AI that works inside your record, not beside it
It reads the market, prices the voyage, predicts where the fixture clears, watches every live voyage and assembles the claim. It never forms a contract, never releases against a bill of lading and never settles above a limit. Those four limits are not configurable.
Built for dry bulk, not adapted to it
It knows that quantity carries a tolerance and an option holder, that stowage factor decides intake rather than deadweight, and that a base form resolves to a known laytime starting position.
- Understands charter forms, clauses and rider sets
- Reads a recap as structured terms, not free text
- Prices emissions per leg, with the right scope factor
It works inside your record
The agents act on the same objects the desk acts on. An estimate, a claim pack or a chain comparison is produced from your own voyages rather than described back to you from somewhere else.
- No separate tool to copy data into
- Context aware of the voyage, cargo or claim in front of you
- Operates inside each user's existing permissions
It takes work off the desk
The capture, the lookup, the comparison, the chasing and the assembly all happen before a person opens the record. What is left is the judgement, which is the part worth paying a trader for.
- Work arrives done, ranked by money at risk
- Every action records what it measured against
- Any capability can be switched off without waiting for us
Six things it does every day
Reads the market
Several hundred cargo and tonnage mails a day become structured orders and open positions, deduplicated across brokers and resolved to real vessels, counterparties and ports. Low confidence fields are flagged rather than guessed.
Prices the voyage
A complete estimate the moment a match appears: route and distance, observed consumption for that vessel, port time from what that terminal achieves, bunkers by grade, emissions per leg, commissions and freight tax.
Predicts where it clears
An opening band, an expected clearing level and a confidence interval from weighted historical comparables, plus the likely next counter and the economic walk-away point for this counterparty.
Checks both sides of the chain
In-charter against out-charter, term by term, before the recap is confirmed. Laytime basis, demurrage rate and bar, performance warranty, off hire, trading limits, emissions allocation, bunker terms, notice periods.
Watches every live voyage
Arrival as a confidence range rather than a single time. Hours gained against the commercial baseline, predicted before the vessel arrives. Stationary time after an early arrival, logged separately.
Builds the claim
Laytime interpreted from the clause in this charter and applied to sourced events, with the working shown line by line. The document pack derived from what the clause requires, and gaps chased while they can still be closed.
Automation in a business that signs contracts
The reputational cost of one bad automated action is greater than the efficiency gained across a hundred good ones. So every capability declares what it may do before it does anything.
Acts
Parses the inbox, builds the estimate, watches live voyages, predicts arrival, accrues gained hours, escalates a bar date.
Bounded. Every action records what it measured against, so the figure can be re-derived rather than trusted.
Drafts
Writes the offer, the laytime calculation, the disbursement query, the owner follow up.
A named person sends it. Always.
Never
Forms a contract, signs or releases against a bill of lading, moves money above a limit, dispositions a sanctions hit.
Not configurable. The agent prepares the work completely and a person performs the act.
Every capability, its level and its gate
| Agent | Level | Gate and reversal | Release |
|---|---|---|---|
| Circular parser | Acts | Creates proposals a trader works. Corrections train it. | V1 |
| Estimate builder | Acts | No gate needed. An estimate creates no obligation. | V1 |
| Offer drafter | Drafts | The trader sends. The draft is refused if it concedes outside the approved ladder. | V1 |
| Chain mismatch detector | Suggests | Raises a task owned by the trader, with the exposure quantified. | V1 |
| Voyage exception monitor | Acts | Suppression is aggressive and the dismissal rate is itself monitored. | V1 |
| Statement of facts reconciler | Acts | Raises discrepancies while the record can still be corrected. A data gap is reported as a gap. | V1 |
| Laytime calculator | Drafts | Reviewed before service. Clause ambiguity is surfaced rather than resolved by preference. | V1 |
| Time bar sentinel | Acts | Alarms only. An underivable bar date escalates as unquantified rather than defaulting. | V1 |
| Early arrival watch | Acts | Every hour booked records the baseline version it measured against. | V1 |
| Disbursement scrutineer | Drafts | Drafts the query, the operator sends. Second approval above a value threshold. | V1 |
| Quote guidance | Suggests | Advisory, comparables always shown. Behind the readiness gate. | V2 |
Autonomy is earned, per desk
A capability enters at its stated level and is raised only on evidence, from two numbers tracked from day one: the share of proposals accepted without material edit, and the share of executed actions later reversed. There is no automatic promotion, and any capability can be switched off by your desk without waiting for anybody.
Open where the market actually clears
Twenty five years of your own fixture history is the one asset no vendor can sell and no competitor can copy.
Comparable band, live context, current recommendation
- An opening range, an expected clearing level and a confidence interval, with every historical case listed and inspectable.
- Fixture probability across rate, terms, timing and counterparty, shown as a curve rather than a single opaque score.
- Comparables weighted by genuine similarity of market regime, season and route, rather than all old data treated equally.
- The likely next counter and the economic walk-away point, derived from how this counterparty has actually behaved.
A recommendation that cannot show its comparables will not survive contact with a trader who has been doing this for twenty years. The comparables stay on screen.
This layer is gated, and we say so before you buy
Quote guidance ships in the second release, behind a readiness test run against your extracted record in week two. Nothing in the first release depends on it. Promising prediction from a history nobody has profiled is the fastest way to lose a desk’s trust in everything else the system says.
Coverage
Enough populated fixtures per route and vessel class to support comparison, rather than simply a large row count.
Outcomes
The losing side of the negotiation is recoverable, at least for recent years. A history of wins alone cannot teach a model where the market clears.
Comparability
A defensible mapping from historical records to today's comparison basis, with the breaks documented.
Time safety
Backtesting that cannot see the future. Otherwise every model looks excellent and none of them work.
The operator predictions ship in the first release.
Arrival distribution, early arrival forecast, the time recovery call and schedule risk all run on position, weather and the voyage baseline. The first release already carries every one of those inputs, so these predictions arrive with it rather than after it.
See it run on one of your own voyages
The fastest way to judge this is to point it at a cargo you already fixed and a claim you already served, and compare what it produces against what happened.