At Computex 2025, NVIDIA unveiled a suite of technologies designed to accelerate AI and integrate it into the workplace. The new Enterprise AI Factory validated design and an expanded collection of AI Blueprints aim to assist companies in deploying AI-powered digital coworkers capable of reasoning, speaking, and adapting.

The company mentions that the concept of an AI teammate is evolving, and businesses will benefit from it. Whether assisting with fraud detection, customer support, or health education, AI agents are being trained not only on data but also on empathy, context, and memory.

For enterprises juggling generative AI, data integration, and infrastructure scale, NVIDIA’s new releases seem to pitch full-stack solutions. But this time, the focus is more specific: transforming AI agents into something meaningful.

NVIDIA also announced plans to invest in an AI supercomputer in Taiwan.

AI Teammates, Not Just Tools

The newly released Tokkio and AI-Q AI Blueprints are designed to help developers build intelligent digital avatars that converse naturally, integrate with enterprise data, and adapt to emotional or contextual shifts.

One example comes from the COACH store in Tokyo’s Harajuku district, where shoppers now chat with “imma,” a virtual stylist powered by Tokkio and NVIDIA ACE. Instead of playing pre-recorded scripts, it handles unscripted, real-time dialogue. For a brand built on personal style, the ability to personalise recommendations live is more than a gimmick. It sounds like a redefinition of customer service.

Meanwhile, financial institutions like the Royal Bank of Canada are deploying agents like “Jessica” — an internal-facing AI that helps employees handle fraud cases. Jessica retrieves the latest updates and documentation on fraud trends, enabling faster and more informed decisions.

At hospitals in Taipei (Taiwan) and Cincinnati (US), AI agents serve pediatric patients and visitors. From providing directions to explaining treatments in child-friendly ways, these avatars reduce cognitive overload, both for patients and overworked staff. This may not be surprising, considering many users go to ChatGPT as their doctor for many things.

These aren’t isolated experiments. It shows the initial forms of agentic AI, in which software is interactive, cooperative, and predictive.

A New Era for Video Analytics AI Agents

With video now accounting for over half of all global data traffic, yet less than 1% of it analysed, NVIDIA also launched a new blueprint to help organisations tap into this underused resource. The AI Blueprint for Video Search and Summarisation (VSS), built on NVIDIA’s Metropolis platform, enables enterprises to develop vision AI agents capable of understanding and summarising vast volumes of video content.

These agents, powered by vision-language models (VLMs) and large language models (LLMs), can perform real-time analysis, generate summaries, and even narrate complex events. From monitoring manufacturing floors to managing urban infrastructure, the VSS blueprint has already been used by companies like Pegatron and smart city operators in Kaohsiung, Taiwan. Results include a 67% reduction in defect rates and up to 80% faster incident response times.

With features like RAG, speech transcription, and GPU-accelerated processing, the VSS blueprint represents a leap in productivity for video-heavy industries. Even the NHL is getting in on the action—using AI agents to tag, summarise and retrieve game footage in real time.

An AI Factory That Scales with Enterprise Complexity

Building these agents is a full-stack endeavour behind the scenes. NVIDIA’s Enterprise AI Factory validated design offers a blueprint for deploying agentic AI on modern, accelerated infrastructure. It is tailored to run on RTX PRO 6000 Server Edition GPUs, NeMo Retriever, microservices, and partner integrations—essentially a plug-and-play back end for conversational, knowledge-aware agents. 

The AI Data Platform reference design, meanwhile, brings compute closer to storage. That’s critical for AI agents that must search millions of internal documents and respond in real time. Companies like IBM, VAST Data and NetApp are embedding GPUs and NeMo Retriever into their storage solutions, enabling AI to reason directly at the data layer.

This infrastructure is already being manufactured at scale. Taiwanese Original Design Manufacturers (ODMs) like Foxconn, Supermicro, ASUS, and Wistron are producing AI Data Platform-compatible systems that include the full NVIDIA stack — from NVIDIA’s BlueField DPUs to Spectrum-X Ethernet.

In short, NVIDIA’s ecosystem now spans blueprints, silicon, software, and deployment partners. With orchestration tools from Red Hat, Nutanix, and Canonical, as well as data tools from platforms like Galileo and DataRobot, enterprises can easily use AI agents with NVIDIA’s tech.

The post NVIDIA is Betting Big on AI Agents for Enterprises appeared first on Analytics India Magazine.