Question 1

What is Portals AI and what can I do with it?

Accepted Answer

Portals AI is a workspace for connecting your data sources, normalizing them into a shared vocabulary, and asking questions about them in chat-style sessions called portals. You can pull in data from CSV and Excel files, Google Sheets, Excel on OneDrive, and REST APIs, define how their fields relate, and then explore or analyze the combined data through portals.

Question 2

How do I connect my first data source?

Accepted Answer

Open the Connectors page, switch to the Catalog tab, and pick a connector definition (for example, File Upload for CSV and Excel files). Configure it with your file or credentials — that creates a connector instance you can use across stations.

Question 3

What is a Station and why do I need one?

Accepted Answer

A station is a workspace that bundles together the views, connector instances and tool packs you want to analyze together. You need at least one station before you can open a portal, because the portal uses the station's data and tools to answer your questions.

Question 4

How do I start asking questions about my data?

Accepted Answer

From a station, click "Open Portal" to open a chat session. Type your question — the assistant will use the station's tool packs (data query, statistics, etc.) and the connected data to respond. You can pin useful results for quick access later.

Question 5

Is there a limit on how many messages I can send?

Accepted Answer

Each plan includes a send ceiling for portal sessions — a small per-minute burst limit and a daily limit that resets at midnight UTC. Sending a message never uses your tool-credit allocation; the ceiling is a separate fair-use bound, and hitting it just pauses new messages until the window resets (a notice in the session says exactly when). Higher plans raise the ceiling, and enterprise plans have none.

Question 6

How do I upgrade my plan?

Accepted Answer

Go to Settings → Subscription & Billing and pick a plan from the list — Subscribe takes you to Stripe's secure checkout. When you return from checkout, the Subscription & Billing tab confirms the payment with Stripe and shows your new plan; if the payment is still processing, it says so, and Subscribe stays unavailable so you can't pay twice. Your usage allocations reset on your new billing cycle from that point on. To change or cancel later, use Manage subscription on the same tab, which opens the Stripe billing portal.

Question 7

Who can manage billing?

Accepted Answer

Anyone with permission to manage billing. Everyone can see the available plans on Settings → Subscription & Billing, but Subscribe and Manage subscription are enabled only for those with that permission — the server enforces this too, so it isn't just hidden buttons.

Question 8

My plan says it's managed — what does that mean?

Accepted Answer

Your organization is on a custom plan arranged with us directly rather than one of the self-serve plans, so the plan list and checkout are hidden. Contact us to make changes to a managed plan.

Question 9

Why is a toolpack marked “Inactive on your plan”?

Accepted Answer

Your organization's plan determines which toolpacks are available. Custom toolpacks registered while they were included stay saved when a plan stops including them — nothing is deleted — but their tools aren't offered in portal sessions and new registrations are blocked until you're on a plan that includes them again. Built-in toolpacks can also vary by plan. Everything reactivates automatically on upgrade; station setups are preserved exactly as you left them.

Question 10

What's the difference between a connector and an entity?

Accepted Answer

A connector instance is the live link to a source (your uploaded file, your Google Sheet, your REST API). A connector entity is one specific data object exposed through that connection — a single sheet, table, or collection. One connector instance often surfaces several connector entities.

Question 11

What are column definitions and why do they matter?

Accepted Answer

Column definitions are reusable field specifications — name, data type, validation rules, canonical format — shared across your whole organization. They let different connector entities map their raw fields to the same shared vocabulary, so analysis can compare like with like.

Question 12

What are field mappings?

Accepted Answer

A field mapping links a raw field on a connector entity to a column definition. Mapping the CRM's "Email Address" and the CSV's "email" to the same email column definition is what makes records from different sources comparable.

Question 13

Why don't records the assistant created show up when I query them?

Accepted Answer

A record is only queryable once a field mapping projects its fields into the entity's columns — records written with unmapped fields stay invisible to queries and tables. When you ask the assistant to create data on a new entity, it now reads your organization's column-definition catalog, maps the columns first (creating field mappings), and only then writes records. If a column it needs has no matching definition, it will tell you rather than write data that can't be read back. To fix older invisible records, add the field mappings for their columns, then re-create or re-sync.

Question 14

How do I validate my data?

Accepted Answer

Validation is driven by your column definitions: each one can declare a data type and a regex validation pattern. When records are synced or written, those rules run automatically — invalid records are flagged so you can fix them. Use "Re-validate" on an entity to re-run the rules across all records.

Question 15

What happens when I sync an entity?

Accepted Answer

Syncing pulls the latest records from the connector source into Portals AI. New records are added, changed records are updated, and the entity's record count and last-sync timestamp are refreshed. A sync runs as a background job you can monitor on the Jobs page.

Question 16

When should I use a SQL filter instead of the filter builder?

