AI Import

Before you start

Use AI Import when a statement, PDF, CSV or screenshot should become structured records only after review. Import is not an automatic bank or broker connection. Nothing should enter the live portfolio until you review extracted rows, skip duplicates and commit only what you understand.

You are in the correct active portfolio.

The file is clear enough to show names, values, currencies, dates and ownership context.

You know the correct import context: broker, bank, private investment, crypto, real asset, rental or debt.

You are ready to review every extracted row before committing.

You have removed or avoided unrelated private data where possible.

Detailed steps

01

Choose Import statement from the right workspace

Start from Assets, Debts or the relevant Add Capital Item path. Use Import statement when your source evidence is a file or screenshot rather than a single manual record.

Assets onboarding panel showing Add assets and Import statement actions. — Use Import statement when your best evidence is a file, screenshot or statement.

02

Select the import context before uploading

Choose the context that matches the file. Broker holdings, bank statements, debt statements and private investment reports need different review logic. The right context reduces cleanup later.

Add Capital Item modal with Public Markets selected and Import Holdings visible. — Choose the import context before uploading so rows are interpreted correctly.

03

Upload a clear source file

Upload a PDF, CSV or screenshot that shows the important fields. If a screenshot cuts off currency, account, balance or date, extraction may be incomplete. Split unrelated documents instead of uploading a mixed file when possible.

AI Import modal for stocks and fund holdings with Upload file action. — Upload a clear PDF, CSV or screenshot that includes names, values, currency and dates.

04

Wait for the editable preview

The preview appears before anything is committed. Treat it as a review workspace, not as finished data. Read the row count, included rows and warnings before doing anything else.

AI Import preview showing extracted public market rows before commit. — The preview appears before anything enters the live portfolio.

05

Review each extracted row

Check detected item, row type, liquidity, amount, currency, source and confidence. Compare each row against the source. Correct obvious extraction mistakes and skip rows you do not understand.

Editable import preview table showing included row, detected item, type, liquidity, amount and currency fields. — Review row type, included state, amount, currency, liquidity and confidence before commit.

06

Exclude duplicates or irrelevant rows

Use exclude or skip for duplicate holdings, zero-value rows, irrelevant items or records that belong in another portfolio. Excluded rows should not become portfolio data.

AI Import preview after one extracted row has been excluded. — Exclude duplicates or rows you do not understand before committing.

07

Treat balance rows and transaction rows differently

For bank imports, a balance row updates the cash or account record. Transaction rows explain movement history. Do not count transaction rows again as separate assets.

Review imported bank statement modal showing balance row, transaction rows and warnings. — Bank imports can include a balance row plus transaction history; review both differently.

08

Commit only reviewed rows and verify the result

Commit only after row type, amount, currency, date, included status and duplicates are reviewed. After commit, open Assets, Debts or Transactions and verify the created records.

Assets screen after importing a public markets broker statement. — After commit, verify the created records in Assets, Debts or Transactions.

Field guide

Import context

The workflow used to interpret the file.

Choose broker, bank, debt, crypto or another context before upload.

Detected item

What Capitelist thinks the row represents.

Compare it with the source before commit.

Row type

Asset, liability, transaction, balance or another extracted type.

Wrong type creates wrong downstream behavior.

Included

Whether the row will be committed.

Exclude duplicates and rows you do not understand.

Confidence / warnings

Extraction quality indicators.

Low confidence requires review, not blind acceptance.

Amount and currency

The value and denomination extracted from the file.

Check signs, symbols and currency carefully.

Commit

Moves reviewed rows into live portfolio data.

Use only after all included rows are correct.

Decision rules

What it adds

Speeds up onboarding from PDFs, CSVs and screenshots.
Keeps a human review step before anything enters the balance sheet.
Separates imported capital rows from transactions when possible.

What it does not do

AI extraction can misread documents and must be reviewed.
Poor scans, missing pages or ambiguous screenshots reduce confidence.
Import is not a live bank or broker connection.

Common mistakes

What to check

Never commit rows you do not understand.
Check currencies, signs and duplicate holdings carefully.
Use warnings and low-confidence flags as mandatory review cues.

Committing rows without reading the preview.

Using the wrong import context.

Uploading cropped screenshots that hide currency, date or ownership.

Accepting duplicate holdings from repeated statements.

Treating transaction rows as new cash balances.

Ignoring low-confidence warnings.

Committing rows into the wrong active portfolio.

After you save

Open the destination table and verify created records.
Check Net Worth when imported rows affect assets or liabilities.
Open Capital Review for low-confidence, stale or unclassified records.
Archive or correct duplicate rows immediately.
Keep the source document available when values are material.
Use manual entry when a source file is too ambiguous for reliable extraction.