AI/ML

MetricWire Launches AI-powered Scheduling Tool to Boost Survey and Customer Support Efficiency

Posted by

onixadmin

About the Customer:

MetricWire is a technology company that provides a platform for researchers to collect and analyze real-world data. Their platform allows researchers to design and conduct studies using smartphones and other electronic devices, enabling them to gather high-quality data on participants’ real-world behaviors without needing to write any code. MetricWire’s platform can notably be used in various fields, including:

Healthcare: Patient monitoring, clinical trials, behavioral health research
Social Sciences: Surveys, observational studies, longitudinal research
Market Research: Product testing, consumer behavior analysis
Transportation: Travel behavior research, urban planning
Environmental Studies: Pollution monitoring, impact assessment

The Challenge:

Researchers needed a flexible system to gather data from participants in their natural environments, but this meant accommodating a wide range of unpredictable factors like participant location, time of day, and even their current activity. Simply put, scheduling surveys with traditional methods was too rigid and often led to missed opportunities or inconvenient timing for participants.

To overcome this, MetricWire wanted to create a dynamic scheduling system that could adapt to these real-world variables. This meant:

Handling dynamic scheduling: The system needed to go beyond simple calendar reminders and factor in various conditions, trigger types (like location or time), and even allow customized prompts to appear based on a participant’s context.
Ensuring scalability: As MetricWire grew and researchers designed more complex studies with larger participant pools, the system needed to handle increasing volumes of data and user requests without impacting performance.
Providing a seamless user experience: Despite the complexity behind the scenes, researchers needed an intuitive interface to easily design their scheduling logic, and participants needed a smooth, unintrusive experience when receiving survey prompts on their devices.

The Solution

Our team leveraged Google Cloud’s Vertex AI and Gemini to develop a sophisticated “Scheduling Agent” tailored specifically to MetricWire’s needs. This solution moved beyond simple rule-based scheduling to intelligently interpret researcher requirements and participant context.

Key components of the solution:

Gemini:This language model has the ability to understand and interpret nuanced language, including complex scheduling requests with multiple conditions and triggers. With Gemini, we extracted key information from researcher inputs, like desired survey frequency, preferred times, and trigger events (e.g., location, activity), to ensure accurate and efficient scheduling.

BigQuery: Acting as the system’s memory, BigQuery, Google Cloud’s highly scalable data warehouse, stored all the necessary data for the Scheduling Agent to function effectively. This included participant data, survey responses, scheduling parameters, and historical interactions. BigQuery’s ability to handle massive datasets and provide fast data retrieval was essential for real-time processing and ongoing refinement of the scheduling model.

LangChain: This framework acted as the connective tissue between PaLM2 and the rest of the system. LangChain streamlined the interaction by managing prompt engineering (translating researcher requests into a format PaLM2 understands), response generation (interpreting PaLM2’s output), and seamless data retrieval from BigQuery.

Cloud Functions: These serverless functions provided the agility to automate actions within the system. For example, when a researcher submitted a new scheduling request or a participant triggered a condition, a Cloud Function would activate the Scheduling Agent to process the information and determine the optimal survey schedule.

Looker: To provide MetricWire with complete visibility and control, Looker, a business intelligence and data visualization platform, was integrated into the solution. A customized Looker dashboard allowed MetricWire to monitor the Scheduling Agent’s performance in real time, tracking key metrics like scheduling accuracy, processing speed, and user engagement. This data-driven insight enabled continuous optimization and informed decision-making.

How it works:

Results

The implementation of the AI-powered Scheduling Agent yielded significant improvements for MetricWire:

Increased Scheduling Accuracy: The system demonstrated a remarkable improvement in scheduling accuracy, ensuring surveys are deployed at the optimal times based on user preferences and historical data.
Reduced Latency: Average latency dropped to under one second per request, even for complex scheduling scenarios, providing users with a seamless and responsive experience.
Improved Scalability: The system efficiently processes over 100 scheduling tasks weekly, showcasing its ability to handle growing data volumes and user demands.
Schedule Generation: Gemini analyzes the prompt, considers historical data and scheduling parameters, and generates an optimal survey schedule.
Enhanced Confidence: Overall confidence in the system’s responses consistently exceeds 90%, reflecting the accuracy and reliability of the AI-powered solution.
Data-Driven Optimization: Continuous feedback and performance monitoring through BigQuery and Looker enable ongoing refinement of the PaLM2 model, leading to further improvements in accuracy and efficiency.

“Partnering with Onix to develop our AI Scheduling tool was a valuable experience. They understood our business challenges, validated key assumptions, and kept the engagement aligned with our ROI goals. The implementation of the AI Scheduling tool exceeded our expectations. Onix played a critical role in our decision to strategically align our roadmap with Google’s AI offerings.”
Brian Stewart CEO,
MetricWire

Conclusion

The new scheduling system has delivered clear improvements in both efficiency and user satisfaction. By reducing scheduling latency to under one second and significantly improving accuracy, the system has streamlined operations and enhanced responsiveness. The use of BigQuery to analyze user data played a key role in continuously refining the system, ensuring it adapts to changing needs. With the ability to process over 100 weekly tasks and maintain over 90% accuracy, the system proves reliable, scalable, and effective.

This case study demonstrates the value of taking a proactive approach with an AI-powered scheduling agent. By leveraging AI, businesses can improve performance, reduce manual errors, and stay ahead of the competition, all while driving stronger user satisfaction and maximizing return on investment. The system also positions the organization to remain agile and competitive in an evolving digital landscape.