WorkTango Recognition Intelligence Engine
An AI-powered scoring system that transforms unstructured employee recognition into quantitative leadership intelligence, processing recognition in real time to surface quality scores, leadership archetypes, and behavioral benchmarks across the entire organization.
My Role
VP, Product Management
Type
AI Product · 0→1
Status
Launched March 2025
The problem we set out to solve
WorkTango had a data problem hiding in plain sight. Every recognition message sent on the platform contained rich qualitative signal about who was leading well, who was developing others, and who was building culture. But it was unstructured text, invisible to analytics. HR leaders had no way to measure recognition quality, identify standout leaders, or benchmark cultural health across the organization. The data existed. The intelligence didn't.
03
Leadership archetype scoring
Recognition received tells a different story: what kind of leader are you? I designed a second scoring layer that evaluates each recognition message for five leadership archetypes: Strategist, Collaborator, Influencer, Executor, and Communicator. Scores accumulate over 365 days, building a composite leadership profile for every employee in the organization. Recipients are grouped into four leader tiers: Advanced, Accomplished, Intermediate, and Developing. The result is a leadership intelligence layer built entirely from organic, peer-generated data with no surveys and no self-assessments required.
04
Dashboard design & product integration
I worked with a data analyst, architect, and scrum team to deliver the full feature set: individual score views, manager direct report tables, department heat maps, and company-wide averages, all with cross-customer benchmarking built in. I also tied the scoring system to platform behavior: high-quality recognition gets pinned to the top of the activity feed, creating a direct loop between insight and engagement. Pinned recognition drives additional interaction through reactions, high fives, and comments, making quality visible in the flow of daily work.
MY CONTRIBUTION
From unstructured text to quantitative intelligence
I led product from earliest discovery through session experience design, AI moderation, and go-to-market readiness, partnering with the CTO and cross-functional teams to take Constellation from concept to a fully operational platform ready for enterprise deployment.
01
Conceiving the scoring methodology
The core insight was mine: employee recognition messages contain enough signal to score both the sender and the recipient in fundamentally different ways. I defined the two scoring frameworks — Recognition Quality and Leadership Archetypes — and personally refined the prompts and data models that generate the outputs. This wasn't a feature request from customers. It was a product insight that created an entirely new category of intelligence within the platform.
02
Recognition Quality Scoring
Recognition Quality scores each message across five dimensions: Specific (25 points), Meaningful (25 points), Impactful (25 points), Timely (10 points), and Celebratory (15 points), producing a composite score out of 100. Over time, senders accumulate scores that place them into quartile tiers: Exceptional, Advanced, Proficient, and Developing. Cross-customer benchmarks give HR leaders real context on how their program compares to other organizations.
WHAT WE BUILT
At scale
In the past year the recognition intelligence engine has processed 25 million recognition scores across 400,000 users, making it one of the largest real-time qualitative-to-quantitative intelligence systems in the HR tech space.
What I learned
The most important product decision was keeping the scoring invisible to employees at the moment of sending. If people knew their recognition was being scored as they wrote it, behavior would change, optimizing for the score rather than the person. By scoring after the fact and surfacing insights only to HR leaders and managers, we preserved the authenticity of the data while still creating actionable intelligence from it.
The second insight was that two datasets live inside every recognition message: one about the sender's culture and communication habits, and one about the recipient's leadership brand. Separating those into two distinct scoring systems was what made the product genuinely useful rather than just interesting.