomes transparent to the user and managed by the Agent.
AI Agents with Natural Language in Your Business
Think of an AI Agent based on a language model as a component to be integrated into your business and configured to provide access to all digital services your company offers to customers, employees, and partners. Just as we created an Internet channel (Intranet or Extranet) in the past to enhance our interfaces and optimize our businesses, now think that the new interface is in natural language, and start asking yourself the following questions:
- Which stakeholders are currently outside your business but could be integrated through this new natural language channel?
- What new services could be provided?
- Which services (internal and external) could be improved with the simplification of access complexity?
- Which processes can be fully (or partially) executed by AI?
- How can this represent a competitive advantage for your business?
These questions can indicate a path for your business to explore this new scenario and transform how technology is used in your organization.
Gaps: Challenges to Overcome
Most of the examples we have seen of AI usage in traditional businesses are still question-and-answer Chatbots, with disclaimers about the quality of the answers, which often are not of good quality. But the future is not far away. There are already platforms and solution providers that allow the creation of AI Agents that function as interfaces to perform business transactions in natural language .
But it’s worth noting that there is still much to be done before providing access to your company’s transactions to an AI. Several challenges need to be overcome with the inclusion of testing, experiments, and the creation of barriers to ensure that:
- The AI used has the ability to deeply understand the context and intention of the user and does not perform transactions in an erroneous manner.
- Ethical issues are adequately addressed to prevent an AI Agent from causing damage or harming people to achieve its goals.
- Information privacy is ensured, and access is restricted to defined authorities and profiles.
- The data used to train the AI is free of biases that could lead to prejudiced decisions or actions.
- The AI improves its ability to deal with ambiguities and complex questions without “delusions” or assuming information that could lead to errors.
- APIs are developed to enable access to your business transactions by agents with the necessary security.
I do not underestimate the difficulty of overcoming these challenges, but I believe they will be overcome soon. Be ready to leverage your business’s strategic vision using AI as soon as possible and take advantage of the opportunity of being a pioneer.
In this first article of the series, we explored significant advancements in the user interface through natural language, represented by the new language models. We have realized that we are experiencing a revolution in human-machine interaction, and technological evolution has allowed communication with machines to be increasingly natural and intuitive. Language model-based Chatbots enable more conversational interactions, surpassing traditional Internet search engines.
The potential of this natural language interface is remarkable, especially when it comes to making technology more accessible for less technical users. By eliminating the need to learn complex interfaces, AI Agents will enable professionals from various fields to focus on their expertise without wasting time on specific system training. This change has the power to promote digital inclusion and offer opportunities for “digital illiterates” in the digital economy.
In the future, Agents equipped with language models integrated into business applications will function as interfaces for conducting transactions, becoming the primary way of interacting with the services offered by companies.
This series of articles aims to precisely project these next stages of evolution, guiding research and development investments, and driving society towards an era of greater interaction and cooperation with technology.
Let’s be pioneers on this journey to build a future where artificial intelligence and natural language walk hand in hand in harmony, bringing innovation and prosperity to all.
Stay tuned for the upcoming articles and leave your comments to contribute to this discussion.
 Brazilian startup creates “ChatGPT for the real estate sector”; https://imoveis.estadao.com.br/noticias/startup-brasileira-cria-chatgpt-do-setor-imobiliario/
 Lais, a digital assistant focused on the real estate market; https://laisriopreto.com.br/
 PM Otto, a digital assistant for Project Management; https://www.pmotto.ai/
 Ágil GPT, a digital assistant for agile software development methodologies; https://argon.tec.br/mentoria-agil/
 Models of communication; https://en.wikipedia.org/wiki/Models_of_communication
 What is an AI agent?; https://zapier.com/blog/ai-agent/
 Build AI Agents at Scale – No Coding Required!; https://fine-tuner.ai/
 AI Agents: Power Up Your Productivity; https://aiagent.app/
Other articles and videos on Artificial Intelligence
- Business Analysis with Artificial Intelligence in Year 20XX
- The Big Debate: Risks and Benefits of Artificial Intelligence
- Few people understood the highest breakthrough of Chat GPT
- The Amazing ChatGPT Functional and Non-Functional Requirements
- Chat GPT knows how to differentiate rules from requirements. Do you?
- How Can ChatGPT Support Business Analysis?
- ChatGPT as a Tool for Business Analysis
- Will Artificial Intelligence Take Place?
- Interview with ChatGPT