By Jonny Palmer, Data Science Lead, Collaborative Conveyancing
Innovation and data science are two words or phrases that are often mentioned together. For both, it isn’t always about finding the right solution immediately – it’s about testing, predicting and asking the question, what if? And knowing when to revisit ideas that didn’t work the first time. In our journey at Collaborative Conveyancing, this has been particularly true with the evolution of our data science department and our use and adoption of Generative AI (Gen AI).
Initially, when we set out to create a solution to streamline conveyancing workflows, the data science team found Gen AI, at that point lacked the precision needed in the legal space. Instead, working alongside the team at Hartree UK, we pursued a different approach, building a sophisticated product that answered the problem we are trying to solve. However, while this worked in principle, it highlighted a crucial lesson: innovation must serve the user, not just the idea of progress.
Fast forward three to six months, and the landscape of Gen AI had transformed dramatically. This was when we revisited it – not as a standalone solution, but incorporating it within our product, enabling a more fluid and intelligent system. The early product that had utilised Hartree’s expertise in Natural Language Processing and designing AI solutions that cater to specific industries remains as the backbone of our product, it’s Gen AI that now powers the product as it stands today – creating a smarter, more adaptable system that understands legal documents, structures knowledge effectively, and enhances efficiency without interfering with professional judgment.
Understanding Generative AI and NLMs
At its core, Generative AI functions by continuously predicting the next most likely word, phrase, or symbol (known as “tokens”) given an input. While models have clearly become incredibly robust and effective at this process, ultimatley, each output from one of these models is still boils down to being a combination of predictions.
This predictive nature presents both opportunities and risks, especially within conveyancing where compliance and accuracy are critical, this approach is not always suitable. Instead, we’ve taken a more structured approach, leveraging AI in ways that enhance but do not replace professional expertise.
Training AI the Right Way: Learning from Experience
A key differentiator in our product’s success has been the role of experienced conveyancers in training our AI. Understanding conveyancing is one thing, but responding in the correct way is another. Our experts have guided the AI, ensuring that it recognises what to look for in legal documents while also respecting the nuances of professional decision-making. This approach allows us to use AI effectively for tools like glossaries and knowledge bases, which enhance workflow efficiency without overstepping into areas where legal expertise is required.
This balance is crucial. Some aspects of conveyancing benefit from AI-driven fluidity, such as document categorisation, summaries, and data extraction, while others must see AI pulled back maintaining rigidity, and ensuring compliance and legal integrity. Knowing where to apply AI and where to maintain human oversight is the difference between innovation that works and innovation that disrupts without benefit.
The Future of AI in Conveyancing
As AI technology continues to evolve, so will our approach. The future lies in refining how we integrate these tools, ensuring they support professionals rather than replace them. By continuously training our AI with real-world expertise and maintaining a careful balance between automation and human judgment, we’re not just innovating for the sake of it. We’re creating solutions that truly work for conveyancers, today and into the future.
Jonny joined Collaborative Conveyancing in June 2023. An experienced Data Anylyst from his time at Dye & Durham UK, Jonny ensures that he and the Data Science team are utilising cutting-edge techniques from across the world of machine learning, AI and natural language processing in order to better understand, interpret and act upon enquiries.
