PDF · DOCX · TextOne row per document, one column per field

Add your documents.
Ask them anything.

Point Document Intelligence at the documents you already have — contracts, resumes, invoices, reports. It reads every one, pulls out the fields you name, and answers your questions with the documents each answer came from.

Try Document Intelligenceis pinned to the top of the page — it stays there as you scroll.

app.futuresmart.ai/document-intelligence

Supplier contracts248 documents · 6 extracted fields · all read

Which contracts renew before October, and what notice do they need?

Four contracts come up for renewal before October. Two of them need notice this month.

Supplier	Renews	Notice	Owner
Northwind Logistics	Sep 01 2026	60 days	Operations
Acme Cloud Services	Sep 30 2026	30 days	IT
Bluepeak Facilities	Sep 14 2026	90 days	Facilities

Answered fromNorthwind_MSA_2024.pdfAcme_CloudServices_SLA.pdfBluepeak_Facilities.docx· open any one to read the exact wording it used

How it works

Add a collection. Start asking.

Your documents live in a collection — one set that belongs together, like every supplier contract or every applicant CV. Each collection is separate, with its own documents and its own fields.

Three steps, none of which is a configuration screen.

Add your documents

Name the collection, then drag in a whole folder, paste text, add a website, or connect a source that keeps itself in sync. Mixed formats, any number, all at once.

Say which fields matter

Name the fields you want from every document — renewal date, owner, total, years of experience. Title, summary and document type come as standard on every one.

Ask, sort, export

One row per document, one column per field. Ask a question and check the documents behind the answer, sort and filter the table, and export exactly what you filtered.

Features

Every answer comes with the documents behind it.

You never have to take an answer on faith. Open the sources and you will see which documents were read, which sections were used, and the original wording — word for word.

Conversations you can follow

Ask in plain language and get a proper answer — with tables and charts where they help. Each question and its answer stay together as one numbered exchange, so a long conversation never turns into a wall of text. Open the sources beneath any answer to see each document’s summary, its extracted fields, and the exact sections that were used. If something is not there, it says so rather than guessing.

Find any document fast

Every document is a row and the fields you chose are columns. Search, sort, filter on several rules at once, hide what you do not need, and export exactly what is on screen.

See what is in the pile

Pick any field — supplier, document type, year — and see every value it takes with a count, then jump straight to the documents behind it. Useful before you commit to a schema.

One question across every file

Batch-run a single question over the whole collection and read the results as one table — one row per document, with the answer and its source on each.

Add more whenever you like

Upload files, paste text, add a website, or connect a source that syncs on its own. New documents join the collection and the originals stay downloadable.

Straight about what it knows

Documents still being read are labelled as such. Fields that are not in a document are reported missing, not invented. Nothing is dressed up.

Also in the boxFields taken per section of long documentsOriginals and plain text downloadableCSV export of whatever you filteredLight and dark

Use cases

The pile someone reads one file at a time.

Each of these is a set of documents a team already keeps, and a question they answer by opening them. The wording below is what you would type.

HR and recruiting

Resumes as a table: years of experience, skills, notice period, location — sortable across every applicant.

Which applicants have five or more years in Python and can relocate?

Everyone who has shipped a data platform, with the line that says so.

Legal and procurement

Contracts with renewal date, notice period, liability cap and owner pulled out of every one.

Which contracts renew before October, and what notice do they need?

Every agreement with an auto-renewal clause, quoted.

Finance and operations

Invoices, claims and policies as rows you can filter, with the original page one click away.

Which invoices over $50,000 are missing a PO number?

Total claimed per vendor this quarter, and the documents behind it.

Sources

However the documents reach you.

Drop in files, paste text, add a website, or connect a source that keeps itself up to date. The same collection is reachable from your own code, so an internal tool can ask the question instead of a person.

Takes in

PDFDOCXPlain textWebsite URLSynced source

Reach it from

PythonJavaScriptcURLChat bubble on your siteFull-page chat link

Generate a key under Share → API. A key can read answers and the documents behind them; it can never change settings.

import requests

r = requests.post(
    "https://app.futuresmart.ai/api/knowledge/v1/71b25bc8-2c76-4028-bb0e-319c35399a33/chat",
    headers={"x-api-key": "<YOUR_API_KEY>"},
    json={"query": "Which candidates can start immediately?",
          "stream": False},
    timeout=120,
)
d = r.json()
print(d["response"])

for source in d["data"]["knowledge"]["relevant_sources"]["data"]:
    print(source)

Your documents

Your collections stay yours, and stay apart.

One collection never mixes with another. Contracts and CVs keep their own documents and their own fields, so what you set up for one cannot spill into the other.

Kept apart

Every collection stands on its own — its own documents, its own fields, its own conversations. Share one with named users without opening the rest.

Originals stay yours

Download the file you uploaded, or the plain text read out of it, at any time. Nothing you added becomes unreachable.

Nothing hidden

Each collection carries an ID you can quote to support or use in your own systems, and an activity trail of who asked what.

Need it inside your own environment? That is a conversation, not a different product.

The answers are already in your documents.

Add a handful of them, ask your first question, and see which document the answer came from. Try Document Intelligence is at the top of the page.

Prefer a walkthrough first? Book a demo.