Contract AI
Summary
Using AI enhanced document parsing service (DPS), we simplified the process of adding complex contract data into Procurify, providing in-context data that will help organizations make more strategic purchasing decisions.
Company
Procurify
My Role
User Research, UX Design, Usability Testing
Timeline
Q1 2024
The Challenge
Procurify’s Contract Repository was developed as a way for clients to store their service contracts and other contract related documents directly within Procurify so that they could be more easily used to make strategic spending decisions. Launched in Q3 2023, after a quarter of GA availability, its adoption remained low. Our team’s goal was to understand why and explore how we could support our clients in using this feature.
8%
The percentage of clients who had uploaded contracts to the newly created contract repository.
5
The number of strategic clients who have identified the manual effort of creating contracts as a blocker to adoption.
ProcurifyAI
Procurify’s Senior Leadership Team had challenged the Product group to find useful ways to integrate AI into the product
Project Goals
Increase the utilization of the contract repository (8% pre-build)
Decrease the time required to add a contract
Pioneer and build best practices around using LLMs to extract document information
Research Goals
- Understand how different types of contracts are formatted to support training our LLM
- Understand how our clients are currently storing their contract files and data
- Understand how clients who have adopted the new contract management features are using them
- Discover how company’s interact with contracts on a day-to-day basis
Research Insights
- Through our analysis, contracts could be as small as one or two pages for a SAAS agreement or as long as 30+ pages for a more complex construction contract
- The most common way we found that clients were storing contract details was in spreadsheets. While this achieved their goal of having them all in a single location, it made managing them a very manual task
- We found a number of cases where clients had started using the contract repository as a vendor document notification tool, using the expiry reminder for all sorts of documentation
User Flow Ideation
By involving stakeholders from Product, Engineering, and Integrations early on we were able to understand the ways in which our application would interact with the data parsing service (DPS)
This made it much easier to identify edge cases and determine how we were able to support users through those scenarios.
Wireframe Exploration
While flow diagrams were helpful in collaborating with more technical stakeholders, moving to low fidelity wireframes made it easy to build alignment with senior design and product stakeholders.
Once we were confident in the established flow, I teamed up with our partners in CX to tap into their user empathy and understanding.
Early Feedback & Insights
AI Language
As a new method of creating content in Procurify, we wanted to ensure users would clearly understand how it works.
Communicating feature limitations
The first iteration would only support certain file types and be able to process a certain number of files at a time.
Supporting related contract files
As a new method of creating content in Procurify, we wanted to ensure users would clearly understand how it works.
Solution
- Allow users to upload one or multiple contracts without adding additional complexity
- Inform users of when and where assumptions were being made
- Allow users to review contract data before making them active
Beta Feedback
Set it and forget it
AI isn’t that scary
Great! (For new contracts)
Success Targets
Utilization
The AI auto-fill functionality should be used to create 20% of all new contracts.
< 1 quarter since delivery. Target currently tracking at 24% (4% greater than expected)
Adoption
The contract module should be adopted by 16% of domains (an increase of 100%) within 3 months.
< 1 quarter since delivery. Target currently tracking at 11% (62% short of target)