The 90-Day Time Bar Kills Demurrage Claims. Hand the Laytime Statement to ChatGPT Work Instead

How much demurrage did you write off last year because nobody built the calculation in time? Not disputed, not negotiated down — simply never claimed, because the statement of facts sat in an operator's inbox until the time bar in the charter party ran out and the money stopped existing.

Every small bulk, breakbulk and coastwise operator knows the answer and almost none of them have written it down. It is not a small number, and it is the cleanest example in this trade of what changed in 2026.

Ninety days, and then the money is gone

Most charter parties give you a fixed window to present a demurrage claim with all supporting documents attached — commonly ninety days from completion of discharge, sometimes less. Miss it and the claim is time-barred. Not reduced. Gone. You can be owed thirty thousand dollars in laytime and recover nothing because the pumping log was never scanned.

The claim is not hard, it is just fiddly. You need the notice of readiness and proof of when and how it was tendered. You need the statement of facts, signed by the master, the agent and the terminal. You need the time sheet built hour by hour against the laytime terms you actually fixed — weather working days, Sundays and holidays excepted or included, whether time counts before the berth is available. You need the weather record for every hour you claim as an exception, the pumping or loading rate log, any letters of protest, and the laytime statement itself. Then a claim letter that cites the clause numbers.

For an operator with a handful of vessels, that packet is half a day of somebody's undivided attention. And it arrives at the exact moment nobody has half a day, because the vessel is already loading somewhere else.

How the packet gets built today

A post-fixture operator opens the fixture recap, opens the statement of facts as a scanned PDF that came off the agent's printer at an angle, and starts a spreadsheet. Times get typed by hand. The laytime clock starts, stops for rain, stops for shifting berth, restarts. Every exception has to be matched to a document, and the documents are spread across three email threads, a shared drive folder named after the voyage, and in at least one case a photograph of a signed page taken on a phone at the terminal.

The workaround everyone pretends is fine is triage: claim on the big voyages, let the small ones go. A goal-based agent is software you hand a finished outcome to rather than a list of steps — "build the laytime statement and claim pack for Voyage 24" — and it comes back with the actual spreadsheet, the letter and the merged exhibits, having worked through your files on its own.

What ChatGPT Work and Claude Cowork changed this year

This is the specific 2026 development. Claude Cowork arrived 12 January 2026 and ChatGPT Work on 9 July 2026, running on GPT-5.6. Both do the same new thing: you state a goal, they connect to your mail and your files, they break the job into steps themselves, they work for hours without being nudged, and they hand back a finished artifact. Not a chat reply. A completed spreadsheet, a document, a packet. Both were built for people who do not write code, which in this trade means the operations manager who has been doing laytime by hand since 2011.

The 2024 version of this idea could summarise your statement of facts if you pasted it in. It could not go and find it, it could not read the recap alongside it, and it certainly could not produce a time sheet that ties minute by minute. That is the difference.

flowchart TD
  A["Ops manager states the goal: laytime statement and claim pack, Voyage 24"] --> B["Pull the signed statement of facts from the vessel mailbox"]
  A --> C["Pull the fixture recap and laytime terms"]
  A --> D["Pull the pumping log, weather record and letters of protest"]
  B --> E["Build the time sheet hour by hour against the terms"]
  C --> E
  D --> E
  E --> F["Return laytime statement, claim letter and merged exhibit set"]
  F --> G["Post-fixture operator checks the notice of readiness and signs"]

Three streams open at once and land in one packet. That fan-out is why the job that took half a day comes back in an hour, and why it now happens on the small voyages too.

Tuesday morning: one goal, four sources, one packet

Discharge completed on the previous Thursday. On Tuesday the operations manager types one instruction: build the laytime statement and demurrage claim for Voyage 24, Houston to Antwerp, using the recap in the voyage folder and the statement of facts from the agent's email of the fourteenth; flag anything the documents do not support.

