Compliance desks can now act on what an AI sees before a human confirms it saw anything at all.
The Summary
- Blockworks launched Agentic Detection, letting exchanges receive compliance alerts directly from AI agents without waiting for analyst review, cutting response time from 10 minutes to 5 seconds.
- Co-founder Jason Yanowitz told Unchained that manual analyst checks averaged 8 minutes per development, the new system delivers alerts in 5 seconds.
- Every alert carries a tag identifying whether an agent or analyst produced it, so compliance teams know what they're acting on.
- This is the first production case of crypto institutions taking action on AI-generated intelligence before human verification.
The Signal
Blockworks just shipped what every compliance desk has wanted but been too risk-averse to build: alerts that arrive before a human can type "checking now" in Slack. The feature, called Agentic Detection, went live Thursday in Messari Monitoring, the product Blockworks acquired when it bought Messari's intel arm. Exchanges can now configure their systems to ingest alerts the moment an AI agent flags a development, no analyst review required.
The speed improvement is brutal. Manual checks took roughly 8 minutes per event, agent alerts land in 5 seconds. That's not iteration, that's a different game. In crypto markets, where a smart contract exploit can drain $50M in the time it takes to grab coffee, 8 minutes is archeology.
"Manual analyst checks took roughly 8 minutes per development, while the new agent alerts arrive in 5 seconds."
Here's the design choice that matters: every alert carries a tag saying whether an agent or analyst produced it. Blockworks isn't trying to hide the AI. They're labeling it explicitly, letting compliance teams decide their own risk tolerance. Want human-verified alerts only? Configure it that way. Trust the agent enough to act immediately on certain alert types? That's an option now too.
This works because Blockworks already had the agent infrastructure running. They've been using AI to scan on-chain activity, social signals, and protocol changes. The agents were producing intelligence, analysts were checking it, then humans were sending alerts. Agentic Detection just removes the human bottleneck for teams that choose speed over certainty. The system enhances regulatory efficiency and adaptability by letting institutions tune their own speed-versus-verification tradeoff.
Key capabilities unlocked:
- Compliance desks can halt trading on a flagged token in seconds, not minutes
- Listings teams can pull a token from consideration before it's officially announced
- Risk managers can act on emerging patterns before they crystallize into losses
The broader implication: this is infrastructure for a market where agents are participants, not just tools. If your exchange is still waiting for an analyst to read an alert, write a summary, and ping the compliance channel, you're operating on human time in an agent-speed market. Blockworks is betting institutional crypto is ready to trust tagged AI output enough to act on it, especially when the alternative is being eight minutes behind.
The Implication
Watch how exchanges configure this. The ones that default to agent-only alerts for high-confidence event types (like smart contract deployments or large wallet movements) will move faster. The ones that require human review for everything will move accurately. Neither is wrong, but the speed gap will show up in how fast they respond to the next Mango Markets exploit or Terra collapse. Compliance is becoming a latency game.
If you run a listings desk or risk function, the question isn't whether to trust AI agents. It's what categories of intelligence you trust them with, and how you build review loops for the edge cases. Blockworks just made that choice explicit. Everyone else in crypto intelligence will have to ship their version or explain why they're still the slow option.