Build interfaces from observed activity
Use Orloi's assistant to turn captured activity into a specification for an interface or dashboard. Then give that specification to Airtable Omni, Zite, or an AI app builder of your choice. This guide uses Omni and Zite as the main examples; you can adapt the handoff prompts for another builder that connects to your intended data source.
Orloi contributes context about how people work with your base: the records and fields they change, recurring operations, and observed handoffs. The builder uses that context alongside your base schema. This is a prompt handoff; Orloi does not build or publish the interface for you.
Analyze the work
Start with an observed base that has enough captured activity to represent the work you want to support. Replace the bracketed placeholders in these prompts.
For an interface tailored to one person, ask Orloi:
Review [person]'s captured activity during [period], along with their
available profile and operations. Which tables, records, and fields
feature in their work? Identify recurring changes and observed handoffs.
State the period and evidence behind your conclusions. Separate observed
behavior from inferred responsibilities, and flag gaps in the evidence.For a team or workflow dashboard, use:
Review captured activity for [team or workflow] during [period]. Suggest
a dashboard that would help us understand or act on this work. Explain
which observations support each proposed chart, summary, or record list.
Do not treat edit counts as a measure of productivity or assume that
every edit represents completed work.Activity shows captured changes, not everything a person does. Reading records, using an existing interface, or doing work outside the captured scope may leave no evidence. A few edits to a linked-record field can justify making that relationship easier to access; they do not by themselves establish a cleanup or approval workflow.
Make the design concrete
Ask Orloi to check the schema before producing the specification:
Inspect the available base schema and turn this analysis into a small
interface specification. For each proposed page or block, give:
- Its purpose and the activity evidence supporting it.
- The exact source table and existing field names and types.
- Visible fields, filters, sorting, and grouping where useful.
- Record detail fields and actions, including which fields are editable.
Separate confirmed facts, assumptions, and proposed additions. Do not
invent fields, statuses, assignment rules, or review queues. If you
cannot verify a field or rule, identify what I need to confirm. Prefer
the smallest design that supports the observed work.Confirm the assumptions with the people who will use the interface. For example, records someone recently edited are not necessarily records assigned to them. A personal queue needs an actual assignment rule.
Choose how to build it
| Implementation | Use it when | Editing afterward |
|---|---|---|
| Omni + standard interface | Native Airtable layouts fit the work | Normal no-code Interface Designer controls |
| Omni + custom interface | You need a bespoke visualization or interaction inside Airtable | Omni or source code, plus exposed no-code properties |
| Zite | You want an external app or dashboard connected to Airtable | Zite's builder |
Omni + standard interface
Omni can create native Interface Designer pages, such as list layouts. Explicitly request a native page: a generic request to build an interface typically produces AI-generated custom elements. Omni cannot currently create Gantt or Blank layouts. Blank is Airtable's older layout for freely placing individual elements. See Omni's capabilities.
Ask Orloi:
Turn the agreed specification into a self-contained prompt for Airtable
Omni. Request native Interface Designer pages using standard layouts
and components, editable afterward through the no-code editor. Specify
the source tables, exact fields, filters, sorts, and allowed actions.
Do not request AI-generated custom elements or a Blank layout. Identify
anything that needs manual configuration rather than promising Omni
can build it. Use existing tables and fields; list proposed schema
changes separately for my review.Paste the resulting prompt into Omni in the relevant base. Check that its plan describes native pages before building. You can adjust the resulting page through Interface Designer.
Omni + custom interface
Omni can generate custom table layouts or dashboard elements backed by JavaScript/React. You can revise them through Omni or edit their source code. The properties panel also supports data sources, filters, sorting, actions, and any custom properties the element exposes. This does not make every part of the generated design editable through no-code controls. See AI-generated interface elements.
Ask Orloi:
Turn the agreed specification into a self-contained prompt for Airtable
Omni to build an AI-generated custom interface. Describe the layout,
data sources, exact fields, and interactions. Specify which settings
should be exposed as configurable properties. Use existing tables and
fields; list proposed schema changes separately. Include only features
whose required data is available, and list unresolved dependencies.For a custom element inside an existing dashboard, open that dashboard page before asking Omni to build it. Otherwise, Omni defaults to a full table layout. Airtable currently lists generated elements as unavailable on mobile devices, so check that limit if the intended users work on phones.
Zite
Zite connects to Airtable to build external apps and dashboards. The handoff needs both the interface specification and a connected data source; pasting a prompt does not connect your base.
Ask Orloi:
Turn the agreed specification into a self-contained prompt for Zite to
build an external app connected to my Airtable base. Include the pages,
exact tables and fields, filters, sorts, and interactions. Distinguish
read-only displays from actions that write back to Airtable. List the
connection and access requirements I must configure. Do not include
credentials or unnecessary personal record values in the prompt.Connect the intended Airtable base in Zite, then use the generated prompt. Check access with the intended user account before sharing the app, especially if it allows record edits.
Keep the data requirements explicit
An interface designed from Orloi analysis can display current Airtable records without displaying Orloi's stored history. A change timeline, before/after comparison, or filter for records changed by an automation requires that history to be available to the builder. Do not assume those values exist as Airtable fields.
Ask Orloi to check the handoff:
Review the builder prompt. Which features use existing Airtable data,
and which require Orloi history, derived values, or new fields? Remove
unsupported features from the initial build and list their data
dependencies separately. Keep the final prompt self-contained so the
builder does not need access to this conversation.For a dashboard that reads Orloi's stored activity or metrics directly, see the SQL and BI cookbook. That requires a separate data connection or export.
Before using the finished interface, verify its fields and record selection, try its filters and sorts, and check any write actions on test records. Compare the result with the agreed specification and refine the prompt where the builder made assumptions.