Introducing AX Pro — Shifting the Focus from Replacing Humans to Empowering Experts

Human-Centric AI - Empowering Experts, Not Replacing Them

The 90% Problem: Why Even the Smartest AI Chatbots Fail at the Final Hurdle, and How We Finally Crossed the Line.

Have you ever tried implementing an AI chatbot for your work? At first, you're amazed at how it answers questions just like a human. But you’ve probably also experienced moments where the chatbot gives an absurd answer to a certain question or mumbles with an unsure tone. It’s that moment when it answers intelligently nine out of ten times, but that one mistake feels critical.

This is what is commonly called the "90% Problem." It's the phenomenon where an AI gets things right about 90% of the time, but the remaining 10% gap causes a sharp drop in trust and usability. No matter how high-performing a chatbot is, it can't handle every situation, leading to disappointing answers. When that happens, users quickly lose faith and look for other methods. Especially in business environments where accuracy and consistency are crucial, 90% accuracy is simply not enough.

The Limitations of AI: Lack of Trust, No Control, Difficult Updates

The problems with AI chatbot reliability and user experience run deeper than you might think. Modern chatbots generate incredibly natural sentences, but they sometimes confidently invent things they don't actually know. This is the phenomenon where they state false information with a very assured tone. From a user's perspective, it's difficult to distinguish which answers are right and wrong, so once distrust sets in, a feeling of unease remains even when the chatbot gives an accurate answer. Ultimately, for important questions, users end up double-checking with a human, cutting the effectiveness of the chatbot implementation in half.

Another limitation is the difficulty of control and updates. A general AI chatbot is trained on vast internet data, so it doesn't immediately know your company's latest policies or specialized knowledge. Teaching it new information requires time and money for separate training. It's also not easy to change the response style or tone to fit your corporate culture. Testing time is also enormous. Because of this, even after adopting an AI chatbot, companies often struggle in the field because it's hard to fine-tune it to their specific needs. If it can't respond to customers or employees with a consistent voice, the user experience inevitably suffers.

In short, current AI chatbots are smart, but they fail to close that final 10% reliability gap, making them insufficient to be fully trusted with real-world business operations.

TecAce's Shift in Perspective: From Replacement to Augmentation

To solve these problems, we applied TecAce's unique philosophy: "Expert augmentation, not human replacement." Simply put, the philosophy is that AI shouldn't completely replace people, but rather augment the capabilities of experts, allowing them to be active in more places, at all times. It's the belief that by combining the knowledge and judgment of human experts with the speed and scale of AI, we can leverage the strengths of both to increase reliability.

Based on this philosophy, TecAce has launched a solution called ‘AX Pro’.

What is AX Pro?: The Philosophy in the Name

AX Pro stands for an AI solution that combines Avatar, AI, and Agent. We also added the concept of Adaptive to encompass the flexibility to learn and evolve according to the situation. Finally, the ‘Pro’ emphasizes enterprise-level professionalism and scalability.

As its name suggests, AX Pro is an enterprise virtual human agent that bundles AI intelligence, Avatar's human-like interaction, and an Agent's ability to perform tasks. It's like an independent digital employee that responds to customers or staff 24/7, but behind the scenes, it has a unique structure where a real-world expert directs and controls the AI's direction.

In other words, the biggest differentiator of AX Pro is that it operates under the control of an expert through an expert-centric control tower. This keeps the AI from making autonomous judgments and "running wild", ensuring it is always under human expert supervision.

How AX Pro Works: Control and Participation by Field Experts

So, how does AX Pro implement this philosophy? First, a domain expert within the company (e.g., an HR team leader, a sales manager) accesses the AX Pro dashboard. From this Control Dashboard, the expert teaches and manages the necessary knowledge for the AI. They build a knowledge base of company policies, FAQs, product information, and frequently asked questions, and even set the AI avatar's speaking style and personality to match their own style and tone.

What's interesting is that you can even customize the AI avatar's appearance to look like the actual person in charge. By uploading a photo or avatar image and adjusting the voice tone and speech patterns, the responding AI can be made to feel like a friendly colleague, an extension of your own team.

