An AI-assisted enterprise tool that helps operations teams validate document packages, extract key data, detect issues, and resolve inconsistencies with greater clarity and control.
The initial MVP covered the core validation flow, but lacked the clarity, flexibility, and consistency needed for everyday use. Key states and actions were missing, issue recovery was limited, and the interface had yet to evolve into a cohesive JTI product experience.
Reduce the time and cognitive effort required to validate document packages by shifting repetitive checks to AI — allowing users to focus only on detected issues and resolve them by providing the correct documents.
— 60% of vendor data checks automated
— Validation time reduced to 2–5 min
— Scaled adoption across multiple JTI brands and teams
01 Core UX problems
How might we make document handling clear and flexible throughout validation?
How might we make validation outcomes and detected issues immediately understandable?
How might we help users fix an issue and revalidate without restarting the workflow?
02 Define the role of AI
The experience separates what AI does, what the user decides, and what the system must communicate.
03 Validation logic
A single issue can affect the overall status while multiple individual checks still pass.
04 Designing for recovery
When validation detects a problem, users can identify the affected document, replace it, and rerun the current validation without starting over.
04 Final experience
Package status, document-level issues, and individual checks stay visible at the same time.
See the overall result clearly.
One document can surface an issue while successful checks remain visible.
Users can rerun validation after correcting the package, instead of starting over with the same document set.
Before saving the result, users can either accept unresolved issues or upload additional documents and rerun the same validation context.
Fixing one document should not require rebuilding the entire validation flow.
Up to three active validations can run in parallel, helping teams manage multiple company packages at once.
Final takeaway
By making validation outcomes transparent, preserving successful checks, and designing clear recovery paths, the product reduced the amount of attention users had to spend on routine verification while keeping them in control of exceptions.