<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=1691976594387096&amp;ev=PageView&amp;noscript=1">

ESTIMATED TIME:   1 minute read

Synthetic data production improves next-gen conversational AI engine

Learn how we helped Microsoft achieve better modeling through structured, synthetic data production resulting in an evolving, scalable conversational AI engine.

The quantity of usable data and the quality of deep learning algorithms are the determining factors when it comes to how “smart” an AI can become. However, most ventures are reaching the end of their data supply long before their algorithms begin to falter.

Microsoft, a leader in technology advancements, acquired Sematic Machines to serve as their next generational artificial intelligence (AI) engine. These AI needed to be trained on practical technologies for accurate responses with the end-users. To drive this result, Microsoft decided to study the AI’s ability to calendar by scheduling relevant meetings, reading daily agendas, and responding to similar tasks. This allowed them to focus on:

  • Improving data quality and product performance
  • Building a tailored data set focused on mitigating identified gaps and training their AI
  • Streamlining data production processes and defining best practices
  • Developing strategies to focus data production on accelerating product development

Microsoft engaged The Spur Group to deep dive into their data production processes and systems, with the goal of improving data quality and acceptance by their highly sophisticated Conversational AI Engine.

The Challenge

  • Limited data production scale failed to accelerate product and feature development
  • Lack of available data hindered deep modeling and machine learning for AI engine
  • Low quality and relevancy standards with outsourced data production
  • Gaps in artificial intelligence knowledge-base due to minimal training in data across key scenarios

The Solution

  • Engaged in joint planning with Microsoft’s leadership to identify key success drivers
  • Developed a data strategy and engineering vision for scaled synthetic data production
  • Designed a program management framework and support for a new, more scalable data production team
  • Streamlined data production processes, with industry best practices and a standardized model
  • Employed machine teaching strategies to automate data production and rapidly improve model performance
  • Created a documentation library of dozens of trainings, process maps, best practices, and other reference materials

The Result

  • Managed AI performance across 100+ product features
  • Improved model accuracy by an average of 14% across managed features
  • Saved thousands of hours in data production through machine teaching and automation strategies
  • Increased data relevancy at intake, reducing downstream costs by 33%
  • Scaled production capacity through process redesigns, specialized trainings, and automation

“We would not be in the position we are today without The Spur Group. They brought a level of expertise and professionalism that was above and beyond. Our partnership significantly accelerated the evolution of our AI engine and our progress toward our product vision.”

Mikko Ollila
Microsoft | Principal Program Manager