Agents don't do tasks here. They run lifecycles.
A lifecycle is the unit of work agents change the economics of: multi-party, regulated, structured, exception-heavy, gated and audit-bound. Six industries have them. OpenEXA's first is live.
Six traits. All six, or it isn't ours.
These are the properties that make a process too costly for humans at low margins — and exactly right for a swarm of narrow, verifiable agents with a shared ledger. Every one of them maps to a layer of the stack.
Multi-party orchestration
Brokers, custodians, transfer agents, clearing houses, shippers, regulators — each with its own state and timing. Agents hold the whole sequence at once.
Regulatory oversight
Rules that must be checked on every action — not sampled after the fact. Encoded as policy the Council enforces on every proposal.
Structured data exchange
Baskets, confirmations, documents, meters and filings in fixed formats. Machine-readable by design; machine-actionable through the MCP server.
Exception management
Partial fills, rejects, breaks, discrepancies and late data — the cases that consume human hours. Narrow agents resolve them; the Council escalates the rest.
Settlement or approvals
Value moves only when a defined gate is passed and someone — or something accountable — has signed. Approve or reject, automatic or human.
Audit-trail requirements
Every transition must be reconstructible, attributable and tamper-evident. Append-only and hash-chained, replayable by an auditor or a regulator.
One infrastructure. Many lifecycles.
OpenEXA is built as a general layer for agentic lifecycles. Financial markets are first because the data is cleanest, the rails are electronic and the proof is measurable in basis points. Everything below the domain agents carries over to the next class.
- LC-01
Financial markets · insurance Live
What agents run: ETF creation and redemption on live capital today; basis, collateral, margin and insurance settlement lifecycles on the same rails next. Parties: brokers, authorized participants, custodians, clearing houses, exchanges.
- LC-02
International trade · letters of credit
What agents run: issuance, document presentation and checking, discrepancy handling, bank-to-bank confirmation and payment. Parties: issuing and advising banks, exporters, shippers, insurers, customs.
- LC-03
Asset securitization · MBS / ABS
What agents run: pool assembly, eligibility and tranche compliance, servicer reporting, waterfall calculation and investor distributions. Parties: originators, servicers, trustees, rating agencies, investors.
- LC-04
Wholesale energy settlement
What agents run: scheduling, metering reconciliation, imbalance charges and multi-party netting. Parties: system operators, generators, traders, utilities, regulators.
- LC-05
Semiconductor manufacturing
What agents run: order-to-wafer orchestration across fabs and packaging houses, yield exceptions, quality-gated release, export-control checks. Parties: fabless designers, foundries, OSATs, logistics, regulators.
- LC-06
Pharmaceutical · aerospace
What agents run: regulated qualification, batch release, traceability and multi-supplier approval chains. Parties: manufacturers, suppliers, regulators, operators, auditors.
To make it concrete: a foundational infrastructure layer for the ETF lifecycle.
Trillions sit in 17,000 ETFs. Every one of them is created and redeemed through a mechanism that keeps price near NAV — and every gap in that mechanism is work a swarm can do that a desk cannot afford to.
Creation. When an ETF trades at a premium to its net asset value, an authorized participant delivers the underlying basket and receives new ETF shares at NAV — then sells them at the premium.
Redemption. When it trades at a discount, the participant buys shares at the discount and redeems them for the basket at NAV.
The gap is the work. It is small, persistent and recurring across thousands of funds — worthwhile only if the cost of operating the lifecycle is a fraction of what it is today. That is what the swarm changes.
Signal
Data agents run real-time analysis of price, NAV, basket composition and venue depth across the ETF universe.
Predict
Proprietary AI-ML models score the persistence of each gap and the risk of moving against it.
Decide
Strategy agents propose create or redeem, size and timing under portfolio-level constraints.
Approve
The Execution Council approves or rejects — automatic, with approval, or manual. Rejected proposals are logged too.
Execute
Execution agents route across clearing-and-settlement, OTC and dark-pool rails, HFT automation and dealer-brokers.
Settle
Clearing and settlement tracked to completion, with counterparty exposure engineered out of the design.
Reconcile
Every fill matched to broker truth; every transition hash-chained into the ledger.
One session, simulated. Move the sliders.
The gap between an ETF's price and its NAV wanders through a trading day. Whenever it runs past the threshold, an agent acts — create above NAV, redeem below — and pulls it back. More agents on the fund close gaps sooner; a lower threshold works smaller ones. Illustrative, not market data.
Dislocations appear every day. The swarm follows them — then leaves finance.
ETF NAV gaps, cross-venue spreads, futures basis and pairs dislocations — each a lifecycle with the same six traits. So are letters of credit, securitizations, energy settlement and batch release. The swarm follows the pattern, not the industry.
Have a lifecycle with the six traits?
Talk to us about running it on OpenEXA — as an institution with a lifecycle, a manager deploying Lifecycle 01, or a partner building on the infrastructure.