LoginSign Up
← Back to all sparks
B

Basedash

ANALYTICS
Velocity7.5

AI-powered business intelligence and database interface with chat-based data analysis, dashboards, and automations

www.basedash.com ↗

Basedash is done answering questions about your data — it now wants to tell you what to do next.

ai-analystprescriptive-analyticsembedded-bienterprise-controlsdeveloper-platformscheduled-reporting
◆Current state
Basedash spent July and early August building the surfaces of an AI-native BI tool: suggestions that propose questions before you type, subscriptions that push dashboards to Slack and email, audit logs that record every query the AI runs, and a developer platform exposing the whole feature set through an API. Tasks, now in research preview, changes the output shape entirely — instead of charts and answers it produces a ranked list of work with a stated rationale and expected outcome, then watches whether the metrics move. A sidebar rebuild the day before quietly names the product's five pillars: Chat, Dashboards, Automations, Insights, and Data.
◆Where it's heading
The arc runs from self-serve querying toward prescription and closed-loop measurement. Each release chips away at the assumption that a human must decide what to look at: suggestions removed the blank prompt, subscriptions removed the visit, and Tasks removes the interpretation step. The navigation rework is the tell that this is now a multi-module product rather than a chat box with extras — and the enterprise scaffolding arriving alongside it, audit logs covering AI queries plus retention controls, is what makes an autonomous analyst deployable rather than a demo.
◆Prediction
Tasks graduating from research preview will be the release to watch; the outcome-tracking loop it describes only has value once it has run long enough to show whether its recommendations worked. Expect Tasks to become a sixth sidebar module and to be exposed through the developer platform API, since that is where every other Basedash capability has landed.

◆Recent moves

  1. 4d ago

    Introducing Tasks: your operations, on autopilot

    ⚡ SPARK

    Tasks is the logical endpoint of the suggestions-and-automation work of the past month: rather than proposing what to ask, it proposes what to do, and then measures itself against the result. It moves Basedash from the analysis side of the loop to the operating side.

  2. 5d ago

    A sidebar that follows what you’re working on

    The sidebar now anchors to five top-level modules — Chat, Dashboards, Automations, Insights, Data — with the list beneath it swapping to match the active one, and favorites pinned across all of them. Shipped the day before Tasks, it reads as structural preparation: a product that had grown from a chat box into five surfaces needed navigation that admits it.

  3. 11d ago

    Introducing Basedash Subscriptions

    Subscriptions productizes the scheduled-delivery work started in early July: any dashboard or chart can now push a snapshot to email or Slack on a chosen cadence. It fits the pattern of removing the visit — the report arrives instead of waiting to be opened.

  4. 12d ago

    Sort and arrange tables without changing the chart

    Sorting, hiding and reordering columns now persist to a personal view rather than forking the chart, and are applied as a validated query change instead of rewriting the author's saved SQL. It is a small change with a real multi-user consequence: exploration no longer costs the chart's canonical definition.

  5. 18d ago

    Introducing Basedash audit logs

    Native audit logs cover access, queries and configuration changes, explicitly including every query the AI itself runs, with export and retention controls. This is the compliance groundwork an autonomous analyst needs before a regulated buyer will let it near production data.

  6. 19d ago

    MotherDuck is now a supported data source

    MotherDuck joins the supported data sources via its Postgres endpoint, queryable alongside existing databases and available to AI chat. Source breadth is the quiet prerequisite for the prescriptive work — recommendations are only as good as the share of company data the analyst can reach.