Store your information in a structured database and then use AI on top of it.
Why?
With text-based information like contracts, you might be tempted to just train a large language model (LLM, or “AI”) on a bunch of text and ask it questions.
You could ask it what the commencement date is. This is an information retrieval question. Give me the data.
You could also ask it whether a given clause poses a risk under a given scenario. This is a reasoning question. Think about the data and give me the thought.
There’s a big difference between reasoning and retrieval. AI is uniquely good at reasoning.
If you depend on AI to store your information as well as reason about it, you’re locked into a very specific form of storage. All of your contract data is represented as billions of numbers (or “neurons”) that are entirely unintelligible to anyone or anything other than the AI you have trained.
This means that it doesn’t play nicely with other systems, you can’t export or transfer it, or do anything with the raw data, including training better AI systems on it.
You get few of the advantages of structured databases and many of the disadvantages of AI.
So... store your information in a structured database and then use AI on top of it.