The CHALLENGE
Contract ingestion relied on human judgment, making it hard to scale and expensive.
Productiv helps companies manage and optimize their SaaS portfolio. Contract data is one of the key ways the platform delivers value, giving customers a single source of truth for purchased applications, renewals, licenses, and cost-saving opportunities.
Since the company’s inception, contracts had been manually ingested by a growing ops team in India. Every month, thousands of files were reviewed line by line through a complex operational process, creating an ever-growing backlog and increasing costs as demand scaled.
RESULTS
$1.2M saved annually
Cost savings by automating manual contract processing.
+3 points
Transitioning to a self-serve model maintained customer loyalty.
$4.5B TCV
Value of contracts ingested across 81k files.
Establishing goals
What the project needed to succeed
Ensure functional parity
Support the same workflows previously handled by operations, without compromising the experience.
Build trust in an AI-driven experience
Clearly communicate progress, surface decisions transparently, and give customers confidence in the output.
Reduce operational effort
Create a self-serve experience that reduced dependency on operations in complex scenarios.
research
Where the manual workflow broke down
I partnered closely with the PM, engineering and the operations team to understand how contracts were processed and where the workflow broke down. The ops team processed thousands of contracts monthly, frequently collaborating with customers to resolve ambiguous cases, resulting in a 2–3 day turnaround, sometimes longer for more complex contracts.
A snapshot of the discovery process
IDEATION
Finding the right interaction model
We initially explored extending the existing contract workflow, since customers were already familiar with it. But as we mapped real ingestion scenarios, it became clear that AI introduced decisions the original experience was never designed to support.
Mapping out flows helped pressure test the designs
DESIGN
Designing for an AI world
Moving from a ops-led workflow to AI introduced a new challenges, decisions previously handled behind the scenes by operations now needed to be surfaced to customers. The experience had to help users understand, review, and recover from AI decisions with confidence.
Keeping users in control
AI should suggest the outcome, but customers needed the final say before contract data changed.
Preserve trust through transparency
When AI modified an existing contract, it should be clear what changed so users have a point of reference.
Defer when confidence is low
AI should defer uncertain decisions to the user. Incorrect data poses a greater risk than incomplete data.
Reviewing AI-generated information
How AI recommendations stayed flexible
Exploration: pattern for reviewing AI decisions
Launch
From beta to launch
We partnered with six enterprise customers through a phased beta to validate the experience in production and improve model quality. These learnings helped shape the final experience and informed ongoing improvements to model accuracy and simplifying the experience.
Following the launch in November 2025, AI-powered ingestion fully replaced the manual ops workflow, reducing review time, preserving retention, and turning a days-long process into a self-serve experience.
81k+ files processed
AI-powered ingestion scaled to fully replace manual operations within six months.
3m 57s → 1m 9s
Review time drastically reduced by improving LLM accuracy and streamlining the review experience.
↓ 40% review workload
Improved file classification reduced unnecessary customer review workload.
Replace the human-in-the-loop ingestion process with an AI-powered experience that could fully support contract processing at scale.