Accepted Answer

Use the builder whenever it can say what you mean: it covers comparisons, lists and text matching on any column. Switch a view's row filter to SQL when it can't, for example rows within a distance of a point, a value inside a JSON column, or one column compared with another (amount > cost). A SQL filter is one condition over the entity's own columns. It's checked when you save, so a mistake is caught then, and Preview shows the rows it selects first. Describe the rows you want and Suggest writes one for you to review. SQL filters need a plan that includes advanced filters.

Question 17

What are entity groups and when should I use them?

Accepted Answer

Entity groups bundle related connector entities from different sources together — for example, your CRM Contacts and your support tool's Users — so you can analyze them as one population. Use them when the same real-world thing (a person, a company, an account) appears in multiple sources.

Question 18

What is a "link field" in an entity group?

Accepted Answer

The link field is the shared identifier used to match records across the group's members. If you pick "email" as the link field, two records — one from each source — are treated as the same entity when they share an email address. The Overlap Preview shows how well your link field actually matches data across members.

Question 19

How do tags work?

Accepted Answer

Entity tags are color-coded labels you attach to your entities to group and filter them. Tags appear as colored badges throughout the app and let you narrow long lists down to the subset you care about — for example, tagging all customer-data entities so you can filter to them in one click.

Question 20

How do I rename my organization?

Accepted Answer

Open Settings → Organization and choose Rename next to the organization's name. Rename appears for anyone with permission to edit the organization. Names don't have to be unique. The new name shows straight away in the header and the organization switcher, and on invoices from then on. The rename is recorded in Settings → Activity, and if you later delete the organization, the confirmation asks for its new name.

Question 21

How do I delete my organization?

Accepted Answer

Open Settings → Organization and scroll to the Danger zone. Only someone with permission to delete the organization can delete it, and the action is permanent: all organization data — stations, portals, connectors, records, and uploads — is destroyed, and every member loses access. You'll be asked to type the organization's name exactly to confirm, and you'll be signed out once the deletion completes. If a background job is still running, wait for it to finish (or cancel it) first.

Question 22

How do I invite someone to my organization?

Accepted Answer

Open Settings → Members and invite them by email, choosing their role. You'll get an invitation link — copy it and send it to them; it's shown only once and expires after 7 days. They join when they sign in with that email. The Members tab appears for anyone with permission to invite members.

Question 23

How do I control who can see and change things?

Accepted Answer

Two ways. Share an individual station, pinned result, or curated view from its Share action — read-only or read & write for stations and pins, read-only for views. Or, on a plan with custom roles, use Settings → Access to build policies, bundle them into roles, and grant them to members or groups. The server checks every request against the same rules, so hiding something in the app and refusing it are the same decision.

Question 24

Why is an action missing or greyed out for me?

Accepted Answer

Each button reflects what you can do with that specific object. If you don't have permission, the action is usually hidden — ask for access to that object. A page's main action, like Create on a list page, stays visible but disabled, and its tooltip says what access you need. If your plan doesn't include it, it points you to Settings → Subscription & Billing. If something temporary blocks it, like a sync or import still running, it's disabled and the tooltip says why; it comes back when the job finishes.

Question 25

Where can I see who changed what?

Accepted Answer

Settings → Activity lists the organization's security-relevant actions — sign-ins, invitations, member and role changes, shares, policy edits, and organization settings such as a rename or a new default station — with who did each one and when. It's visible to anyone with permission to read the audit log, and entries are kept for about 24 months.

Question 26

What are tool packs?

Accepted Answer

Tool packs are bundles of capabilities you enable on a station — data query, statistics, regression, financial analysis, web search, entity management, access & roles, curated views, visualization, and GIS (maps and geocoding). Your plan decides which ones are available. The packs you turn on determine what kinds of work the assistant can do during portal sessions on that station.

Question 27

Why is a tool pack on my station greyed out, and why won't the assistant use it?

Accepted Answer

That pack isn't included in your current plan. Plans decide which tool packs are available, and a station can keep a pack that a later plan change excluded — so instead of removing it, we mark it inactive: the chip stays visible but dimmed, the pack is badged "Inactive on your plan" on the Toolpacks page, and it can't be added to another station until the plan includes it. Its tools are absent from portal sessions, which is why the assistant will tell you the capability isn't on your plan rather than pretending it doesn't exist. What's gated is the *assistant's* toolset, not the app: where a screen exists for the same work — creating an entity and its columns, for example — you can still do it yourself regardless of plan. Nothing was deleted, and nothing needs re-attaching: change the plan in Settings → Subscription & Billing and the pack is live again in your next portal session.

Question 28

How do I save results from a portal session?

Accepted Answer

When the assistant returns a chart, table, map, or text block you want to keep, click the pin action on that portal result. Pinned results show up on your dashboard and on the Pinned Results page so you can return to them without rerunning the conversation. A pinned chart, table, or map keeps the query behind it rather than a copy of the data, so it's always current; a result that has no query behind it can't be pinned, and the pin action says so.

Question 29

What do people see when I share a pinned result?

