Turning Individual AI Wins into Scalable Business Advantage with HKSTP, FabriXAI, and beNovelty

As AI adoption accelerates across Hong Kong, many professionals and organizations are discovering the same challenge: while it is easy to achieve individual AI wins, turning those wins into scalable, enterprise-ready solutions is far more complex.

On 16 December 2025, beNovelty was honored to partner with Hong Kong Science and Technology Parks Corporation (HKSTP) to deliver a hands-on ecosystem event titled: “Automate & Accelerate AI Agents on HPC: Level Up Your Skills, From Zero to Master AI Agents.”

Hosted at Hong Kong Science Park, the event brought together founders, technologists, and professionals for an afternoon of learning, discussion, and hands-on API-powered AI building. The focus was on how AI Agents, APIs, workflows, and HKSTP’s high-performance computing (HPC) come together to enable real-world AI automation.

Keynote: Turning Individual AI Wins into Scalable Business Advantage

In the keynote session, Ms. Kyla Chan, AI Project Manager at beNovelty, explored a reality many teams face today:

AI adoption is happening everywhere, but AI is not scaling.

While individuals can quickly boost productivity using AI tools, organizations often struggle with fragmented adoption, duplicated effort, rising costs, and a lack of governance. Kyla’s keynote reframed the problem and the solution. Instead of treating AI as isolated tools, she emphasized building AI as shared infrastructure, grounded in four core principles.

1. API-first AI Architecture

APIs are the foundation of scalable AI. Without APIs, AI remains limited to conversations. With APIs, AI Agents can take action, fetching data, updating systems, and triggering workflows.

2. Shared AI Components

Scalable AI requires reusable agents, prompts, and API actions that teams can build once and use across the organization.

3. Governance by Design

Token control, access management, and usage visibility are essential to keep AI secure, compliant, and cost-predictable.

4. Agentic AI with Guardrails

Agentic AI enables autonomous, multi-step execution. It works best when paired with workflows, monitoring, error handling, and versioning.

The key takeaway was clear. AI does not scale until it becomes shared, governed, and operationalized through APIs and workflows. That means successful prompts, agents, and actions must be packaged as reusable building blocks, with clear access control, usage visibility, and cost guardrails. With this foundation in place, teams can reliably plug AI into real processes and scale adoption across departments without creating tool sprawl. Rather than treating AI as a collection of isolated experiments, the session emphasized building AI as reusable infrastructure from the start, so today’s wins become a durable platform for long-term business impact.

Panel Discussion: Rewrite Your Future, Winning Tomorrow with AI

The keynote was followed by a panel discussion titled “Rewrite Your Future: Winning Tomorrow with AI.” Speakers and guests explored how LLMs and AI Agents are changing what it means to work, learn, and build in an AI-first economy. The conversation covered shifts in professional skill sets, the growing importance of AI literacy and lifelong learning, and the new opportunities for entrepreneurship as automation becomes more accessible and faster to deploy.

A key theme throughout the discussion was that meaningful adoption is not just about individuals using AI tools to work faster. The real advantage comes when teams redesign how work gets done, turning AI into a shared capability across functions. When AI Agents are connected to APIs and embedded into workflows, supported by governance and reliable infrastructure, they move beyond being helpful assistants and become operational collaborators that can execute tasks consistently and safely at scale.

Hands-on Workshop: From Zero to Master AI Agents on HPC

The event concluded with an interactive hands-on workshop collaborated by beNovelty and HKSTP, allowing participants to experience how AI Agents move from concept to production.

The live demonstrations showcased the following capabilities:

1. Deploying AI on HKSTP HPC

HKSTP demonstrated how enterprise-grade AI models hosted on HKSTP’s HPC infrastructure can power production-ready AI applications, ensuring performance, security, and trusted data boundaries.

2. Building an AI Agent MVP with FabriXAI Studio

beNovelty showcased FabriXAI Studio, a low-code AI agent development platform that enables teams to:

• Build AI agents using reusable templates
• Select AI models hosted on HKSTP HPC
• Publish AI agents as Chatbots, Web Apps, or OpenAPI services
• Manage tokens, costs, and model versions via a built-in AI Gateway

3. Integrating AI Agents into Workflows via OpenAPI and n8n

Participants experienced how FabriXAI-generated OpenAPIs can be integrated into n8n workflows, allowing AI agents to:

• Classify and prioritize unstructured customer inquiries
• Convert free-text inputs into structured outputs
• Automatically insert results into enterprise data tables
• Keep data within HKSTP HPC’s trusted infrastructure boundary

This end-to-end demonstration highlighted how AI moves from experimentation to operational, governed, and scalable deployment.

Building Hong Kong’s AI and API Ecosystem Together

beNovelty is proud to collaborate with HKSTP in advancing Hong Kong’s AI and API ecosystem. As AI becomes core digital infrastructure, success will depend not only on better models, but also on APIs, governance, workflows, and operational excellence.

Events like this help bridge the gap between experimentation and production. They empower professionals to move from zero to mastering AI Agents, and they enable organizations to translate AI potential into sustainable business advantage.

We look forward to continuing this journey with HKSTP and the broader innovation community. Our sincere thanks to HKSTP for the invitation, and for creating a space where builders, technologists, and innovators can learn, exchange ideas, and build together. We are grateful to be part of this ecosystem effort, and we look forward to supporting more AI Agent adoption with HKSTP and the wider community.