Chartering & post-fixture
For post-fixture operators, demurrage analysts and claims handlers who live in recaps, SOFs and laytime statements. We train the desk to push a model through those documents and leave a trail a colleague can check.
AI training for maritime desks · Singapore
Navigatewithai teaches one industry only. Every exercise uses maritime paperwork rather than generic business examples: statements of facts, recaps, noon reports and emissions returns. A cohort leaves with prompt patterns tied to its own document set, ready for Monday morning on the desk.
Founded in Singapore in 2026.
Generic AI courses spend their hours on summarising and e-mail drafting. A post-fixture desk needs something narrower: a model that can be pushed through a statement of facts, read against a charterparty laytime clause, and still produce a note a claims handler can check line by line. When the classroom example is a fictional marketing plan, the prompt patterns collapse on a poorly scanned SOF, a handwritten amendment or a BIMCO rider that qualifies notice of readiness. Maritime desks do not have spare hours to re-translate generic advice into their own forms, and they cannot afford an invented shifting time in a file that may later be read by lawyers.
Narrowing the audience is what makes the exercises usable on Monday morning. We write in the language of the desk — NOR, SOF, laytime, demurrage, CII, SIRE 2.0 — and we refuse to pad the syllabus with adjacent industries. The catalogue is therefore small on purpose: six programmes, forty-eight practice items in a maritime corpus, and a teaching habit of stopping the room to re-derive a figure by hand. Breadth is easy to advertise. Depth is what a claims file, a verifier pack or a PSC folder actually needs when the ship is already on the berth.
Each programme is built for a named shore-side function, not for “the business” in general. Titles, durations and typical attendees sit beside a short description; the full exercise lists live on the programmes page.
For post-fixture operators, demurrage analysts and claims handlers who live in recaps, SOFs and laytime statements. We train the desk to push a model through those documents and leave a trail a colleague can check.
Superintendents and fleet managers work from defect notes, planned maintenance systems, class status and inspection history. The programme turns that paperwork into prompts that produce work orders and inspection packs without inventing facts.
Bunker buyers and operators compare tenders, bunker delivery notes and laboratory results under time pressure. Exercises keep the arithmetic visible and force exception notes that a superintendent can sign.
For desks assembling MRV/DCS, CII, EU ETS and FuelEU Maritime files. The work is evidence trails and verifier-ready narratives, not slogans about environmental strategy.
HSEQ officers who must structure near-miss text, root-cause drafts and internal audit checklists. The course forbids the model from filling gaps in the facts or softening language that the SMS requires to stay precise.
Crewing coordinators handling circulars, certificate-expiry matrices and joining instructions. Emphasis on multilingual drafts that never expose personal data to a public tool.
A session is not a slide lecture. Two weeks before we meet, the desk sends a redacted sample of the paperwork it actually handles — an SOF, a recap, a noon-report extract or an inspection finding. We open with how these models behave, including the ways they fail on numbers, dates and scanned columns. The rest of the time is live work: each person at a laptop, the same documents on the table, and a verification drill before anyone treats an output as finished.
We collect a redacted sample of the desk’s real paperwork two weeks ahead, with counterparty names, rates and personal data already removed or replaced.
A short model-behaviour session, including where these tools fail on numbers and dates, and why a fluent paragraph is not the same as a correct one.
Three hours of live work on those documents, everyone at their own laptop, using the tools the company has already licensed.
Verification drills where every output must be traced back to a clause, a log entry or a reading before it is allowed to leave the room.
A model asked to compute laytime from a poorly scanned statement of facts will happily invent a shifting time that never appeared in the log. The illustration is not a war story from a named client; it is the failure mode we drill against. The exercise is to require a clause reference for every event the model proposes, and then to re-derive the totals by hand once, on paper, before anyone files the figure.
Hallucination in this setting is not a colourful paragraph. It is a wrong clock, a missing NOR condition, a CII comment that cites a year the noon reports do not support, or a SIRE 2.0 response that invents a corrective action the vessel has not taken. Two rules the course enforces without exception: nothing leaves the desk without a human signature, and every numeric output is re-derived from the source document at least once. Prompt craft without those habits is theatre. With them, a model becomes a drafting assistant that a claims handler, a superintendent or a verifier can actually use.
We work from client-supplied redacted samples or from our in-house synthetic corpus of forty-eight exercises built on public forms and invented voyages. Counterparty names, freight and hire rates, crew personal data and anything under legal privilege are stripped before a public tool is even discussed. Where the company holds an enterprise tenancy or an on-premise deployment, the session stays inside that boundary. Consumer chatbots are treated as hostile to commercial paperwork. Retention settings, paste discipline and the instruction to keep a human signature on anything leaving the office are part of the syllabus, not an afterthought. The full redaction rules sit in data handling on the method page.
Thirty minutes is enough to establish which tools are already licensed, how documents actually move across the desk, and the two or three tasks that consume the week. We ask about scanned PDFs, enterprise tenancies, flag or class constraints on data handling, and whether the cohort should sit in HarbourFront, on site, or on a remote evening slot for Europe-facing teams. The outcome is a written outline before anyone commits to a date — programme, format, seat count and the workflows we would actually run. Call +65 6371 5248 during Monday–Friday, 09:00–18:00 SGT, or send the form and we will reply within one working day.