Story beat 01 of 7: Hook
An autonomous data analyst · Proof of concept
Dataloqui is an AI data analyst for teams that don't have one. It connects to your databases, spreadsheets and business systems, notices what changed, works out why, and tells you before you think to ask. In plain words, in your language, typed or spoken.
Mon 08:00 · Morning briefing
Nobody asked
Good morning. Three things since Friday.
Act today
94% sure
Wireless earbuds will sell out in about 3 days.
Demand is up 71% in two weeks and your supplier needs 5–7 days. Reorder today to avoid a gap.
Watch
88% sure
Imported goods just got about 2.8% pricier.
The dollar rose 3.2% this week. Electronics margins are now below 5%.
Good news
97% sure
Online orders keep climbing: 34% → 41% of sales.
Four weeks in a row. Mobile accounts for 70% of the growth.
The problem
Most businesses collect plenty of data, then decide on gut feeling. Not for lack of caring: nobody has the hours to question every chart. A chart shows you a line. It can't tell you what happened, why, or what to do about it.
Revenue by store · weekly · k€
Same twelve weeks, two views. Dataloqui greys out what doesn't matter, highlights what does, and writes the explanation onto the chart itself, following the principles of Storytelling with Data (Knaflic, 2015).
How it thinks
Language models write well and count badly. So Dataloqui splits the job: statistics find the signal and produce every number, and the language model only turns those verified numbers into words.
Statistics engine · output
What you read
When the dollar strengthens, orders for imported products drop about three days later. The link is strong (−0.73) and has held for six months. The dollar rose 3.2% this week, so expect imported orders to fall 10–15% by the weekend. This is a correlation, not proof of cause. Confidence: 86%.
Hover or tap an underlined phrase to see the number behind it.
Statistics Language model
01Statistics
Connect
Databases, APIs, spreadsheets and ERP data land in one unified layer, cleaned and lined up on a shared timeline.
02Statistics
Watch
Scheduled scans compare every metric with its learned normal, so seasonality and holidays don't cry wolf.
03Statistics
Compute
Statistics do the finding: z-scores and IQR for anomalies, seasonal decomposition, change points, lagged correlation across sources.
04Language model
Narrate
The language model turns verified results into a story: what happened, why it matters, what to do next.
05Statistics
Check
Every claim is checked against the numbers. If the text says “rose” and the data says “fell”, the text is rewritten.
How it’s built to behave
Autonomous
It scans your sources around the clock and ranks what it finds by size, reach and urgency. You get the three findings that matter, not three hundred.
14:06 · nobody asked
Dataloqui:
Returns jumped to 12% in the last hour, three times the usual rate. 80% are one product, and the reviews mention damaged packaging. Worth a word with the warehouse.
Conversational
Ask in plain language, typed or out loud. It keeps the thread, so a follow-up like “and by region?” just works. No SQL, no ticket to the data team.
You:
spokenWhy are online orders growing?
Dataloqui:
Mostly mobile. It drove 70% of the increase. Want it by region?
You:
spokenYes. Just last month.
Dataloqui:
The coast leads at +12%. I put the chart on your screen.
Learning
Three layers of memory: this conversation, past ones, and what it has learned about your business. Correct it once and it stays corrected.
Dataloqui:
Feb 14: orders up 300%. Looks like an anomaly.
You:
That's Valentine's Day. Happens every year.
Dataloqui:
Noted. I'll treat it as normal from now on.
One year later
Dataloqui:
Orders up 240%, the usual Valentine's spike. But that's 15% fewer than last Valentine's, which is worth a look.
Voice
Ask out loud, the way you would ask a colleague. Dataloqui answers in plain words, turns the numbers into a short story, and says it back. On a call, in the car, across the shop floor.
Listening…
You, out loud
How are sales doing today?
Dataloqui
Up 12% on yesterday. One thing stands out: electronics orders jumped from about 50 to 127.
You, out loud
Why did that happen?
Dataloqui
Two likely reasons: a competitor ran out of stock, and the dollar dipped 1%, so your prices look sharper. At this pace you sell out in three days. Want me to plan a reorder?
A preview of voice mode, which is on the roadmap: speech in, speech out, and a chart on your screen when a picture says it better.
Languages
Dataloqui writes each story directly in the language your team works in, and gets the small things right: where the percent sign goes, how digits are grouped, which way the sentence runs.
Sales fell 18% last week, almost all of it in one store. Foot traffic held steady, but your best-seller has been out of stock since Tuesday. Restocking it should win back about €15,000 a week.
Ten languages shown here; the language models underneath cover many more.
Fits in
No migration, no new warehouse. Dataloqui reads from the systems you already run, lines everything up on one timeline, and finds the stories that only show up when sources meet.
Many sources in · one story out
Databases × Sales & POS
Harbor lost 18% in week 9. It’s a stock problem, not a demand problem.
Each connector is a short config file. Add one and Dataloqui discovers the schema, works out what the columns mean, and asks you only about the ones it can't tell.
connector: name: Central bank FX rates type: rest_apiconnection: base_url: https://api.example-bank.org auth: { type: api_key, value: ${FX_KEY} }sync: schedule: "0 10 * * 1-5" # weekdays, 10:00metadata: freshness: daily reliability: 0.95Industry profiles tell it which metrics matter, what “normal” looks like, and which dates are special, from Valentine's Day in retail to maintenance windows on a factory floor.
Trust
An analyst you can't trust is worse than none. These are the house rules every story follows.
Every finding carries a confidence score, and doubt is said out loud.
“I'm about 60% sure. Last year's returns data would settle it.”
“These move together” and “this caused that” are different claims. It only makes the first unless the data supports the second.
Every number traces back to its source and to the query that produced it. Ask, and you get the raw rows.
A briefing that only ever warns is a briefing you stop reading. What's working is part of the story.
It can run entirely on your own servers. Personal data is masked before analysis, and nothing is sent to third parties.
Where it stands
Dataloqui is a working proof of concept under active development. Here is the plan, in the order it's being built.
Working
A retail pilot that joins a sales database, a central bank exchange-rate API and a CSV budget. It catches stock-outs, currency effects and sales trends, and turns plain questions into SQL on real schemas.
In progress
Connectors, the unified data layer, chat, and the anomaly and trend engines.
Planned
Agent orchestration, autonomous monitoring, cross-source analysis, confidence scores.
Planned
Three-layer memory, learned preferences, baselines that adapt.
Planned
Speech in, speech out, interruptions handled, spoken alerts.
Planned
Industry profiles, ERP connectors, guided onboarding, email and Slack briefings.
Early access
We're looking for early partners: teams that have the data and the questions, but not the analysts. Tell us what you'd want your data to tell you, and we'll show you where Dataloqui is today.
Built by
Creator of DataloquiBackend Developer & AI Integrator