"Talk to your data" usually fails on trust, not vocabulary. A chatbot that answers "you sent 12,480 messages" is only useful if you can check the number against the dashboard beside it, and most natural-language analytics tools make that check hard: the answer comes from a separate query, a separate aggregation, or a model's guess, and the reader has no way to tell which. Devotel Orbit ships a different shape of the feature. Ask your data, inside Insights → Analytics, answers plain-English questions against the exact series the page already charts, so every sentence is checkable against the cards and charts around it.
What Ask your data is
Open Insights → Analytics and the Ask your data panel sits below the KPI cards. Type a question about your own messaging, voice, or contact metrics and the panel answers in one sentence over the date window you asked about, with a compact trend chart when the metric has a daily series to chart.
Three properties define the surface:
- It reads the same snapshot as the page. The answer is computed from the same tenant-scoped metric snapshot that backs the KPI cards next to the panel. The sentence can never disagree with the figures around it, because it is the figures around it.
- It is deterministic, not generative. Questions resolve through a fixed phrase catalogue to one metric, one window, and (when named) one channel. The same question always returns the same metric and the same phrasing, and nothing about your data is sent to a language model or a new API surface.
- It names what it answered. Each answer states its metric and its window, and adds the channel when a named channel is honoured, so “you sent 4,120 messages over the last 30 days on Viber” is a claim you can verify on the page, not a chatbot guess.
The panel covers nine metrics across the platform: message volume, delivered messages, failed messages, delivery rate, calls, talk time, voice spend, active contacts, and total spend.
Per-channel answers you can audit
Since October 2026, a question can name the channel it cares about. Ask "how many Viber messages did we send?" and the message-volume answer filters to Viber's own column in the same daily per-channel series the page's big volume chart renders. The filter works for all six channels the page tracks: SMS, WhatsApp, RCS, Email, Viber, and Telegram.
The audit story is the point. When a channel is named:
- The sentence names the channel. The answer reads "… on Viber" rather than silently answering the all-channels total. Before this shipped, a channel-named question returned the unfiltered figure with no signal the filter had been dropped.
- The chart re-renders to the channel. The governed mini area chart under the answer rebuilds from that channel's column of the same daily series, so the trend under the sentence is visually checkable against the page's own volume chart.
- The filter is only applied where it can be proven. Only message volume has a daily per-channel series on the page. Ask for a channel alongside a different metric (say, delivery rate) and you get that metric's all-channels figure with the channel left out of the sentence — the answer never implies a filter the data cannot support, and never claims a slice it cannot show.
That last restraint is what separates auditable analytics from a chatbot guess: the panel would rather answer the plain all-channels figure than print a channel-qualified number it cannot corroborate.
Worked examples
Four questions, and how each resolves:
"How many Viber messages did we send this month?" Metric: message volume. Channel: Viber, honoured and named in the sentence. Window: the question's "this month" pivots the page's date filter to the 30-day preset, and the answer and the mini chart both render over Viber's daily column for that window. Check it by switching the volume chart to Viber and reading the same total.
"How many messages did we deliver this week?" Metric: delivered count, resolved from the verb "deliver" rather than falling back to overall volume. Window: 7 days. The answer reports the delivered figure for the week, distinct from the submitted total a lazier match would have returned.
"What's my deliverability today?" Metric: delivery rate, as a percent over the last 24 hours. Delivery rate has no daily series on the page, so the answer stays text-only. The panel only charts numbers the dashboard already shows per day, and says so by omission elsewhere.
"How much did we spend on voice over the last 20 days?" Metric: voice spend, in your billing currency. Window: 20 days is not one of the page's presets, so the panel clamps the filter's custom calendar range to that span instead of snapping to the nearest preset and mislabelling the answer.
Questions that name no recognisable metric fall back to overall message volume with a "name a metric" hint beside the answer, so you always know when the panel guessed.
Where it sits in Insights
Ask your data is the exploratory half of the Insights workflow; the push half already ships beside it. Scheduled reports email a fixed digest of volume, deliverability, top contacts, queue performance, or spend to up to 50 recipients on a daily, weekly, or monthly cadence, so the Monday review reads one page. Ask your data is the tool for the moment a digest raises a question: open Insights → Analytics, type the question, and read the answer against the same series the digest summarised.
For the full mechanics, the docs guide Ask your data — natural-language questions on Analytics lists the answerable metric catalogue, the window phrases the parser recognises, and the resolution order when a question matches more than one metric. The companion guide to the page itself, Insights analytics, covers the charts and KPI cards the panel reads from.
Frequently asked questions
Does Ask your data send my questions or my data to an AI model?
No. Questions resolve in your browser through a fixed phrase catalogue against the metric snapshot the Analytics page has already loaded for your tenant. There is no language-model call, no third-party request, and no new data surface.
Which channels can I name in a question?
SMS, WhatsApp, RCS, Email, Viber, and Telegram, the six channels the Analytics page tracks. A named channel filters the message-volume answer and its chart; ask for a channel alongside a metric without a per-channel daily series and the answer simply reports the all-channels figure, with no channel in the sentence to imply a filter that was not honoured.
What happens when my question doesn't match a metric?
The panel answers with your overall message volume and shows a hint that it wasn't sure which metric you meant. Naming a metric (messages, delivery rate, calls, talk time, contacts, or spend) gets you the specific answer.
What if I ask about a window the page can't represent?
The selected window stays as it is. The parser only picks windows the page's filter can express (today, week, month, quarter, year, or an explicit day count clamped to the custom range), so it never snaps to a misleadingly different preset to look helpful.
Is the chart under an answer computed separately from the dashboard?
No. It is built from the same daily per-channel series as the page's message-volume chart, summed per day, and re-rendered to a single channel when the question names one. A number the big chart cannot corroborate never appears in the small one.