As the SDLT deadline rapidly approaches and transaction volumes surge, conveyancing firms are caught in the eye of a perfect storm: firms must somehow maintain service quality while accelerating completions – all in an environment where mistakes carry ever-increasing costs.
This pressure coincides with a technological inflection point. 2024 marked a watershed year for artificial intelligence in residential conveyancing. Capabilities that seemed out of reach just months ago are now used daily to accelerate title checks and reduce risk.
The Conveyancing Association’s Annual Conference in January provided the perfect opportunity to reflect on this rapid pace of change and tackle one of the industry’s most pressing questions directly: ‘Is AI good enough for conveyancing?’
I’ve summarised the key points from that presentation below, so let’s dive in.
AI became a lot more intelligent
If you formed an assessment of generative AI or AI applications at the beginning of 2024 or even halfway through the year, it is out of date and needs to be refreshed.
The evolution of AI last year wasn’t just quick, it was relentless. While ChatGPT’s launch in late 2022 sparked public interest in AI, 2024 marked the year this technology developed genuine reasoning capabilities that matter for conveyancing.
As large language models (the engines that power generative AI applications like ChatGPT) became much more intelligent, OpenAI faced serious competition for the first time. Anthropic and Google led the chase to build the most intelligent LLMs, and others including Meta and Mistral joined the race. This competition drove unprecedented advances in capability.
The turning point came in September 2024. OpenAI’s o1 model series demonstrated unparalleled accuracy: 83.3% on complex mathematics and 78% on PhD-level scientific problems. By December, its o3 series raised the bar even higher, achieving 96.7% accuracy on the American Invitational Mathematics Examination and 87.7% on complex PhD-level scientific questions, surpassing most human experts.
Benchmarked on PhD-level science questions, the leading LLMs improved by 56% accuracy in the space of one year: that’s not a trivial improvement. That is a huge step change.
So, what does this leap in intelligence mean for conveyancing?
Whilst earlier models could handle basic document review, today’s AI demonstrates what psychologists like Daniel Kahneman call System 2 thinking: the kind of careful, methodical reasoning essential for a multitude of tasks in conveyancing.
For example, when reviewing a property’s title or analysing lender requirements, the latest ‘reasoning models’ today will break tasks down into a sequence of logical steps and show their workings, much like a junior lawyer explaining their findings to a senior colleague.
AI’s memory expanded dramatically
The second significant development in 2024 was AI gaining the ability to process a lot more information at once, known as the context window.
This window represents everything a model can ‘think about’ at one time, from input documents and instructions to its generated responses. Context windows are measured in tokens.
By the end of 2023, most leading LLMs could only process about 32,000 tokens – roughly two typical lease agreements worth of information. Though there were techniques for using context windows more efficiently, this was a significant limitation for conveyancing. Fast forward a year, most leading models now have a context window of 128,000 tokens. That is roughly equivalent to 300 to 400 pages, allowing them to analyse an entire contract pack for a residential property transaction.
Google’s latest release, Gemini 2.0 Flash, pushes this even further with a million-token context window. This could mean every bit of data in a conveyancing transaction, from documents and emails to phone calls, could soon provide LLMs with all the context on the transaction that lawyers have access to today.
Combined with the enhanced reasoning capabilities outlined above, AI can handle complex tasks like responding to enquiries and conducting thorough title checks with greater accuracy and understanding of the full context.
This marks a significant step forward in Generative AI’s practical value for residential conveyancing firms.
AI learned to hear, see and speak
In 2024 AI transcended text-only interactions.
While previous LLMs could either process text or generate images, today’s LLMs engage with the world much like humans do through sight, sound and conversation.
Beyond asking Dall-E to create an image for you, you can now have a natural conversation with ChatGPT, ask it to analyse images, or have Gemini comment on live video feeds.
For conveyancing, this ‘multimodal’ development opens up powerful possibilities. AI could help lawyers:
- Verify that all documents have been signed and executed correctly
- Compare property boundaries in plans against written descriptions
- Analyse satellite and streetview imagery of properties to check for breaches of covenants or planning regulations restricting development
- Handle communications with clients and estate agents.
At Orbital, our AI Engineers and Legal Engineers are using these advancements to develop new and exciting product features that were previously impossible.
AI Agents arrived on the scene
By late 2024, AI evolved beyond simply processing information to actively completing complex tasks. These new systems, called AI Agents, represent one of the most significant advances in artificial intelligence to date.
Back in January 2024, our CTO, Andrew Thompson, wrote a blog predicting AI Agents would be the future of legal technology. Within months, lawyers began exploring how these autonomous systems could transform their work.
If LLMs are the brain capable of reasoning with complex information, AI Agents are like putting that brain in a body. It allows the body to do tasks on behalf of the brain.
An AI Agent builds upon the analytical capabilities of an LLM by taking independent action. While a model can process and analyse information, an agent can interact with external systems and complete complex tasks autonomously, like navigating the internet to perform tasks and retrieve information that previously required manual work or involved complicated and protracted software integrations. In residential conveyancing, Orbital Residential demonstrates this evolution.
When provided with a contract pack, searches and some additional context on the transaction, our AI Agent first processes and structures the documents, understanding their relationships and wider context. It then conducts a comprehensive title check across more than 300 different points, identifies potential issues, suggests necessary enquiries and retrieves additional information from sources like HM Land Registry and the UK Finance lender handbook.
So, is AI really good enough for conveyancing?
Whilst some AI benchmarks have been put forward by Vals.ai and others for specific practice areas or real world legal tasks, there isn’t yet any conveyancing benchmark for AI. Given how specialised conveyancing is, a generalised legal benchmark may not be that helpful as compared to a more domain-specific approach.
But we can map Generative AI’s capabilities against a lawyer’s career progression. By this metric, we believe LLMs were ‘scraping through the bar exam’ in 2023 – definitely not as capable as a trainee lawyer.
Today, after a year of rapid advancement, the most sophisticated AI Agents operate at the level of a junior lawyer – they can do several conveyancing tasks to a high standard but are still learning and require supervision.
There’s no doubt there are already a lot of tasks AI can do at a speed and scale an army of conveyancing lawyers could never match. But the reality is much more nuanced: there remain many basic and essential tasks that AI is currently incapable of, that lawyers can handle easily.
The balance is shifting rapidly, though.
We expect AI innovation to accelerate again this year. Already in 2025, Chinese AI company DeepSeek stunned the world when it unveiled its ‘R1’ reasoning model, which they claim was developed at a fraction of the cost of the other major players, and proceeded to make it freely available open source. Since then, Elon Musk’s xAI, and Anthropic have released their latest LLMs which they each claim outperform OpenAI.
Even more striking is Project Stargate – a $500 billion joint venture between OpenAI, Microsoft, Nvidia, Softbank and others, announced by President Trump. Their ambitious goal is to achieve artificial general intelligence (AGI); a system that has the ability to understand or learn any intellectual task that a human can.
To summarise, the practical applications of generative AI in residential conveyancing took major strides forward in 2024. As the first UK legaltech to become an enterprise partner with OpenAI, at Orbital we’ve had a front-row seat to this step change.
Our combined expertise in Generative AI and property law and practice (no fewer than 12 experienced real estate and conveyancing lawyers are embedded in Orbital’s product teams, making up about 15% of the current team) have guided the evolution of our products that support residential conveyancing.
As we look ahead, 2025’s early developments suggest that far from slowing down, we can expect to see even more innovation that could have important ramifications for the applicability and uptake of AI in the conveyancing sector. This is just the beginning.
Edmond Boulle is Co-Founder and CSO at Orbital