Accepted Answer

They see live data as you see it, not as they would. A pinned chart, table, or map runs its query with your access every time it's opened, so someone you share it with sees your current view of the data, even data they couldn't open themselves, and never more than you can see. A shared pin is labelled with whose view it shows. If you lose access to that data, or leave the organization, the pin empties for everyone: people you shared it with see that its data isn't available, and you see why. To share a view that outlasts a change in your access, have someone who keeps access pin it.

Question 30

What's the difference between a portal and a portal result?

Accepted Answer

A portal is the whole chat session — the conversation, the messages, the context. A portal result is a single piece of structured output (a chart, a table, a text block) produced by the assistant inside one of the messages. You pin portal results, not portals.

Question 31

Why do some results appear inline and others as a separate streamed table?

Accepted Answer

It depends on size, and the assistant doesn't choose — it's automatic. A small result comes back inline inside the message. A larger one returns a query handle: the full rows stream into a table or chart you can scroll, while the assistant keeps only a small preview so a big result never crowds out the conversation. You still see every row either way.

Question 32

How do I refresh a chart, map, or table with the latest data?

Accepted Answer

You usually don't have to. Every visualization — chart, map, or table, at any size — remembers the query behind it, so when you reopen a session it reloads current data on its own instead of showing a stale snapshot; no need to ask the assistant again. If you want to be sure it's current right now, each one has a refresh button, and an “Updated … ago” note tells you how fresh the data is. A refresh re-runs the whole query, so a “top 10” list picks up a new leader rather than just restating the old rows. If a refresh can't reach your data, a widget in a session keeps showing the last good result and says so instead of going blank. A pinned one keeps no copy of the data, so it opens with current data every time, as the person who pinned it sees it. Refreshing is always free — it never uses your usage allocation. (If a visualization predates this feature, it'll say so and you can re-run the prompt to regenerate it.)

Question 33

Can I show my data on a map?

Accepted Answer

Yes, if your data has location. When a source carries geographic shapes (for example an ArcGIS parcel layer) they're imported as a geometry column; latitude/longitude number columns work too. If your data has only text addresses, the assistant can geocode them into map points — one address on the fly, or a whole column as a background job (a repeated address is cached and free). Then just ask — “show all vacant parcels on a map, colored by zoning” — and the assistant renders an interactive map you can pan, zoom, and click to inspect. Spatial questions like distances and areas are answered directly from your data as well. Mapping is part of the GIS tool pack, available on the plans that include it (see Pricing); add it to a station like any other tool pack.

Question 34

Do failed tool calls use up my usage allocation?

Accepted Answer

No. Some tools consume usage units from your organization's monthly allocation when they run in a portal session — mainly web search and heavier analytics like clustering. You're charged only when a tool call succeeds: if a call fails (for example, a search provider is briefly unavailable), it costs nothing. Your allocation resets at the start of each billing period, and you can see what's left on the Settings → Organization page and in a portal session's details — the Itemized usage button there lists every charged call (which tool, how many units, when, and who ran it).

Question 35

Why are the assistant's answers vague or missing my data?

Accepted Answer

Almost always because the station has nothing to answer from. A portal answers from records that have actually been imported into the station's entities — not from general knowledge — so a station with no connected source, or one whose sync hasn't run yet, can only answer in generalities. Open the station and check that it has a view attached and shared with you, or a connector instance, and that its entities have records. A view you can't read shows on the station with a lock. If a source is connected but empty, run a sync and wait for the job to finish, then ask again.

Question 36

How should I word my questions to get better answers?

Accepted Answer

Two things help more than anything else. Ask about one thing at a time — a narrow question routes cleanly to the right tool, while a compound one gives the assistant several jobs at once and no clear place to start. And use the station's own vocabulary: name entities, columns, and values the way they appear in your data, because that's what the assistant matches against. If your column is called subscription level, ask about subscription level rather than plan tier.

Question 37

Why does the assistant say my data is incomplete while an import is running?

Accepted Answer

Because it is, for now. While a background job is importing or syncing a connector instance, that data is still arriving and the records it owns are read-only — syncing, editing fields, and deleting are paused until the job finishes, and the connector's page (and each affected entity's page) says so while it runs. Ask again once the job reaches a terminal state and the answer will reflect the full data. Checking for a running job is worth doing before you trust a number that looks off.

Question 38

What do job statuses mean?

Accepted Answer

Pending — queued, not started yet. Active — currently running. Completed — finished successfully. Failed — stopped because of an error. Stalled — appears stuck and may need attention. Cancelled — stopped by you. Awaiting confirmation — waiting on input from you before continuing.

Question 39

Why did my job fail?

Accepted Answer

Open the job from the Jobs page to see the error details — the most common causes are invalid source credentials, malformed data that fails column-definition validation, network timeouts to the source, or insufficient permissions on the connector. Fix the underlying issue and re-run the job.

What Portals AI does

Connect anything you already have

Agree on what the fields mean

Ask, don't query

Run it in your own cloud

Extend what the agent can do