Rahoto and AI dashboards

As announced

What AI dashboards do well

Both are new, and both are built for ad-hoc exploration: ask a question about your company data and get a live dashboard back. This is how each company described its product at launch.

• Claude Dashboards (beta, announced 8 October 2026)

  Ask a question in plain language and Claude builds a live dashboard on your company data, from cloud warehouses and Salesforce, and keeps it current. Click any number to see the query behind it; each chart shows when its data was last refreshed. Teams edit dashboards together, and admins control sharing beyond the team.

• Data agent in ChatGPT Work (announced 10 September 2026)

  Connects to approved warehouses, databases and files, investigates business questions in conversation, and builds dashboards that can be edited, shared and refreshed, including in existing BI tools. It shares results through Slack or email and performs approved actions through connected tools. Admins manage connections, and existing table, row and column permissions apply.

Descriptions are taken from each company’s announcement on the date shown; we have not tested either product for this page, and both may have changed since.

Claude Dashboards announcement, 8 October 2026 · OpenAI Data agent announcement, 10 September 2026 · Data agent for ChatGPT Work, OpenAI community post, 10 September 2026

Side by side

An answer on a screen, or a report that goes out

A dashboard is for the people who open it. A recurring report goes to people who never will: clients, a board, a regulator, every region manager. Each row is one thing that report needs.

What Rahoto is built for

What a delivered report needs

• One report, a version for every audience

  Give each client, region or board its own version of one report: hide or redact pages and elements, scope the rows to that audience, add a watermark, and give each version its own schedule and recipients. A recipient can get their own slice of the data. If a version’s data scope can’t be applied, the run stops rather than sending unscoped rows.

• A finished report, delivered

  PDF, PowerPoint, PNG or a live web link, sent by email, Slack, Teams, webhook, S3, Google Drive or SFTP. Someone who receives the PDF doesn’t need a Rahoto account to read it.

• Reports that run themselves

  Set the cadence once; Rahoto refreshes, rebuilds and delivers. “Data as of” is filled in on every run, readers can open earlier editions, and if a source fails you can choose to send the last good edition, labelled with its date, instead of a gap.

• Your data rules, everywhere the data goes

  Rules set by your data owners hold in analysis, in the editor, in reader questions and in what gets delivered, and a report goes out with its publisher’s access, never more. A report bound to sensitive personal-data columns, such as health information or national ID numbers, won’t publish, and reader questions for restricted readers are never answered from fewer than five rows.

• Your brand, held

  Your fonts, colours and logo as a brand kit, applied in one step, with the properties that must not change locked. The page stays a freeform design canvas, so the report looks like your best work in March and in December.

• Numbers checked before they go out

  Before each scheduled issue, Rahoto checks the text against the data; if something material changed, the run waits for your review. Preflight checks every page; it advises and never blocks your download or share. Every figure in a draft must come from a query Rahoto ran on your data: a query naming a field your data doesn’t have is refused, and a page with an untraced figure goes back before you see it.

• Data beyond the warehouse

  Read from 73 data sources: warehouses, databases, SaaS tools, files and documents, alongside the warehouse you already trust.

• AI under your control

  Names and identifiers are swapped for stand-ins before an AI request leaves, and AI requests honour your residency policy, falling back to Rahoto’s local engine rather than leaving the region. Route the tasks you choose to your own model, set spend limits, and export a scrubbed log of AI requests.

• Built for a team

  Edit the same report together, comment, compare saved versions, hold a run for approval before it goes out, and keep an exportable audit log, append-only and hash-chained on Enterprise.

Reports that run themselves · Data rules and security · Rahoto and general AI chat

Fair use

When an AI dashboard is the right tool

• Exploring a question

  Working out what is going on in your warehouse data, one question at a time, is what AI dashboards are built for.

• A live screen for people with access

  When everyone who needs the numbers can sign in and open a dashboard, a dashboard is simpler.

• Use both

  Explore in an AI dashboard. Keep the report that goes out every month, to every audience, in Rahoto.

Limits

What Rahoto does not claim

• The descriptions of Claude Dashboards and the ChatGPT Data agent come from their announcements on the dates shown. “Not described in the announcement” means only that; it is not a finding that the product can’t do it.
• Rahoto has no view that shows the query behind a single number. Authors see where each dataset comes from.
• Sending the last good edition when a source fails is a setting you choose for each publication; it is not the default.
• The text check compares words with the data and holds a run for review; it does not prove a statement true.
• Rules set by your data owners hold everywhere only when their author governs the source; otherwise a rule covers its own report.
• A web link is available on every plan; the other delivery channels come with Pro.
• Runs that only publish a web link skip the text refresh.
• Using your own model is a plan feature, with one endpoint per organisation.