Frequently asked questions

Sail deploys AI agents that behave like mini-employees in your workflow. They monitor inboxes, extract shipment files, classify HS codes with reasoning, cross-check data across documents, and assemble declaration drafts around the clock. For ambiguous or high-risk cases, they stop and escalate to your team for final judgment.

Yes. Sail continuously improves using your historical classification data, accepted rulings, and corrected decisions. The more you use it, the more accurate and aligned to your specific trade profile it becomes.

Sail integrates via API with major ERP and TMS platforms. We also support email-based ingestion and direct file uploads, so you can start using Sail without changing your existing processes.

Sail is available internationally and without restrictions. The customs brokerage solution is currently available in the UAE and Switzerland.

Your team does. Sail automates the high-volume work and attaches a confidence score and rationale to every suggestion, and your experts approve, override, or correct the result. Low-confidence classifications and unresolved discrepancies route to a review queue instead of passing through silently, and every decision lands in the audit trail.

Technical configuration typically takes about a month, up to six weeks for more complex setups. In practice the schedule is usually driven by internal IT and security approvals on the client side rather than by the configuration itself, so we recommend starting that review early and we support it with the documentation your security team will ask for.

You can still start. We offer an email-in, report-out workflow where your team sends documents to a dedicated address and receives validated results back, with nobody logging into anything new. The platform's ISO 27001 certification also tends to shorten the security screening once the full review does happen.

Pricing follows your monthly declaration volume across five tiers, from Essential for up to 49 declarations a month to Enterprise for 1,000 and more. Every tier includes the full platform, so you're choosing capacity, not a feature list. If a volume tier doesn't fit how you work, we also offer per-document pricing.

All of them. Sail supports any format of documents, including images, PDFs, Excel spreadsheets, and CSVs with undefined structure. Documents in any language are processed automatically.

No. Sail classifies and extracts documents without templates or layout training, including layouts it has never seen before. You can start by forwarding documents from your existing inbox, and the system begins working on day one. Accuracy then keeps improving as your team's corrections and decisions feed back into your model.

No. Sail comes with hundreds of extraction fields configured by default, covering the data that trade documents actually carry: parties, line items, weights, values, terms, and references. If your business needs something specific, we add custom fields on top, and they behave exactly like the built-in ones.

No. Sail ships with hundreds of default validations, so totals, weights, quantities, party data, and cross-document consistency are checked from day one and you're up and running immediately. From there you can tune tolerances and add your own rules per client, country, and trade lane whenever you need more control.

Sail splits and extracts what it can read with confidence, and when a value is genuinely illegible it flags the field for review rather than guessing. Nothing gets fabricated to fill a gap, so a bad scan produces an open question for your team, never silent bad data.

Matching strictness is configurable. Where letter-perfect agreement matters, as in letter of credit checks, you can keep exact matching on; everywhere else fuzzy matching absorbs special characters, abbreviations, and spelling variants. We can also match names against a curated master list so known parties always resolve to the same record.

Yes. Product codes, material codes, and SKUs can be matched against your master data as they're extracted, which also resolves cases where several candidate codes appear on the same document. Values that don't match anything are flagged for review rather than passed through.

Yes. Units, country codes, ports, currencies, and reference numbers are all configurable, so if your systems expect KG rather than KGM, or you want your own reference numbers used as job identifiers, the output follows your conventions. Export files are shaped to the format your downstream systems import, with no reformatting step in between.

It handles the repetitive execution work between an email arriving and a declaration being ready: monitoring shared inboxes, extracting and sorting shipment files, consolidating line items, cross-checking documents against each other, proposing HS codes, and assembling declaration drafts. Teams typically save 30 to 60 minutes per shipment and cut overall processing time by 50 to 70%.

The agent watches your shared inboxes around the clock. When a new shipment arrives, it pulls the attachments, classifies each document, and routes everything to the right job before anyone touches a keyboard. Jobs can auto-assign by the sender's email domain or by client, and it handles invoices, packing lists, bills of lading, and certificates in any format or language.

Yes. The agent extracts every row, normalises the data, and flags wrong quantities, mismatched totals, and missing fields automatically. Your team can then standardise and edit in bulk instead of correcting the same value hundreds of times. Extraction runs at 99% precision with zero manual data entry.

Yes. When the packing list gives weights by HS code or product group rather than per item, Sail allocates weight across every line automatically using the verified totals from your uploaded documents. That replaces the usual round trip through Excel formulas and rounding fixes.

The agent flags it before it reaches customs: totals that don't reconcile, weights that differ between the packing list and the B/L, party details that don't match, missing or duplicate documents. Your team sees the discrepancy in context, and ambiguous or high-risk cases always escalate to a person for the final call.

Yes, and it's a first in the industry. Instead of clicking through forms and dropdowns, your operators ask the agent questions about a shipment and get answers on the spot: why a document was flagged, where a value came from, what's still missing. Explanations reference the underlying documents so your team can verify instead of taking them on trust.

Sail assembles declaration-ready packages and formats the output for whatever your filing channel expects, whether that's a broker system, a national portal upload, or a structured export file. Submission itself stays with you or your broker, so your legal relationship with customs doesn't change.

It gives multinationals a single dashboard to validate shipment documentation, route clean packages to customs brokers, and monitor declaration status across entities, countries, and partners. Compliance and operations teams see what's ready, blocked, or pending at a glance instead of chasing spreadsheets.

Your brokers receive structured, validated records instead of raw PDFs they have to interpret. Documents are extracted, cross-checked, and pre-classified before handoff, and export packages arrive in formats partners can import directly. Errors get caught on your side before they become broker escalations or customs delays.

