The opening line at this year's AI Agent Conference was perfect. As Simon Chen took the stage, he smiled and said: “Over 3,000 human agents are coming here by bus, by plane, by train, in the most traditional way possible. I'm no longer sure if this conference is about AI agents anymore.”
The room laughed. But that joke contained the entire signal: three thousand founders, engineers, investors, and enterprise operators converged on New York for two days to confront the most consequential shift in how companies are built since the internet. I was there — not just as an attendee, but as a founder who has built and exited tech companies, and as an LP investor in Firsthand VC, the fund that helped architect this conference from a small gathering into a movement of thousands.
What I witnessed was not a hype cycle. It was an industry crossing a threshold.
Three Themes That Cut Through the Noise
1. The Agentic Enterprise Is the Only Viable Posture
Speaker after speaker — from Datadog, Crew AI, Fetch.ai, IBM, and enterprise CTOs across financial services and government — converged on one message: workflow and organizational change is happening in months, not years. The AI agent era is not something you prepare for. It has arrived, and the enterprises winning are those already running supervised agentic workflows in production.
The conference was organized around three tracks: Agentic Enterprises, Agentic Engineering, and Agentic Industries. All three pointed to the same structural shift: AI agents are no longer the demo. They are the product. The Crew AI CEO shared that 42% of their own internal work is now AI-assisted. Not AI-augmented — AI-assisted. That distinction matters enormously for how you think about building a company in 2026.
2. The Real Barrier Is Adoption, Not Technology
This was the most counterintuitive insight of the conference, and I heard it from CTOs, investors, and operators alike: the hard problem of agentic AI is not the model. It is the organization.
The Pearson CTO put it cleanly: companies succeeding with agentic AI have clear scope, defined adoption paths, and organizational alignment before they build. Technical excellence without adoption strategy produces zero ROI. The C-suite panel cited that 85% of AI projects fail to demonstrate ROI — not because the technology underperformed, but because change management was absent.
The driving analogy resonated deeply: “Don't toss the keys to your 16-year-old.” Human-in-the-loop is not a limitation of current AI — it is the correct enterprise posture. The companies building governance scaffolding now, while autonomy is still limited, are the ones who will safely expand agent scope over the next 18 months.
3. Open Source Models Are Winning the Enterprise Back-End
The VC panel was explicit: closed models will become too expensive at true agent scale. When an AI agent runs for hours on a complex workflow, the token economics of frontier model APIs become a budget catastrophe. The analogy used was Linux vs. consumer OS: open source wins the infrastructure layer every time.
This is a massive signal for where the enterprise AI infrastructure investment opportunity sits. The companies building orchestration, observability, governance, and tooling layers on top of open source models — not the models themselves — are where durable enterprise value is being created.
What This Means for Asia
Asia's enterprise software stack is significantly less entrenched than the US or EU. That is not a weakness — it is the most important asymmetric advantage for builders in our region. Enterprises in Hong Kong, Singapore, and Greater China can adopt agentic workflows faster precisely because they are not carrying 20 years of legacy infrastructure debt.
At N+, our thesis has always been that Asia's next category-defining companies will be infrastructure, not applications. The AI Agent Conference confirmed what we already believed: the infrastructure layer underneath agentic enterprises — orchestration, data governance, observability, identity — is the generational investment opportunity. Not the agents themselves. The rails they run on.
“The first firm to rebuild this infrastructure layer at scale across Asia captures a generational opportunity.” — N+ Investment Thesis
A Personal Note on Simon and Firsthand VC
Simon Chen built this conference from nothing. A few hundred people in a room to over 3,000 practitioners in New York. That kind of community building reflects a genuine conviction that the agentic AI ecosystem needs a home — a place where founders, engineers, and investors can share real operational experience, not polished vendor narratives.
As an LP in Firsthand VC, I have had a front-row seat to how Simon thinks. He doesn't just write cheques. He builds the connective tissue around companies. Watching 3,000 people fill a conference he architected is one of the clearest signals I have seen that the right community infrastructure, built with the right conviction, compounds faster than almost any other investment.
Building the Agentic Infrastructure Layer in Asia?
N+ is co-building with founders who see what we see. No pitch deck required — just conviction and domain depth.
Start a conversation →