How the AI fits into the editor
A first draft that starts from your data
Rahoto’s AI starts from your dataset, your report and your brand kit, not a blank chat box. It checks what it writes against your data and shows every change before it’s applied. You review, refine and ship.
- Starts from your data
- Checked against the data
- Previewed before applied
- Identifiers pseudonymised
- Residency-aware
- Bring your own model
It starts from your data
Rahoto’s AI starts from your dataset, your report and your brand kit, not a blank chat box. That is also why it can be careful with what it sends out.
AI providers see the shape of your data (columns, types, summary figures), not your rows.
Names and identifiers are swapped for stand-ins before a request leaves, and swapped back when the answer returns.
AI requests honour your organisation’s residency policy; if no in-region provider is available, Rahoto uses its local engine instead of sending data out of region.
Bring your own model: route the AI tasks you choose to your own OpenAI- or Anthropic-compatible endpoint.
Rahoto does not train on your data unless your organisation opts in, and withdrawing removes every example it contributed.
AI capacity per plan is on Pricing. Read the full detail in our Privacy Policy or the Trust Center.
It is checked against your data
The checks run before you see a draft and before a scheduled issue goes out. They reduce the chance of a wrong figure; they are not a guarantee that every statement is correct, which is why every change is previewed and a run that moved waits for you.
| Check | What it does | Where it applies |
|---|---|---|
| Fields that exist | If a draft names a field your data doesn’t have, Rahoto fixes it or drops it before you see it. | Drafting, in the editor |
| Words against numbers | Before each scheduled issue, Rahoto checks the text against the data. If something material changed, the run waits for your review. | Scheduled issues |
| Previewed, then editable | Every AI change is shown before it’s applied, and afterwards it’s an ordinary, editable part of the page. | Editor |
| Reader questions | Readers can ask the published report questions and get answers computed from the data published with it. | Published reports whose charts are ready for questions |
What the AI can do
Each of these is in the product today. Where one needs an AI provider configured for your workspace, or has a limit, it says so beside it.
| Capability | What it does | Limit |
|---|---|---|
| Chart from a description | Describe a chart in plain English and Rahoto adds it to the page as an editable chart. | The AI only picks which of your columns to chart; with no AI provider configured a rule-based picker chooses instead. With no dataset bound, the chart is added empty until you bind data. |
| Report commentary | Rahoto drafts the written commentary for a report from your data, and undoes any AI rewording that introduces a figure your data doesn’t support. | Without an AI provider, the commentary is built from your data’s own figures, with no rewording. It is not a guarantee that every statement is correct. |
| Tone and audience | Rewrite a passage more formally, in your report’s brand voice, or for an executive audience, and review the change before it applies. | Works on the passage you select, one at a time, and needs an AI provider. |
| Critique | While drafting a report, Rahoto reviews its own draft for hierarchy and readability before you see it. | Needs an AI provider. It makes small refinements such as type sizes, heading levels and spacing; it does not delete pages or replace charts. |
| Repair | If the draft has problems such as overflowing text or elements off the page, Rahoto repairs them before you see it. | Needs an AI provider. It repairs layout and wording that is grounded in your data; problems with data bindings are surfaced for you rather than guessed. |
| Explain a chart | Select a chart and ask Rahoto to explain it or suggest insights. | The request carries a privacy-safe summary of the chart, not raw data values. With no AI provider configured, you get a statistical summary (range, total, trend) computed in your browser instead. |
| AI spend limits | Workspace admins set daily AI spend limits per person and per workspace, and choose whether a request over the limit is blocked or handled by Rahoto’s local engine. | Limits are in US dollars per day and apply to paid, cloud requests. Admin role required. |
| AI request log | Rahoto keeps a log of the AI requests in your organisation, with personal data scrubbed, and workspace admins can export it through the API. | Export is an API call, not a download button in the app. Entries are kept for your organisation’s retention window, then reduced to anonymised counts. |
Then it drafts the report
Describe the report and Rahoto drafts it in your design system for you to edit. Drafting needs an AI provider available to your workspace, and it stops rather than guess when a dataset’s profile is incomplete.
Readers ask the report questions
Readers can ask the published report questions and get answers computed from the data published with it. It works on reports whose charts are ready for questions.
The draft and this exchange are illustrations, not live sessions.
Try it on the report you send most often
Bring a CSV or connect your warehouse. Describe the report and Rahoto drafts it in your design system for you to edit.
Need enterprise governance? See the Enterprise page for bring-your-own-model and residency controls.