This prepared AI agent is then deployed to various business channels. For example, it can be placed in an internal messenger (like Teams) as an HR support bot, instantly answering employee questions about company vacation days or benefits. At the same time, this AI can be connected to website or mobile app chat support, or even offline store kiosks, all with the same level of knowledge. No matter where the question comes from, AX Pro provides a consistent and accurate answer. For instance, if a customer at a store asks about a new phone's specs or inventory, the sales representative's AX Pro provides a seamless guide, reflecting even the very latest information.

The important part is that all of these conversation logs are recorded and monitored in real-time on the admin dashboard.

< A screenshot of a chat on Microsoft Teams>

< A screenshot of a Dash Panel>

What if the AI fails to answer a question perfectly? The AX Pro administrator can review that conversation log without missing it. They can then easily leave feedback, such as, "It would have been better to answer this part differently." For example, if an employee asked about a new benefits policy and the AI's answer was insufficient, the manager can add the correct information.

This feedback is immediately reflected in the AI's next response. AX Pro learns the model answer written by the administrator and evolves to answer the same question correctly next time. Because human expert knowledge and judgment are continuously injected, the quality of AX Pro's responses becomes more refined over time, resulting in a structure that is far more reliable than operating an AI independently.

AX Pro also provides a dual safety net through active monitoring. While the AI is responding, a quality-checking radar runs in the background via TecAce AI Supervision. It scores and evaluates multiple items in real-time, such as the response's accuracy, professionalism, speed, friendliness, and policy compliance.

If the AI says something incorrect or provides an unsatisfactory response to a customer, the system automatically detects it and sends an alert to the administrator. For instance, it might send a signal like, "Current response may be factually incorrect" or "May contain sensitive information." The administrator can immediately see the dashboard alert, intervene to correct the case, or take swift action, such as temporarily suspending the automated response to that query if necessary. Thanks to this automatic monitoring function, even if there's a mistake that a person misses, it can be addressed before it becomes a major problem.

The AX Pro Control Dashboard literally serves as an integrated control center for the AI agents. Through a KPI board in an easy-to-read radar chart format (Powered by Supervision), you can see how well your AI avatars are responding and where their weaknesses lie. Operational metrics like the number of conversations, aolved inquiry rates, and average response times are also updated in real-time, allowing you to view them just like a call center's real-time monitoring. Thanks to this centralized control, a company can coordinate dozens of AI employees working in various places as if they are moving as one brain.

The Look of a Future-Ready Organization: A Company Where AI Avatars Work Alongside You

When we turn our eyes to the future vision, the picture AX Pro paints is even more exciting.

In the near future, we may see AI avatars equipped with specialized knowledge posted in various departments within a company. TecAce is already showcasing the HR AX Pro for HR/General Affairs and the Sales AX Pro for on-site sales. This is said to be infinitely expandable into areas like finance (Finance AX Pro) and IT support (IT AX Pro).

This means that multiple virtual employees will be active within one company, helping human employees in their different professional fields. And all these AI employees will be connected to a central expert dashboard, collaborating under the direction of real managers. The company's collective intelligence is accumulated and shared within the AX Pro platform, creating a massive, living knowledge network where any member can instantly get the knowledge they need.

The emergence of AX Pro shows a paradigm shift in AI utilization. In the past, there was a primary concern that adopting AI would diminish the human role, but we are thinking in the direction of AI amplifying the human role. By adding AI's learning ability and scale on top of an expert's experience and insight, we can achieve a level of reliability and performance that is difficult to reach with either one alone. In effect, it’s about pulling AI's completeness, which was stuck at 90%, up to 99% through collaboration with humans.

Ultimately, what's important isn't the technology, but the operating philosophy. There are new possibilities that can only be obtained when AI is viewed as a human-centric partner. TecAce's AX Pro is a case of this operating philosophy being brought to reality, showing what new value human-centric AI can create.

Now, the AI chatbot is approaching us not as a simple experiment, but as a future-oriented colleague that grows together with experts.

https://www.tecace.com/ai-supervision