Case Study

Saving $80K+ with employee questionnaire automation

We built a NLP-based neural network designed to automate question generation, saving $80K+ in the first year for one of Canada’s leading HR software businesses.

Moon Graphic

The Client

A SaaS company providing solutions for digital onboarding, employee records, time-off management, automated benefits, and payroll synchronization. Among other things, the software enables businesses to onboard new employees, complete tax forms, sign and store employment documents electronically, set up and track time-off policies, automate and manage payroll.

The Challenge

One of the key promises of the client’s solution is its ability to understand and account for personal circumstances of each employee, ensuring ease of use and employees’ satisfaction.

While employees use benefits plans in a somewhat similar way, each employee’s use case often involves a circumstance that needs to be accounted for to enable the desired degree of personalization. Dealing with multiple third-party insurers and benefit providers, the client employed several people whose job was to read benefits brochures, insurance policies and similar content, identify provisions relevant to specific use cases and update the knowledge base with each new scenario. Given that policies frequently change, maintaining the knowledge base represented a significant cost for the client.

The Solution

Crater Labs proposed an NLP-based neural network capable of composite understanding of the text in order to formulate context-aware questions and provide intelligent responses.

The research project benefited the client in multiple ways. In addition to greatly reducing human involvement in the development and maintenance of the knowledge base, it enabled the client to develop a self-learning question/answer chatbot, extract relevant benefits data and automatically generate/update benefit FAQ, and enable intelligent content analysis to identify cross-selling opportunities.

IP Generated

Crater Labs developed an RNN encoder-decoder architecture with the global attention mechanism enabling the model focus on certain elements of the input when generating each word during decoding and enabling intelligent context extraction. The model outperforms commercially available solutions.

Benefits & ROI

This project resulted in ~$80,000 annual cost savings in addition to more than 100% return on investment from leveraging Canadian investment tax credits. This ROI does not account for the potential value of the developed IP.

Keep Reading

More Moonshots Worth Celebrating

Identifying Potential Security Access Violations

We developed a semi-supervised neural network to identify anomalous security access card swipes in a large corporation with numerous facilities throughout the world, increasing secure access violations by over 10%.

Read More

From 4 Hours to 4 Minutes: Expediting routing times for North America’s leading transportation software

We built an industry-first, AI-driven engine that calculated the best delivery route in four (4) minutes vs. the industry average of 4-6 hours.

Read More

Energy demand prediction for generation management

We developed a deep reinforcement learning model to assist energy producers in scheduling maintenance and minimizing operating costs.

Read More