About the project:
DataAI is a data assistant for property developers. Built entirely by our team.
The product’s purpose: owners and executives should get an answer about their own figures straight away — not through an analyst and not through a data dump into a spreadsheet. The system sits on top of a developer’s operational data — both sales and construction — and answers a question asked in plain words: it calculates a metric, builds a table, draws a chart.
This is a service for many companies at once: each has its own data, its own users, its own plan. The interface is translated into four languages, and amounts are calculated in seven currencies.
A question in words instead of a request to an analyst:
The usual path from question to figure looks like this: an executive frames the task, an analyst pulls the data, assembles a table and sends over a file. A day or two passes — and the answer is already about the past.
Here the question is asked right inside the system. “Revenue for March on closed deals”, “top 5 managers by number of deals for the quarter”, “warehouse stock by item” — in reply comes a short text with an explanation, a metric card, a table or a chart.
You can ask in four places, and these are different scenarios rather than one button in different wrappers: a separate chat across all the data; an assistant panel right on the dashboard — it already knows which period and which widgets are currently on screen; a chart builder that assembles a new widget from a description; and the report wizard covered below.
You don’t have to hunt for a chart in a list. It’s enough to describe it in words — “deals by manager for February” — and the assistant clarifies the metric and the chart type, shows the result and offers to pin it to the dashboard.
An answer doesn’t vanish — it becomes a report:
The key difference from a “chat on the side” is that a good answer here doesn’t disappear along with the conversation.
It turns into an AI report. The wizard walks through four steps — goal, table structure, metrics, preview — and the result is a permanent report: it appears in the side menu next to the built-in ones, recalculates when the period changes, exports to CSV and Excel, and keeps its own description — what it counts, what a single row means, which filters are applied.
You don’t have to start from scratch: the wizard holds 37 ready-made templates — 21 for sales and 16 for construction. You take a template as a starting point, adjust the wording to fit and send it off.
As a result the analyst drops out of the process not for a single question but for good: a report that used to be assembled by hand every month now lives in the menu and updates itself.
Dashboards built for the task:
There isn’t a single dashboard for everyone here. There are several — main, sales, marketing, construction — and each is assembled from blocks to suit its role.
The blocks come from a catalogue: deal and revenue trends by month and by day, average deal size, a breakdown by manager, plan versus actual, lead sources, payments by building. They are placed with the mouse, rearranged, removed. Alongside them — the ones the assistant assembled.
Every metric has a definition: exactly what went into the figure and which date it is counted by. It sounds trivial right up until you discover that two people in the same company understand “sold in February” differently.
Why the figures disagree:
The most common complaint about any analytics is “your report says one thing and the system says another”. Usually that’s met with silence. We took the causes apart right inside the product.
The interface has a section that shows what makes up a precise question — period, metric, breakdown, filters — and picks apart the typical traps of language. For a developer, “contract” means both the sale of an apartment and a works contract with the construction side. “Sold” is either the total value of deals or the money that actually reached the account. “For February” is the deal date, the payment date or the lead date. Next to each trap is a wording that removes the ambiguity.
A company can add its own terms too: if it counts a “qualified lead” by a special rule, that rule is written into the glossary — and the assistant counts it the same way.
For a reporting system this is a rare honesty, and it pays off: the conversation shifts from “your figures are lying” to “we were asking about different things”.
Sales and construction in one window:
A developer’s business has two halves, and they usually live in different systems: sales in one, construction in another. Here they sit side by side, in one interface and over one period.
On the sales side — the pipeline of leads and deals, day-for-day period comparison, a manager ranking, units on sale, contracts signed. On the construction side — works contracts, acceptance certificates, the work schedule, the warehouse.
Accounts receivable has been taken all the way to practice: not “how much is owed in total” but a list of payments — the scheduled date, the unit, the amount, the overdue balance as of today. From a list like that you can see straight away who to call today. For a client, a reconciliation statement is assembled: plan, actual, balance.
A small touch that says a lot about the care behind it: in period comparisons the colour is worked out by meaning — a rise in revenue is green, but a rise in debt is red, even though formally that’s a rise too.
Budget, access and languages:
AI costs money, and the product doesn’t hide it. A company has one shared budget for all AI features — chat, reports, widgets — with a visible counter: how much has been spent, how much is left, when it resets, which actions draw it down and which are free. Not “unlimited until the first invoice”, but a figure you can see before it runs out.
Access is granted by roles: owner, member, viewer. A dashboard is opened to a colleague by invitation, and shared outward with a link and a passphrase — the recipient doesn’t need an account.
The interface is translated into four languages and amounts are calculated in seven currencies: the product is built for more than one market.
Result:
The owner gets an answer about their own figures themselves and within a minute — not in a day and not through an intermediary. A good question turns into a permanent report that then lives on without anyone’s involvement. Sales and construction are viewed in one window, every figure has a definition, and discrepancies with the operational system are explained by the product itself, instead of being left as grounds for an argument.
The system serves several companies at once: each has its own data, access rights, plan and AI budget.