Workflow library — 12 maritime tasks we train on

Laytime statement from an SOF

The task

On a typical post-fixture desk the statement of facts arrives as a scan, often with handwritten amendments in the margin. The operator must line those events against the laytime clause in the governing charterparty, convert local times, and produce a statement a claims colleague can check. Spreadsheets help with arithmetic, but they do not decide whether a shifting time counts, or whether NOR was validly tendered.

In the exercise

The model extracts events into a dated table. Attendees must cite the clause for each deduction, re-derive the total by hand, and reject any invented shifting time that does not appear on the SOF. A finished-looking statement with no source line is treated as a fail, not a draft.

Demurrage claim narrative with clause citations

The task

When laytime is exceeded the desk drafts a claim that must read as a reasoned argument, not a demand. Every delay event needs a clause citation, a time stamp from the SOF, and a reservation of rights if correspondence is still open. A narrative that merely restates the spreadsheet total will not survive a charterer’s reply, and invented facts are worse than a thin claim.

In the exercise

The model assembles a first draft from the SOF and recap. The drill is to keep only events that can be pointed to on the page, to insert clause numbers, and to strike any sentence the attendee cannot source before the text would leave the office.

Charterparty recap comparison

The task

A recap is compared with the governing form and with rider clauses that arrived later by e-mail. The desk must spot conflicts on laytime, NOR, bunkers on delivery, and law and arbitration, then produce a working note for operations. Missing a single rider can undo a week of fixture work. The comparison is slow because the documents are not aligned and the language is not uniform.

In the exercise

Two texts are placed side by side. The model lists differences; attendees verify each difference against the source pages and mark any clause the model invented or silently dropped. The working note may contain only verified items.

NOR validity checklist

The task

Notice of readiness is tendered against conditions that vary by form, port custom, and the recap. The operations desk must check whether the vessel was in free pratique, whether holds were accepted, and whether the notice was given in the required window. A defective NOR can collapse a demurrage position. The work is a checklist, not a paragraph of prose.

In the exercise

Attendees build a prompt that asks only the questions the clause actually poses. Every yes or no must point to a line in the SOF, a survey remark, or the recap. The model is not allowed to assume pratique or to fill a blank with port custom.

Noon-report anomaly review

The task

A fleet desk reads noon reports for speed, consumption, weather, and position. Anomalies hide in rounding, in copied-forward figures, and in weather that does not match the route. The task is to flag readings that need a master or superintendent to look again, without turning the review into an accusation. Volume is the problem: dozens of reports a week, each with the same columns.

In the exercise

A batch of synthetic noons is reviewed. The model highlights outliers; attendees must reconcile each flag with the previous noon, the voyage instructions, and bunker figures, and discard any flag that cannot be shown on the page.

PSC preparation by port and flag

The task

Before a call the technical or HSEQ desk compiles deficiency history, outstanding conditions of class, and the inspection focus of that port and flag. The pack must be usable on board, not a dump of last year’s findings. Time is short, and the same vessel may face different regimes in consecutive weeks. The work is assembly and ranking, not legal advice.

In the exercise

Attendees prompt for a jurisdiction-specific checklist from a redacted history. Every item must be traced to a recorded deficiency or a circular. The model may not invent a detention, a code, or a closed-out item that is still open.

SIRE 2.0 observation responses

The task

After a SIRE 2.0 inspection the office drafts responses to observations. Each reply must address the question asked, describe the corrective action without promising what has not been done, and stay inside the company’s SMS language. Over-claiming is a risk. The documents are long, and several people may comment before the pack is sent to the inspecting party.

In the exercise

A synthetic observation is answered with the model as a first pass. Attendees strike any fact not in the observation or the SMS extract, and they keep a human signature on the version that would leave the office. Tone is secondary to traceability.

Defect note to work order

The task

Superintendent defect notes arrive as e-mail, photographs, and half-sentences. The planned maintenance system wants a work order with a system code, a priority, and a spare-part hint. Translating one into the other is slow and easy to get wrong when the note mixes several defects. The desk must not invent a part number or a due date the note never stated.

In the exercise

Notes are turned into PMS-ready orders. The model proposes structure; attendees verify codes against the company’s list and refuse any spare that was not named in the note. Mixed defects are split before they are coded.

Bunker tender comparison

The task

Bunker buyers receive tenders in different units, with different delivery windows and quality caveats. The desk must put them on a common basis, including location, grade, and payment terms, then write a short note for operations. A cheap offer that cannot be stemmed on time is not cheap. Arithmetic errors in unit conversion are common and expensive if they pass unnoticed.

In the exercise

Tenders are normalised into one table. Attendees re-derive the comparison by hand, check the BDN-related caveats, and reject any figure the model cannot show in a tender. The operations note may quote only verified numbers.

MRV/DCS evidence assembly

The task

Emissions reporting needs an evidence trail that a verifier can follow: noon reports, bunker delivery notes, port logs, and the voyage list. The desk spends days reconciling figures that were never designed to add up. Gaps must be explained, not papered over. The output is a pack, not a single number, and EU ETS or FuelEU narratives sit on top of that pack rather than instead of it.

In the exercise

Attendees assemble a trail from a synthetic voyage. The model lists missing documents; every reconciling line must cite a source. Allowance narratives are written only from those sources, never from a remembered factor.

Near-miss text categorisation

The task

Near-miss reports arrive in free text, often written at sea under time pressure. HSEQ needs categories that can be analysed without inventing a cause. The original words must remain available. Over-tidying the text destroys the signal the report was meant to carry. The desk’s job is structure, not investigation, and not a rewritten story that reads more neatly than the event.

In the exercise

A batch of synthetic reports is tagged. Attendees check that categories match the SMS scheme, that no root cause was invented, and that the source sentence is still attached to each tag. Clean language that cannot be sourced is discarded.

Crew circular in three languages

The task

Crewing teams issue circulars on joining, certificates, and company notices. The English original must be rendered in two further languages without changing meaning, dates, or document names, and without exposing personal data. Machine translation that invents a medical instruction is not acceptable. The circular is a controlled document, not marketing copy, and names of forms must survive the rendering intact.

In the exercise

A redacted circular is translated. Attendees verify names of forms, dates, and prohibitions against the English source, and they remove any personal data the model echoed from a sample. Meaning is checked line by line, not by gist.