SMEs make up 99.2% of UK businesses (ONS), yet only 24% can find an affordable legal services provider (LSB 2017). In fact, YouGov and CEBR found SMEs often completely fail to resolve their legal issues, resulting in legal costs worth £13.6bn a year. Genie's data shows that service-based businesses have increased contractual needs due to their variety of clients and partners. Despite this, a survey of 43 users on Genie's platform stated "understanding" legal documents was their biggest challenge. Our proposal is to create a state of the art, large scale transformer language models that can explain and summarize legal clauses, but secondly, to augment it with retrieval capability so that it can also explain what is market standard, whether that's a liability cap in a service agreement or a salary in an employment contract. This means when a small services company is trying to offer their services, employ someone, transfer IP or raise investment, they can receive unparalleled automatic guidance of how to create/review legal agreements. This includes summarizing clauses to aid understanding, explaining what is often negotiated and which commercial terms are market standard.

no public description

Two of the greatest obstacles towards adoption of artificial intelligence in UK services is the acquisition of confidential data, and the explainability of black-box neural models. This research will draw on a number of academics from leading research institutions, large commercial partners and Ginie AI, a machine learning startup to tackle these issues. In particular, commercial products that advance state of the art algorithms will be developed. These solutions will draw on the latest body of research in computational privacy and machine learning. The technology will be researched, tested and trialled in a commercial setting. In addition, key stakeholders and regulatory bodies will be engaged with to provide an industry wide protocol of how to enable access to data for the rapid adoption of machine learning in services.