Technology
Widgets
An answer is not always a paragraph. A widget is the surface Scalata answers in — a workbook, a feed, a portfolio, a document — built live from your own systems and open to inspection rather than handed over as a picture.
Table AI
A workbook that understands the model inside it
A spreadsheet in the shape you already build in — several sheets, formulas that reach across them, and charts sitting on their own tabs. Drivers on one sheet, the build-up on the next, the summary and its charts after that.
What is different is what sits behind the cells. The workbook is wired to the systems your numbers come from, so a model is refreshed each period rather than rebuilt — and it can be asked, in plain language, to extend itself. Formulas stay live: change an assumption and every dependent cell, on every sheet, recalculates. Ask what feeds a figure and Table AI names the cells, the sheets and the source rows behind it.
Document AI
A document you edit by asking
A real document — title, sections, figures, citations — except nothing is typed into the page. Every change is a request: add a section, pull the numbers, redraw the chart, restyle the whole thing. Word, PowerPoint, PDF, Markdown: the format is an output, not a different tool.
It is built in sections rather than as one continuous page, so an edit lands where you meant it and the rest is left alone. Ask for something the document does not yet know and it searches the web, writes the section and leaves the citations attached to the sentences they support. Ask for a different look and the theme, headings and body type change across every section at once — no find-and-replace, and no reformatting afterwards.
The same work comes out in whatever shape you need it. Ask for a deck and the sections become slides; ask for a PDF, a Word file or Markdown and you get that instead — with the figures crossing over as native charts rather than pasted images, and the theme carrying with them. It reads the same formats going in, so an existing report or deck arrives as sections you can edit, question and rebuild rather than as a file you have to start again from.
App AI
Ask for an app, get one running
Describe the app your team keeps asking for and it is generated, wired to the systems that hold the data, and deployed — a real application with screens, filters, permissions and actions that write back, not a dashboard you can only look at.
It builds against your connectors rather than a copy of your data, so the app reads what your systems hold now and its actions land back in them: an approval updates the record it came from and posts where your team is already watching. Access follows the permission model the rest of Scalata uses, so the tool cannot reach further than the person using it.
And it is not a black box. Every app has source you can open, read and change — edit it in place, or take it away. Ask for a change in plain language and it is made, redeployed and versioned, so the fastest way to build an app is not also the last time anyone can maintain it.
Newsfeed AI
Ask what happened, get the stories
Ask about a company, a ticker, a market, a regulation or a competitor and the answer arrives as a feed of stories rather than a paragraph about them — each one sourced, timed, and open to read in full.
A summary you cannot check is worth less than an article you can. Every card names where it came from and when, and opens into the whole piece — with the figures taken from the filing or release behind it rather than restated at second hand. Nothing appears without a source attached to it.
It is a feed of your things, not a category. It follows the tickers, suppliers, counterparties and rules you actually care about — including the ones it can see in your own systems — so a story reaches you because it moves a number you track, and it tells you which one.
Grid AI
Put the documents side by side
Attach a stack of contracts, filings, policies or reports — in whatever languages they arrived in — and get a grid: one column per document, one row per thing you need to compare, filled in and cited.
Language is not a separate step. A German framework agreement, a Japanese services contract and an English master agreement are read as they are and answered in yours, so a set of documents nobody on the team can read end to end still becomes one table you can act on.
Every cell keeps the page it came from. Open one and the clause itself appears — in the original first, then in translation — because a comparison you cannot trace back to a sentence is a summary, and nobody signs off a contract from a summary. Change the criteria and the grid is rebuilt against the same sources.
Portfolio AI
Change the book by describing the change
Your holdings, live, with the analysis around them — and a way to alter them that is a sentence rather than a screen full of forms. Trim a position, move a sleeve, model a hedge, add the positions sitting in a file someone just sent you.
It works from whatever the answer needs: prices from a market-data connector, positions from your custodian, a spreadsheet dropped straight into the conversation. Ask a question the current data cannot answer and it goes and gets it rather than telling you it cannot.
Nothing executes on a sentence. A change comes back as a proposal — the before, the after, and the orders it would take — and stops there until someone approves it. Every figure keeps the mark and the timestamp it was priced from, so a decision can be explained afterwards as easily as it was made.
Turn weeks of expert work into minutes
See how Scalata fits your team, your data, and your controls.