Analytics Answer Desk with Metric Citations

In a data and analytics team, the ask-data channel fills with the same questions every week: what were returns last week, how much stock is on hand, what is the margin on one channel. Answering by hand means finding the agreed definition, the right console and the latest load, and a figure from a stale source can spread before anyone notices. This Botlit concept shows an answer agent for that channel, designed to answer routine questions from SemanticFed's published metrics and BigConsole consoles, to say where every figure came from and how fresh it is, and to hold back when the data is not ready.

The illustration opens the Conversations view of an Ask Data agent for a sample company, badged active, with a badge for the team's ask-data channel and a Metrics: asker's access badge. The design envisions metric answers running through SemanticFed's governed model under the asker's own access policy, while answers read from a console follow that console's sharing. Asked about returns at a group of stores, the agent gives a returns figure and its share of net sales against the week before. Three numbered citations sit beneath: two SemanticFed metrics from a versioned retail model and a BigConsole console with its region filter, beside a Data as of badge. A Sources for this answer panel lists the metric with its model version and publish date, the console with its widget count, and a returns source covering all channels with its last refresh time.

The next two questions show the agent declining to guess. Asked about tent stock, it replies in an amber card that the inventory source is late against its four-hour target and that it will not quote stock until the source refreshes, flagged by a BigConsole freshness board. Asked about margin through wholesale partners, it says no published metric covers that yet; a No matching metric badge sits above a line showing the question was filed as a numbered request in PlanMagnet through FluidGrids. A Today card counts the day's questions, those answered with citations, those held for late data and those sent to the data team.

In the design, Botlit handles the conversation, SemanticFed supplies the agreed definitions, and BigConsole supplies consoles and freshness signals. A FluidGrids workflow is designed to turn questions no published metric covers into PlanMagnet requests, so the data team sees demand instead of losing it in chat. Each answer is designed to be kept in Botlit's execution records with its tokens and cost. The concept is meant for analytics and BI teams, metric owners and the colleagues who bring them questions.