Build and Test a Document Assistant with IBM watsonx.ai | Practical AI training | Applied Academy

FreeBeginner · IBM watsonx.ai · Python · Docling · English

Build and Test a Document Assistant with IBM watsonx.ai

Extract contract terms into JSON with IBM watsonx.ai, flag playbook deviations and compare two prompts against expected answers.

How do you want to learn?

Duration: 3 hours, live
Where: Classroom or remote
Lab setup: Free IBM trial account
Certificate: On completion

The case · ShopNow

ShopNow’s procurement team must check every supplier contract against its playbook.

DocumentsTen synthetic contracts and NDAs

RulesA procurement playbook

You buildA document assistant that extracts five contract fields into JSONwatsonx.ai extracts parties, term, renewal date, liability cap and termination notice, and Python rules flag deviations from the playbook.

DeviationsFlagged with the source clause

PromptsTwo versions compared on held-out documents

IBM watsonx.aiPythonDocling

ShopNow is a fictional retailer; all documents are synthetic.

After this course you can

01Extract five key contract fields into a defined JSON structure using a foundation model.
02Validate JSON output, show supporting clause text and mark missing information without inventing values.
03Use simple Python rules to flag differences from a procurement playbook.
04Measure field accuracy and JSON validity, and compare two prompt versions on the same test documents.
05Explain how scanned-PDF text extraction with Docling can affect the assistant’s results.

Steps

Includes a short break.

Define the contract-review taskReview the five fields, JSON structure, expected answers and playbook rules, and tell missing information from values that need human review.Lab · Inspect sample contracts and define the expected output for a clause with missing information.

Are you ready?

Made forData engineers, software developers, technical analysts and AI practitioners with basic Python and JSON skills who want to build and test a document-processing workflow.

You'll needBasic Python skills: read and edit simple code and run notebook cells.Basic knowledge of JSON objects, fields and values.No prior IBM watsonx.ai experience or legal training is required.

Setup checklistInstructor-ledIBM Cloud trial account; access lasts 24 hours.A computer, browser and internet connection.To join remotely, access to the live online session.

Questions

Instructors

Saleem Pothiwala

Data engineering, AI engineering, strategy

28+ years in data. Built Kermit Tech to help Nordic and Middle East enterprises turn data into competitive advantage. He helps teams build and use data and AI solutions, with hands-on experience across finance, research, and nonprofit organizations. An IBM StreamSets Black Belt, Saleem teaches through practical examples and guided exercises. His courses help learners turn technical concepts into skills they can use at work.

Syeda Eman Fatima

Machine Learning, RAG, Agentic AI

Syeda Eman Fatima is an AI/ML Engineer working across the full machine learning lifecycle. She builds data pipelines as code, with version control.