What is AI Context?

AI Context lets you add business rules and interpretation notes to the datasets, columns, and dashboards Luzmo IQ uses to answer questions.

Names, descriptions, and data types are useful, but they don't capture the rules your team applies every day. AI Context is where you write those rules — directly on the object they describe. Luzmo IQ reads them when the related object is relevant to a question. It's guidance that shapes interpretation, not a hard filter or permission.

Use it to explain things like:

  • which date field to use for timeline questions
  • whether revenue is gross, net, monthly, or annualized
  • what a status value means in your business
  • which rows to exclude from standard analysis
  • how a dashboard should be interpreted for its audience

AI Context is optional — empty (or whitespace-only) context is ignored and changes nothing.

Where can you add AI Context?

Put guidance where it belongs: a field's meaning on the column, dataset-wide rules on the dataset, and audience or purpose on the dashboard.

Dataset AI Context

Rules that apply to the whole dataset.

This dataset contains completed orders only. Cancelled orders are excluded.

* Use `order_date` for sales trend questions.
* Revenue is net of refunds and excludes VAT.
* Fiscal years start in April.

Column AI Context

Field-specific meaning, caveats, or units.

Negative values represent refunds. Do not treat them as data errors.
This is the account creation date, not the first purchase date.

Especially helpful for ambiguous names, coded values, custom metrics, dates, and monetary fields.

Dashboard AI Context

Guidance for a specific dashboard experience — its audience, purpose, and interpretation.

This dashboard compares regional sales performance.
North America includes the United States and Canada only.
Use net revenue unless the user explicitly asks for gross revenue.

Dashboard AI Context helps Luzmo IQ answer in a way that matches the purpose of the dashboard, not only the raw datasets behind it.

When should you add AI Context?

Add it when a smart person could still misread your data from names and descriptions alone:

  • generic column names such as status, type, value, date, or amount
  • metrics with business-specific definitions
  • datasets with filtered or pre-aggregated rows
  • fields where null, zero, or negative values have a special meaning
  • dashboards used by a specific team, segment, or workflow

You don't need it everywhere — start with the fields that cause the most confusion or the questions Luzmo IQ answers least reliably.

How to add AI Context

The AI Context editors appear once AI-powered features are enabled on your account (they're hidden in the embedded dataset editor).

  • Dataset — on the dataset details page, in the AI Context section below the description.
  • Column — on the dataset details page, click the AI context (✨) button on the column's row; an editor opens below it. Only do this for columns that need extra guidance.
  • Dashboard — either the AI Context section on the dashboard details page, or in the Dashboard Editor: Settings → turn on AI context → Edit.

What should you write?

Specific, short, operational rules Luzmo IQ can apply.

Use `completed_at` for fulfillment timing. Use `created_at` only for order creation.
An active customer has at least one paid invoice in the last 90 days.
Rows with `test_account = true` are internal test data — exclude from business summaries.
`mrr` is monthly recurring revenue in EUR, already normalized to a monthly value.

Markdown (bullets, short headings, code formatting) keeps context easy to scan. Keep it concise: dataset and dashboard context is used up to ~2,000 characters and column context up to ~500 when sent to the AI, so put the most important rules first.

What not to write

Avoid context that is broad, stale, or misplaced.

  • Less useful: "This is a very important dataset. Please answer carefully."
  • Better: "Powers the executive revenue dashboard. Use net revenue; exclude rows where is_internal_account = true."

Never store secrets, credentials, or personal data in AI Context, and don't use it to enforce access — it's guidance for the AI, not a security control. Use permissions and your data model for anything that must always apply. And keep each field focused on the object it describes.

Before and after example

A column called revenue with no context leaves Luzmo IQ guessing whether it's gross or net, includes refunds, or includes tax. Add:

Revenue is net of refunds and excludes VAT.
Negative revenue values represent refunds.
Use `order_date` for revenue trends.

Now Luzmo IQ has the business rules it needs to interpret revenue questions more consistently.

Best practices checklist

  • Add dataset context for global rules and filters.
  • Add column context for ambiguous fields and key metrics.
  • Add dashboard context for audience, purpose, and interpretation.
  • Keep each field focused, and use examples when values aren't obvious.
  • Spell out units, time zones, fiscal calendars, and date semantics.
  • Note important exclusions, such as test accounts or cancelled orders.
  • Review AI Context when your business definitions change.

How to evaluate whether AI Context helps

Build a small set of questions your users actually ask, and compare answers with and without the context:

  • does the answer use the right fields, filters, and definitions?
  • does the explanation match your business rules?

Because AI answers can vary between runs, test a few times. AI Context helps most when a question depends on hidden business meaning — it won't fix a wrong chart, aggregation, or dataset choice, which need separate data modeling or IQ-quality work.

FAQ

Does AI Context replace descriptions?

No. Descriptions stay useful for people and general metadata quality. AI Context specifically guides AI behavior with business rules and caveats.

Will empty AI Context change my answers?

No. Empty or whitespace-only AI Context is ignored.

Is AI Context translated per language?

No. It's one shared text field for all languages — unlike names and descriptions, it isn't localized.

Should I add AI Context to every column?

No. Start with high-impact fields: metrics, dates, statuses, IDs, coded values, and fields users often misread.

Can I use Markdown?

Yes. Luzmo IQ reads the text you write (the source, not rendered HTML), so bullets, short headings, and code-formatted field names all help.

Can AI Context enforce a rule?

No. It guides Luzmo IQ's interpretation; it isn't a database constraint, permission rule, or guaranteed query filter. Use your data model and access controls for hard enforcement.

Can I manage AI Context through the API?

Yes. AI Context is exposed as ai_context on the relevant API resources — useful for syncing from another system or managing it programmatically.

Related articles

Need more information?

Do you still have questions? Let us know how we can help.
Send us feedback!