Autonomous execution of legal and commercial contracts
Autonomous execution of legal and commercial contracts
The world’s economy is underpinned and regulated by pdfs of contracts written in legalese. Amazon invested extensively over decades to completely automate the contract of buying arbitrary things on the internet and having it show up at your door. While e-commerce contracts are unusually simple, we can now do the same for more complex documents that underpin other large subsectors like:
Much work has gone into legal informatics and particularly computable contracts to figure out how to take legal documents and turn them into the autonomously executable (i.e. computable) code while retaining the intentionally flexible and vague nature of legal clauses. Researchers have even developed programming languages specifically for this task.
On the spectrum from stochastic to deterministic, GenAI’s on the left and computable contracts are close to the right. They run the same circuit with the same output every time and have the same accuracy/specificity as the phrasing of the written contract. This attribute is obviously essential in the execution of contracts, especially In highly regulated domains. It’s also the only way that more complex automation like agentic systems can be deployed at scale in such industries, otherwise each autonomous transaction requires manual checking.
However, until now that process has been entirely manual. o1 is the first model capable of translating insurance documents into computable contracts.
We’re particularly excited about a system trained to
A startup could pick one vertical whose manual‑execution spend is already measured in the billions like health‑insurance claims or ISDA post‑trade confirmations to prove unit‑economics. 1% penetration of that single vertical yields $100M+ in revenue. From there, the company could expand throughout the value chain within that subsector (e.g. it could work on the generation of health claims to make them work seamlessly with their system) or it could roll out the trained AI system mentioned above to new sectors.
Current global spend on directly relevant manual contract execution is on the order of $100B. That doesn’t include the $100Bs in costs avoided by making the infrastructure as efficient as Amazon’s nor drag that compliance and regulation have on global GDP.
Towards Robust Legal Reasoning: Harnessing Logical LLMs in Law
Equitable Access to Justice: Logical LLMs Show Promise
Codifying Medicare Using Computable Contracts for Improved Understandings of Medical Insurance
Insurance Portfolio Analysis as Containment Testing
Specification and analysis of legal contracts with Symboleo
Knowledge Aware Automated Health Claims Processing with Medical Ontologies and Large Language Models
Symbium: Using Logic Programming to Streamline Citizen-to-Government Interactions
