Narvar’s 10,000+ Contract Migration Made Possible with Coheso

Coheso is an AI-native legal front door and work management platform. Most teams know it for intake, triage, and structured workflows. Narvar, a post-purchase ecommerce platform trusted by more than 1,500 global brands, saw a deeper capability inside the platform. They recognized that Coheso’s QnA engine, which enables business self-service and agent-powered legal intake, could also help them migrate more than a decade of contracts from Salesforce into a new Contract Lifecycle Management (CLM) system.

Over a Decade of Contracts and a Four-Person Team

When Jason Sokel, Senior Corporate Counsel at Narvar, first assessed the work ahead, the scope was significant. The company needed to review, extract data from, and validate more than 10,000 commercial contracts that spanned over ten years.

Narvar’s legal team supports more than 74 billion online sales interactions a year with only one attorney and three support staff. They did not have time for a large-scale manual review.

"A manual review would have taken months of our collective dedicated time, just focused on this," Jason said. "It was not practical."

The complexity extended beyond volume. Narvar’s contract history included multiple template variations. The team had Terms of Service functioning as MSAs with accompanying implementation SOWs, Data Processing Agreements, Business Associate Agreements, along with various unique exhibits. Over the years, offline processes moved online, adding to the complexity. 

Migration Critical to Multiple Teams

Contract migrations affect far more than legal. For Narvar, several groups needed accurate data to support their function.

Finance needed contract values and pricing terms to support revenue recognition work. The deal desk needed insight into commercial terms, pricing structures, and product SKUs. Legal needed to know which customers were operating under which versions of Narvar’s terms, including new fraud protection language.

Success required a fast and reliable extraction process. The data needed to be complete, accurate, and actionable.

Adapting Coheso for a New Use Case

Coheso’s team saw an opportunity to demonstrate the flexibility of the platform’s underlying extraction engine. If the system could answer questions based on a repository for live legal work, it could also be adapted to process historical documents at scale.

The teams agreed on a phased approach that emphasized clarity, communication, and accuracy.

Phase 1: 200 Document Pilot

The project began with a pilot set of approximately 200 documents. Coheso used this phase to understand Narvar’s contract variations and to align with the legal and deal desk teams on required extraction points.

"We went through a pilot test period where we got to refine what we were pulling," Jason shared. "Coheso was very communicative as we worked through different considerations."

Phase 2: Confidence Calibration

The next decision centered on confidence scoring. Coheso’s system can extract data at different certainty thresholds. Setting the right level determines how much human review is needed.

“We were able to calibrate the confidence score however we wanted to," Jason said. "We kept it very conservative. While it is never perfect, I wanted it to be as perfect as possible.”

The goal was to dramatically reduce legal time spent manually extracting data. Strong accuracy would shrink the project from months of work to targeted validation.

Phase 3: Iteration and Expansion

As the pilot progressed, Narvar’s deal desk identified additional commercial terms that would be valuable to extract. Coheso incorporated those requirements before running the full set of over 10,000 documents.

"That process of iterating was very helpful," Jason said. "We did not want to have to run the process again."

The investment in refinement paid off when the full extraction began.

Speed Made the Migration Possible

The goal of the project was to produce reliable results that could easily be verified in a fraction of the time it would have taken to complete the review manually. 

"The time it takes to validate results as opposed to actually getting those results in the first place is significantly less," Jason noted. 

Coheso handled more than 95 percent of the work required for the migration project. What remained was a focused review of flagged items and edge cases. Without AI support, the migration would not have been possible.

"The fact that we were able to speed up this entire process made it doable," Jason said. "I do not think it would have been practical to do without some sort of AI."

Immediate Value for Legal, Finance, and Commercial Teams

The migration served multiple groups with one extraction effort.

Legal gained visibility into customer terms and contract history. The deal desk gained insight into commercial terms and pricing structures. Finance used the extracted data on their accounts receivable reconciliation and revenue recognition validation.

"Finance is using it to cross-check contract numbers against revenue data," Jason shared. “It has been very helpful."

A Flexible Platform With Strong Core Capabilities

While document migration was a new use case, the project validated the architecture behind Coheso. The same extraction and workflow engine that powers AI-enabled legal intake can support adjacent problems that require accuracy, structure, and scale.

As Jason put it, "It was extremely helpful to have a partner guide us through the process and take on the bulk of the work, while being very clear about how they could best serve our needs. The process felt collaborative, and their thoughtful approach helped us get the results we needed."

When asked about the value the project delivered, Jason was clear.

“Even at 10x the price, it would have been worthwhile. Coheso could charge a lot more for the service they provided.”

Narvar transformed a multi-month manual project into a fast and accurate AI-powered engagement. Coheso delivered a unified extraction process that supported legal, finance, and commercial operations with data the entire business could trust.