The Optimization Era: Engineering the AI "Ironman Suit" with Victor Fang, PhD

In the rapidly accelerating world of Agentic AI, most leaders are preoccupied with the engine, AKA the raw processing power of Large Language Models (LLMs). 

But according to Victor Fang, PhD in Computer Science and CEO of AnChain.AI, an engine without hands is just a stationary brain. Victor views "hands" as the proprietary data, specialized tools, and action-oriented capabilities that allow a general AI brain to execute meaningful tasks within a specific vertical. Without these hands, an AI remains a general processor; capable of reasoning but unable to interact with the unique data and context required to solve high-stakes problems

To survive the Optimization Era, organizations must move beyond general automation and begin building specific "Ironman suits" for their teams.

These are tools that augment human expertise with deep, vertical-specific context. Let’s dive into the biggest takeaways from our interview and see how you can apply this to your organization. 

Who is Victor?

Dr. Victor Fang is a pioneer in the AI space, having specialized in the field long before it became a household term. With a background building high-scale AI products for giants like Google and FireEye, he founded AnChain.AI in 2018 to apply those same principles to the world of blockchain.

What AnChain.AI does: The name implies its mission: analyzing blockchain with AI. AnChain.AI is a Silicon Valley startup that builds AI agents to fight fraud and streamline anti-money laundering (AML) compliance. Their technology is trusted by both the private sector and the US Government to explore sophisticated exploits and catch bad actors in the high-stakes world of cryptocurrency.

As a founder who spends his days caught between packed VC boardrooms and deep technical sessions with his engineers, Victor has seen firsthand how AI can either be a transformative signal or just another layer of noise.

Outcomes Over Algorithms: Why Goal-Oriented AI Wins

Many organizations misunderstand the fundamental nature of bringing AI into existing workflows. They treat it as a linear process, or a simple one-to-three-step automation. But Victor points out that real-world workflows, especially in security and compliance, are an entire graph of complex, non-linear interactions.

When companies jump straight into technology without a clear objective, they fail to see the results. Victor’s philosophy echoes a core principle of the Optimization Era: Start with a goal, not a technology.

"Start with the goal. What exactly are we going to create? What exact value are we going to bring to the customers? Technology comes as secondary." — Victor Fang, PhD

In Victor’s vertical, this means moving beyond general AI chatbots. While a model like ChatGPT is trained on the entire Internet, it lacks the specialized "hands" needed for complex tasks. 

For AI to be effective, it must have access to proprietary, vertical-specific data. In Victor’s case, this is the labels, threat intelligence, and sanction databases that turn a general processor into a specialized agent.

Breaking the Black Box: Why Context is the Real Moat

One of the most significant risks in the Optimization Era is "survival bias,” trusting the AI simply because it provides an answer. Victor warns that Large Language Models are designed to be helpful, even when they are wrong.

"The dangerous thing is they will spit out results... if you are not familiar with transformer technology, it may spit out a result that is totally wrong because the model does not know how to answer and will try to push something out." — Victor Fang, PhD

In the vertical of crypto-security, a hallucination isn't just a nuisance; it’s a security failure. To solve this, Victor’s team focuses on data integrity and auditability. By using Chain of Thought reasoning and strict system prompts, they ensure that the AI's logic is visible and grillable by human investigators.

Systematizing the Intelligence: From Observation to 1,000x Gains

Before you can automate, you must observe. Victor, an electrical engineer by training, views every workflow as a system of inputs and outputs. By breaking these systems down into key components (Intelligence, Investigation, and Reporting), his team can identify exactly where the friction lies.

• The Observation Phase: Observing how analysts use their current tools to identify invisible parts of the workflow that have become muscle memory.
• Systematizing the Steps: Breaking down complex graphs into measurable components to ensure that any automation built is actually improving the process.
• The Result: When these invisible steps are documented and automated correctly, Victor has seen improvements ranging from 6x to 1,000x in efficiency.

💡 Pro-tip: Leveraging Scribe to document these workflows is the first step toward building an AI agent's "hands.” By capturing how your threat hunters or compliance officers actually navigate their tools today, you provide the structured data and context needed to train a specialized AI agent.

The Augmented Professional: Using AI to Level Up

A major concern in the world of work is that AI will take over entire professions within months. While Victor acknowledges that 80% of easy tasks will likely be replaced, he offers a more empowering perspective for the individual contributor.

"Consider AI as an Ironman suit. It’s making you as a human stronger... don't let it control you. You always have to stay on top of it to figure out if it’s doing the right things." — Victor Fang, PhD

The future of work isn't about human versus machine; it’s about the human who uses AI versus the human who doesn't. Victor encourages everyone to level up by finding a problem they are passionate about and using AI to solve it.

Relational Resilience: Where AI Cannot Compete

While AI can optimize the what and the how, Victor believes it will never replace the human element of building relationships. AI can summarize an email or analyze code, but it cannot replace the true connection between a salesperson and a customer or the reading the room required for strategic leadership.

The Optimization Era isn't about moving faster; it’s about being better. If you want to move beyond the shiny project phase and see real ROI, your Monday morning move is clear:

1. Define the Goal: What specific vertical impact are you trying to make?
2. Understand the Limitations: Learn the technical boundaries, like context windows and hallucinations, to build effective guardrails.
3. Build the Ironman Suit: Don't just buy a general AI; provide it with the proprietary data and documented workflows it needs to act as a specialized agent.

When you understand the work at a systemic level, you don't just optimize a process, you empower your team to take down the giants (like Thanos). 🦸

Catch his full session below.