Measure success of AI models

Problem

Eigen needed to roll out new confidence models within its B2B document extraction platform.

The challenge:

• Integrate these models without disrupting existing user workflows

• Ensure users understood and trusted the updated confidence scores

• Drive adoption of the new precision settings and reduce reliance on manual review

Solution

Led the UX strategy, including:

• Facilitated ideation and user journey mapping to identify key touchpoints for explanation and interaction

• Designed a step-by-step visual guide that clarified:

  • How confidence scores were calculated

  • The relationship between target precision settings and model output

  • When and why to adjust those settings

• Integrated the explanation seamlessly into existing workflows without creating friction

Conducted validation sessions with internal teams and key users to refine clarity and usability.

Outcome

• Lowered cost of operations by reducing need for human QA

• Supported improved model performance visibility and customer satisfaction

• Helped scale the platform with less dependency on manual intervention