Problem: A big-4 professional services firm wanted to know the outcome of 80000 historic legal judgements. Reading these documents by hand would have taken over 10,000 hours and cost hundreds of thousands of pounds.
Solution: In just a few hours, 3 lawyers used Humanloop to train a high quality model that was able to provide the outcome on all the historic documents. Insights from the model allowed the firm to win new work.
Problem: A customer service start-up, Blastable, wanted to improve the efficiency of their customer support workers so they could take on more business.
Solution: They used Humanloop to train a model that can read a customer query and automatically select which response template is most appropriate. When the model is not confident, it routes the message to a human agent and learns from them in real time. This continuously improving system lets a small team scale to thousands of queries.
Problem: A leading NLP start up, Black Swan Data, needed to manually label thousands of social media posts in order to track consumer trends. The need for annotation made expanding to new regions extremely expensive.
Solution: Using Humanloop's actve learning technology they were able to select only the hightest value data to label and so train high performing models with greater than 50% savings.
Problem: A real estate technology start-up, Office Hub, needed to extract key information from commercial property websites, such as locations and rental costs. Normal web-scraping wouldn't work because the information appears in too many varied ways.
Solution: Using Humanloop they were able to quickly train a model that could take in the unstructured text of a brochure or website and return the structured data they required.