Good question, here is what makes the project both technically more challenging and much more impactful in comparison to your typical "AI for X" startup:
1) We enable businesses to create a repository that in the back powers an individual ("per account") vector DB. This is the only way that allows the real-world use (i.e. in business) of up-to-date data (as opposed to pre-trained or proprietary LLM models) while at the same time effectively preventing data "spillovers" to other users and safeguarding data overall.
2) In contrast to the tons "generative" AI startups you see, our approach evolves around leveraging LLMs capability of interpreting context even when it has some human flaws. We then transpose the context in a standardized, structured data set, effectively making human-written content usable on a big scale. We extract metadata, detect clauses and find potential flwas in contracts. This is important because while LLMs can solve a part of this challenge in the short term when the context window is small, but it reaches its limit with very datasets, when they are unstructured.
3) On a non-AI note: We have created the first true end-to-end system aimed at non-enterprise, non-tech businesses that covers the entire process from import/creation of a document until it is securely signed and stored. This is what we believe the only viable foundation if we really want to convince people to abandon their 40 year old MS Word driven workflows. Having AI features neatly integrated in our workflows effectively allows our customers to benefit from the possibilities LLMs offer.