Yes. Every validation rule is configurable per client, country, trade lane, document type, or any combination of them. You set tolerance thresholds in percent for acceptable variances, so a weight off by 2% can pass while a consignee name mismatch gets flagged. The rules then apply automatically to every shipment without manual review. New rules and changes are configured with our team today and typically land within days, and plain-language rule creation by your own users is on the roadmap.

Yes. The Control Tower tracks assignment, progress, and turnaround by broker and market, so you always know who has the file and what's pending. Declaration throughput, duty exposure, and bottlenecks roll up into one operational view, and every document, validation decision, and handoff event is archived for audits.

Sail uses a classification model fine-tuned on your own data: product descriptions, bills of material, and your prior rulings and corrections. Every suggestion comes with a transparent rationale, a confidence score, and the alternatives that were considered, so your analysts review and approve instead of researching from scratch.

Yes. For products where classification depends on composition, the model evaluates the bill of material and sourcing context rather than the description alone. Judgments your team has already made on similar compositions carry over, so the same product family stops getting re-classified from zero on every export.

Yes. Classification runs with destination-specific tariff logic and local schedule requirements, so the same product gets the right code for each market you ship to.

Yes. The Compliance API processes catalogs in batch for migrations, recertification, or onboarding new product lines. You send product descriptions or structured specs and receive structured JSON with HS codes, origin outputs, duty context, and the reasoning behind each decision, ready to write back to your ERP or PLM.

We track the rollout closely. Dubai Customs is extending its tariff from 8 to 12 digits in phases under Customs Notice No. 10/2025, and practice at the counter still lags the official schedule. Sail maps existing 8-digit codes to their 12-digit successors so your declarations stay valid as the requirement tightens, and our blog keeps a current read on where things really stand.

No, and we don't claim it. In live programs classification accuracy runs in the high nineties, and most of the remaining misses trace back to gaps in the source data rather than to the model. That's why every suggestion carries a confidence score, low-confidence cases route to your experts, and corrections feed back into your model so accuracy keeps climbing after go-live.

Both. Catalogs can be classified on a schedule, weekly for example, or on demand as new products appear, and API calls return classifications in real time. You choose per workflow, and you can mix the two.

It determines preferential origin across your whole catalog, lane by lane, so every eligible shipment can claim the preference instead of paying the full duty rate. Sail evaluates each product's bill of material against the actual rules of the applicable trade agreements, manages the supplier declarations behind every claim, and your team reviews and signs off. Non-preferential origin is covered too, for the invoices and markings that need it.

It works through the rule alternatives of each agreement in order: wholly obtained, specified processes, tariff shift, and value thresholds. Every attempt is recorded with its arithmetic, and each run pins the exact BOM version, rule version, and declarations it used, so you can show a customs auditor precisely how a determination was reached.

Each agreement writes its own product-specific rule, and the tests calculate different things. One agreement might cap non-originating materials as a share of the ex-works price while another demands a minimum regional value content on a different basis, so identical figures can pass one test and fail the next. We walk through a real example across ten agreements on the blog.

Sail watches everything that can flip a status: new BOM versions, fresh batches, price moves, and updated agreement texts. Affected products are marked stale and re-run automatically, and the statuses in your ERP stay current without anyone maintaining a spreadsheet. New products and new lanes get picked up as your data syncs.

Sail's AI reads each incoming declaration, whatever the format or language, matches it to the supplier and the components it covers, and checks validity on arrival. Suppliers can submit over SFTP, direct API, or plain email, and the system routes each declaration to the right supplier and SKUs automatically. Renewal requests go out before expiry and chasers follow up on missing declarations, with a person confirming before anything is used. A claim that would rest on expired evidence is blocked, never silently passed.

Yes. The what-if panel simulates a different supplier, site, or price and shows the effect on preferential status with the same transparent trace as a live determination. Value-rule determinations also show your headroom, meaning how far costs can rise or prices can fall before status flips, and an uplift report finds products where one sourcing change would create preference that's not there today.

Yes. Reporting shows duty claimed under preference and duty left on the table, broken down by product and site, with the gaps ranked by value and the action that closes each one. Reports export to Excel, PDF, and CSV, and can be scheduled to a distribution list.

Lanes are derived from your historical shipment data, typically the last 12 to 24 months, so evaluation starts from where you actually trade rather than from a manually maintained list. New SKUs can inherit all active lanes or follow whatever logic you agree with our team.

Yes, and volume is rarely the constraint. A catalog of 100,000 SKUs runs through in about a week when the source data is clean, so data quality drives the timeline far more than product count. We'll tell you early if the data needs work before determinations are worth running.

Search by SKU, open the decision trace, and export the evidence package: the rules that were applied, the document versions, and the full history of the supplier declarations behind the claim. What took days of file-chasing becomes a lookup your team can do during the audit meeting itself.

The platform is ISO 27001 certified, and every decision, correction, and document handoff is recorded with full traceability.

Yes. Every classification, validation outcome, and correction captures the source document, the rationale, and a timestamp, so the file is ready for customs audits and post-clearance reviews without reconstruction work.

Every change, addition, and correction is tracked from the moment a document enters the system, so your team sees exactly who changed what, when, and why. Overrides are role-controlled and require written justification and sign-off, and both the system's original decision and the override are kept, so the record always shows the full history.

Yes. Validation thresholds, classification logic, and document requirements adapt per client, country, and trade lane, and new markets go live without code changes.

No. Classification and origin work can run on a restricted feed that exposes only the fields a decision actually needs, such as core material and composition data, while the rest of the BOM stays inside your systems. We already run this pattern with manufacturers whose formulations are trade secrets.

You do, all of it. Your documents, decisions, and history are exportable on request, and at termination we delete or retain the data according to your instruction.