Knowledge Docs
A knowledge doc is your company's own rulebook, written down where the AI can read it. Your tariff sheets, operating procedures, quality standards, customer-specific terms, and glossaries live as knowledge docs, and the AI reads them on demand while it works. It answers questions and checks documents against your rules and your prices — not a generic guess about how logistics usually works.
What a knowledge doc is
The AI already knows general logistics. What it cannot know is how your company runs: the rate you charge on a given lane, the tolerance your quality team accepts, the free-time a customer agreed to, the term your team uses for a specific charge. A knowledge doc holds exactly that. It is reference material — the AI searches it and reads the relevant part when a task calls for it, the way an experienced dispatcher reaches for the tariff binder instead of guessing.
Knowledge docs are read on demand, not memorized. The AI pulls the doc that applies to the task in front of it and cites what it finds. Nothing is baked into a model, so when you change the doc, the next answer follows the new version.
What belongs in a knowledge doc
The test is simple: put in a knowledge doc only what the AI cannot already know. Your prices, your procedures, your thresholds, your customers' agreed terms — the facts that are specific to your business and change when your business changes.
Leave out general knowledge and methodology the AI already has. A doc titled "how to write a professional email" or "what a bill of lading is" only slows the AI down, because it re-reads common knowledge it already carries. If a competent operator would know it without looking, so does the AI.
| Belongs in a knowledge doc | Does not belong |
|---|---|
| Your lane rates, surcharges, and rate-sheet effective dates | How freight pricing works in general |
| Your standard operating procedure for order intake | A generic description of order processing |
| Your quality thresholds and accept/reject tolerances | Industry background on quality control |
| A specific customer's agreed free-time and payment terms | Standard Incoterms definitions |
| Your in-house glossary of charge codes and abbreviations | Common trade terms the AI already knows |
How the AI uses your knowledge docs
Once a knowledge doc is in place, the AI reaches for it across everything it does:
- Answering questions. Ask "what's our rate on the HCM to Đà Nẵng lane?" and the AI quotes the figure from your tariff sheet and cites it — "per your rate sheet effective 1 June, that lane is …" — instead of estimating.
- Cross-checking work. When the AI reads an incoming invoice or a supplier quote, it compares the numbers against your rate sheet and flags anything that doesn't match the agreed price.
- Carrying out procedures. When the AI intakes an order or drafts a document, it follows the steps and rules in your SOP — the checks, the required fields, the sequence your team actually uses.
- Grounding workflows and app agents. Automations and the AI agents built into your apps read the same knowledge docs, so a workflow that classifies an order or an in-app agent that drafts an estimate applies your rules, not generic defaults.
Because the AI cites the rule it used, you can see why it answered the way it did and check the source.
Creating and maintaining knowledge docs
You create a knowledge doc two ways: chat with the AI to write one from what you tell it, or upload a file you already have — a rate sheet, an SOP document, a standards guide — and it becomes a knowledge doc the AI can read.
Editing is just as direct. When your rates change or a procedure is updated, you edit the doc, and the AI's answers follow the current version from the next question on. Every edit is kept as a version, so you can see what a doc said at any point and restore an earlier one.
You control where each doc applies. Turn a doc on for the whole workspace when it is a company-wide rule, or scope it to a specific table so the AI reads it only where it is relevant — for example, a repair-rate sheet that applies only when working in the damage-estimates table. Sharing a doc with a teammate makes it available to them, and each person decides whether to switch it on, so an irrelevant share never clutters someone else's answers.
Knowledge docs also ship as installable content packages. A set of standard docs — say, a group of SOPs and standards used across every branch — can be published once, installed into many workspaces, and upgraded in place as you revise them. Upgrades refresh the shipped docs without overwriting local edits unless you approve the change.
Knowledge docs versus instructions and skills
Knowledge docs are one of three ways to give the AI your company's know-how. They differ in when the AI applies them:
| What it is | When it applies | Example | |
|---|---|---|---|
| Instructions | Standing rules always enforced | Automatically, on every relevant action | "Never create a supplier without a tax code." |
| Knowledge docs | Reference the AI looks up | When a task calls for the information | Your tariff sheet, quoted when someone asks for a lane rate. |
| Skills | A step-by-step procedure for a named task | When you or the AI runs that task | Your customs-clearance checklist, run on an inbound shipment. |
Use an instruction for a rule that must hold every time. Use a knowledge doc for facts the AI should consult when relevant. Use a skill for a repeatable procedure the AI should follow the same way each time.
Examples in logistics operations
A forwarder's tariff and surcharge sheet. Load your published rates, surcharges, and their effective dates as a knowledge doc. When a customer asks for a quote, the AI builds it from your sheet and cites the lane and date it used. When a carrier invoice arrives, it checks the billed rate against the same sheet and flags a mismatch before the invoice is approved.
A depot's repair-rate table. Put your agreed repair rates per damage type and container size in a knowledge doc, scoped to your estimates table. When the AI drafts a damage estimate from a survey, it prices each line from your table instead of guessing, so every estimate uses the rate your customers agreed to.
A distributor's delivery SOP. Write your order-intake procedure — the required fields, the credit check, the cut-off times, the FEFO picking rule — as a knowledge doc. When an order arrives, the AI works through your steps in order, applies your checks, and holds anything that fails one, so intake follows the same process no matter who is on shift.
Frequently asked questions
What should I put in a knowledge doc?
Only what the AI cannot already know: your prices, your procedures, your thresholds, and your customers' agreed terms. Leave out general knowledge and how-to methodology — the AI already has those, and restating them only slows it down.
How is a knowledge doc different from an instruction?
An instruction is a standing rule the AI always enforces on every relevant action ("never ship without a signed delivery order"). A knowledge doc is reference material the AI looks up only when a task calls for it, like your rate sheet consulted when someone asks for a price.
Do I need to retrain the AI when my rates change?
No. Edit the knowledge doc, and the AI's answers follow the current version from the next question on. Nothing is retrained, and every edit is kept as a version you can review or restore.
Where does a knowledge doc apply?
Wherever you turn it on. Activate it for the whole workspace for a company-wide rule, or scope it to a specific table so the AI reads it only where it is relevant.
Can the AI use my knowledge docs in workflows too?
Yes. Workflow automations and the AI agents built into your apps read the same knowledge docs, so automated steps apply your rules the same way a chat answer does.
Can I reuse the same knowledge docs across workspaces?
Yes. Publish a set of docs as a content package, install it into each workspace, and upgrade it in place as you revise the docs. Upgrades refresh the shipped docs without overwriting local edits unless you approve.