← Learn
Learn · AI

Working with financial documents

Seven lessons on building AI systems that read credit agreements, financial statements, appraisals and rent rolls: getting clean text from PDFs and scans, extraction schemas with checks that make the numbers add up, a source for every value and sentence, drafting memos from templates, answering due diligence questionnaires from a library of approved answers, and running it all with audit trails and a person signing off. Each lesson has a document to work on: tap the passage behind a value, or the line that fails a check.

Take first: How LLMs work, Retrieval and RAG, Agents and workflows.

  1. The documents

    Credit agreements, appraisals, financial statements, rent rolls, invoices and operating statements, and what makes each one hard to read.

    6 questions

  2. Getting clean text

    Native PDF text or OCR, layout and tables, page and position metadata, and handling a file of hundreds of pages.

    6 questions

  3. Extraction schemas

    Designing the JSON for a document type: field types and units, validation rules, and what to do with missing or conflicting values.

    6 questions

  4. Provenance and sources

    Storing the page, section and text span for every value and every generated sentence, showing them, and checking each one holds up.

    6 questions

  5. Generating memos and reports

    Template-driven drafting, each section from its own retrieved facts, numbers computed in code, and edit workflows for reviewers.

    6 questions

  6. Questionnaire automation

    DDQs, RFPs and security questionnaires: an answer library, retrieving and adapting answers, stale and contradicting answers, and approval.

    6 questions

  7. Running it in a regulated setting

    Audit trails, access control, personal data, human approval gates, evaluating on labeled documents, accuracy per field and cost per document.

    6 questions
More info

How it works

Each lesson is a short read with examples, then a few practice questions. Some ask you to type an answer (a number, an address, what a piece of code prints), some have you pick one, and some have you tap steps or lines into the right order. You can try again as often as you like; after two misses there’s a button to show the answer.

What do I need to know first?

How LLMs work, Retrieval and RAG, and Agents and workflows, or the same ideas from your own work: prompts, JSON output, embeddings and retrieval, and a person in the loop. No finance background needed; the documents are explained in lesson 1. The examples are short Python you only need to read.

The other AI and data engineering courses

Where is my progress saved? In this browser, like the rest of the site. Nothing to sign up for.

“Python” is a registered trademark of the Python Software Foundation. This course isn’t made by or affiliated with it.

For learning only, not legal, compliance or investment advice. Every company, document and number in it is made up. Rules on records, privacy and model use differ by country and by firm; check yours with your compliance team.