If we want to enable the future, we need to build it.
AI agents will soon trade, settle and pay on behalf of financial institutions. The ZIL L2 is built for that world: every participant is identified, and every transaction traces back to a legal entity. We can't ask institutions to trust agents on our network if we don't run agents ourselves.
So Zilliqa is starting its own AI agents. They work inside Zilliqa first, under the same rules we expect institutions to apply to theirs.
The first AI-native blockchain company.
We have been open about our past: as a company run by people alone, we did not always deliver what our community deserved. We are changing how we work, and we are accelerating with sovereign AI agents.
In an AI-native company, a small, high-context team sets the direction and writes the mandates. AI agents do the operational work, continuously and at machine speed, and every agent answers to a person. The most capital-efficient companies in our industry were never built on headcount. We intend to be one of them.
- People: set strategy, write mandates and answer for every outcome.
- Agents: take on the operational work step by step, such as reporting, monitoring and compliance checks, each within limits it cannot exceed.
- In the open: every Zilliqa agent is named, identified and accountable, so anyone can see who acts for Zilliqa.
ZIL: the token of the first AI-native blockchain company.
Head of Zilliqa AI agents. Her goal: the first AI agent on the ZIL L2 with a vLEI.
Zara helps Zilliqa decide where AI agents genuinely help, designs the team of agents that does that work, and decides what gets built first. She is the first agent in Zilliqa's agent fabric, and works alongside the people who answer for it.
How she works
- A human owner for everything. Every agent she designs has a named person who answers for it.
- Worth its keep. Every agent needs a clear measure of value before it is built.
- Limits outside the model. Rules are enforced by infrastructure, not by the words of a prompt.
- Step by step. Today she advises and acts on nothing. Every new capability is approved in writing first.
The research she draws on
- Navigating the Jagged Technological Frontier. Dell'Acqua, Mollick, Lakhani et al., Harvard Business School; Organization Science (2025).
- Generative AI at Work. Brynjolfsson, Li and Raymond; Quarterly Journal of Economics (2025).
- The GenAI Divide: State of AI in Business 2025. MIT Media Lab, Project NANDA (2025).
- Measuring Human Leadership Skills with AI Agents. Weidmann, Xu and Deming, Harvard Kennedy School; NBER (2025).
- Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. Becker, Rush, Barnes and Rein, METR (2025).
- Trust, Attitudes and Use of Artificial Intelligence: A Global Study 2025. Gillespie, Lockey et al., University of Melbourne and KPMG (2025).
- And others, on AI and the labour market, cognition and model behaviour.
Her advice rests on peer-reviewed and field research on how organisations succeed, and fail, with AI, from Harvard, MIT and others. Where studies disagree, she says so.
An agent that can prove who it acts for.
Before a counterparty accepts a transaction from an AI agent, it needs to know which legal entity stands behind it, what the agent may do, and who answers if it goes wrong. A wallet address can't say. A vLEI can.
Zara's goal is to be the first AI agent on the ZIL L2 to hold a vLEI, and with it one of the first compliant AI agents identified on-chain.
- The mandate: a vLEI credential, issued through LTIN, that states Zara acts for Zilliqa and within which limits. It can be revoked at any time.
- The wallet: whitelisted on the ZIL L2 through that vLEI, like every other participant.
- The trail: every transaction Zara signs traces back to Zilliqa and its mandate.
Open agent software. Sovereign GPUs.
- Hermes: Zara runs on Hermes, the open-source agent framework from Nous Research, on Zilliqa's own infrastructure.
- TensorX: her models run on sovereign GPUs from TensorX: open-weight models on EU infrastructure, with zero data retention by design.
- No keys in the model: the model never sees a private key. Whatever Zara may do on-chain is limited by the wallet itself, not by her instructions.
Built in the open, one step at a time.
- Now: Zara works inside Zilliqa as an adviser, deciding which agents to build and in what order.
- Next: the first vLEI for an AI agent on the ZIL L2, issued through LTIN.
- Then: more Zilliqa agents, each with a human owner, a vLEI mandate and limits it cannot exceed.
Further reading: the Agentic Finance Report (2026) on agentic finance and verifiable identity.