We have seen a lot of use cases in finance and are currently working with a few firms. Accuracy is generally their primary requirement. We've been focusing on not just accuracy but also the audit experience which includes confidence scoring. Your right about security and we are currently undergoing a soc 2 audit.
Generally firms are outsourcing the data entry and are already manually auditing with shortcuts like summing values.
Relating to formatted in the table they expected. We extract the data directly into their template format so it is always in the table they expect.
This is something we are hyper focused on. shoot me an email at max@midhsip.dev - you can also try our financial analysis template in our playground for free: https://app.midship.ai/demo.
Hi! In our playground we only included a few templates, but in our app you can create any template you'd like! So yes we can extract from academic papers! shoot me an email: max@midship.dev
This is an interesting use case! We've heard similar stories from people dealing with pensions. Today we are ready to solve out of the box the extract data from a factsheet into excel or CSV step. Shoot me an email at max@midship.dev!
We're excited for foundational models to improve because we hope it will unlock a lot more use cases. Things like analysis after extraction, able to accurately extract extremely complex documents, etc!
We saw initial traction with real estate firms extracting property data like rent rolls. But we've also seen traction in other verticals like accounting and intake forms. The original idea was very ambitious and when talking to potential customers they all seemed to be happy with the existing players.
Humans in this space tend to make mistakes like, "Added rent as per-square-foot instead of absolute value," or, "Missed a rent escalation for year 3."
These tend to be easy to catch, even for the same person who's reviewing the data. They would see that rent steps looked strange (Y2 and Y4, but not Y3) or there was an order-of-magnitude difference in rent from one month to another.
AI can do something like invent reasonable-looking rent steps. They're designed to create output that seems reasonable, even if it's completely made up.
When humans are wrong, they tend to misread what's there, which is much less insidious than inventing something.
And if you have a human reviewer for all the work this AI does, what's the point of the AI in the first place? The human has become the source of truth either way.
+1 to what mitch said. We believe there is a large market for non-technical users who can now automate extraction tasks but do not know how to interact with apis. Midship is another option for them that requires 0 programming!