By late morning there is a spreadsheet with the laytime clock running line by line, each exception tagged to the document that justifies it. There is a draft claim letter quoting the demurrage rate and the relevant clauses. There is a single merged PDF of exhibits in the order a charterer's claims handler expects to see them. And — the part that earns its keep — there is a short list of gaps: the weather record for the second afternoon is missing, the statement of facts is unsigned by the terminal, and the notice of readiness was tendered outside office hours, which may be arguable under the fixture.

The operator then does the only work that ever needed a human: decides whether the notice of readiness was validly tendered, chases the two missing documents, and sends it.

What the owner has to change about assigning work

This is the part owners underestimate. You have spent years assigning tasks — "type up the time sheet", "scan the pumping log". Now you assign an outcome and an acceptance test, and you have to write that test down.

For a claim packet, the test is roughly: every hour claimed as an exception has a document behind it; the statement ties to the signed statement of facts with no invented times; the letter cites clause numbers from the actual fixture, not generic language; and anything uncertain is listed rather than smoothed over. Say that once, keep it in the voyage folder, and reuse it on every claim. The quality of what comes back is almost entirely a function of how precisely you wrote that paragraph.

The second change is scheduling. If the packet takes an hour instead of five, it can be built the day after discharge completes rather than the week before the bar. Make it a standing Tuesday job. The time bar problem is a calendar problem more than a workload problem.

The claim arithmetic

Assumptions, all illustrative and all worth replacing with your own: 34 voyages a year, 11 with demurrage exposure, 4 of those historically never claimed in time. Demurrage at $18,500 a day pro rata, average exposure 1.6 days, and a realistic recovery of 60 percent after the charterer's counter-arguments.

The hours saved are rounding. The recovered claims are the case. Prove it the cheap way: take three voyages from last year that you never claimed, rebuild the packets now as a dry run, and see whether the numbers you get back would have survived a charterer's review. You will know inside a week whether this is worth a subscription.

Where this still needs your post-fixture operator

Interpretation stays human, and it is most of the value in a contested claim. Whether a notice of readiness was validly tendered when the berth was occupied. Whether a swell that stopped work counts as a weather exception under the words you actually fixed. Whether the terminal's version of when hoses were disconnected beats the master's. These are arguments, not calculations, and they are settled by someone who knows the charterer, the trade and how hard to push.

Two other honest limits. A scanned statement of facts with a smudged signature and a handwritten correction in the margin will still be read wrong sometimes, so the operator checks the times against the original before anything goes out. And nothing here fixes a bad fixture: if the laytime terms you agreed are vague, a faster packet only tells you sooner that you are going to lose.

Frequently asked questions

Does this work if our statements of facts are scans and photographs?

Mostly yes, and better than it did a year ago, but treat every extracted time as a draft until a person has compared it to the page. The safe habit is to have the packet come back with each claimed time linked to the exact document and page, so checking is a two-minute scan rather than a re-keying job.

Is it safe to let software read our charter parties?

Reading is the low-risk half. Keep it read-only on the fixture folder, and keep sending under human control — the agent drafts the claim letter, a person sends it. The commercial exposure in this workflow is in what leaves the building, not in what gets read.

We only fix six or seven voyages a year. Is this worth it?

If even one of those voyages carries an unclaimed demurrage exposure, yes, and the maths is not close. The smaller the operation, the more likely the packet never gets built at all, because there is no post-fixture desk — just an operations manager with a berthing problem in front of them.

What about despatch owed the other way?

Same packet, opposite sign, and worth running for exactly the same reason. Charterers are considerably better at claiming despatch than small owners are at claiming demurrage, and a statement built the day after discharge is your best defence when their claim arrives.

One knock-on effect worth planning for: claims that actually go out generate replies, and replies come by phone from charterers' operators who want to argue a time. CallSphere builds AI voice and chat agents that answer the line when your operations manager is on a berthing call, take the vessel and voyage number, and book the callback in the right slot. It will not argue laytime with anybody — that is your job — but it stops a contested claim from going quiet because nobody picked up.