Cross-silo analytics, with its working shown

Ask a question that no single database can answer

Your finance system knows what you owe a supplier. Manufacturing knows their lead times. Retail knows the return rate. The decision needs all three, and today that means a person and a fortnight.

Aitelico queries each system where it lives, joins the results, and shows you every step it took to get there.

Talk to us See how it works

What it does today

Everything below is built and running. What is not yet built is further down, and marked.

Questions that span systems

Finance, manufacturing, retail and subscriptions stay in their own databases. A question queries each of them and the results are joined in a scratch workspace, so nothing has to be copied into a warehouse first. Each connection belongs to one organization, its credentials are encrypted at rest, and columns can be withheld from the assistant one at a time.

Every step shown

Each query, its source, its row count and the first rows of its result are recorded as you go, and stay with the conversation. You can open any intermediate result in full.

Charts you can interrogate

Results are drawn with Apache ECharts — zoom, change the shape, export a PNG. Axis bounds come from the data, never from the model's guess at them.

Dashboards that recompute

Pin any chart or figure to a dashboard. It keeps the chain of steps that produced it, so refreshing replays that chain against live data instead of showing you a stale picture.

Live external data, feed by feed

Exchange rates, weather and air traffic can join an internal result. Every feed is opt-in, and one switch stops all outbound calls for the whole organization. When a feed cannot be reached the reading is modelled — and labelled as modelled, everywhere it appears.

Rotate the application secret without losing credentials

Stored credentials are encrypted with a key derived from the application secret. Changing that secret no longer means re-entering every password: a management command reads them with the retired secret and rewrites them under the new one, refusing to write anything at all if the secret it was given turns out to be the wrong one.

Your brand, not ours

Colours, fonts, logos and chart palettes are set per organization and apply to the whole product, its charts and its Word exports. Saved themes switch the whole look at once.

The problems we keep hearing about

Companies spend heavily, and wait months, to get dashboards and reports built over their operational data — and the question that prompted the work has often moved on by the time it lands.

These are the three we set out to fix.

The question and the schema sit with different people

The person who knows what to ask rarely knows how another team modelled its data — so every cross-team question turns into a queue.

The incumbents are priced for specialists

Tools like Qlik and Power BI are capable, expensive, and need trained staff to build anything. Data that sits outside them tends to stay outside them.

The data is there and goes unused

Operational teams have rich, current data and no practical way to interrogate it, so assumptions go unchecked and problems are found late.

How it works

No warehouse to build first, and no ticket to raise. The question decides which systems get touched, the person who has the question gets the answer, and the answer carries its own evidence.

How a question is answered Four separate systems are queried where they live; their results are joined in a scratch workspace, which produces an answer with a recorded chain of steps. Finance Manufacturing Retail Subscriptions Workspace joined here Answer chart + figure every query kept openable results recomputes later Systems stay where they are — nothing is copied out.

What happens when you ask

  1. 1

    You ask in plain language

    No SQL, and no need to know which system holds which column.

  2. 2

    Each system is queried where it lives

    Read-only, one query per source, checked against the real planner before a row is read.

  3. 3

    Results are joined in a scratch workspace

    The join happens locally, for this question only. Nothing is written back to your systems.

  4. 4

    You get an answer with its evidence

    A chart or a figure, plus every step that produced it, each one openable in full.

  5. 5

    Keep it, and it stays current

    Pin it to a dashboard and the whole chain replays on refresh, against live data.

Not built yet

On the roadmap, and listed here so you can tell the difference. If one of these is what you need, it is worth a conversation — it moves our order.

  • Planned

    PowerPoint and PDF export

    Word export is branded and shipping. The other two formats are not built yet.

  • Planned

    More connectors

    PostgreSQL today. Which system matters most to you is a useful thing for us to hear.