Context Inventory Checklist

Before you build a prompt, list what it actually depends on. A prompt built from stale rate cards or an undocumented assumption fails quietly — the answer still looks confident. Check off what you've confirmed; nothing here is saved or sent anywhere.

Source data

Which system is the source of truth for this data — and is that the system you're actually pulling from?
How current does this data need to be for the task, and how current is it actually going to be?
Does the export contain anything the AI tool shouldn't see — other counterparties' data, personal data, full account numbers?

Policies, rate cards & reference documents

Which policy, rate card, or approval matrix does this task actually depend on?
Is that document the current version — who owns it, and when was it last confirmed accurate?
If the document conflicts with what someone remembers or assumes, which one wins?

Constraints the AI needs to be told, not assumed

What format does the output actually need to be usable — a number, a table, a drafted message?
What must never appear in the output (amounts, names, internal thresholds) even if it's in the source data?
What should the tool say when it doesn't know, instead of guessing?

Ownership & review

Who is responsible for keeping each source document current — and do they know this task now depends on it?
Who checks the AI's output against these sources before it's relied on?
When was this inventory last reviewed — is anything on this list now stale?

Once your inventory is clear, turn it into an actual prompt with the Treasury Prompt Builder — task, context, constraints, and format, explicit instead of implied.

← Back to path

Contact LinkedIn ↗Terms and Conditions Privacy Policy Cookie Policy (EU)

We use a small number of cookies to understand how the hub is used (Google Analytics). Nothing loads until you say yes — see theCookie Policy.