The real-time data to power financial automations

In 2021 we decided to solve financial intelligence.  What does this mean in practice?

Organizing, cleaning, standardizing the world’s financial information, making it retrievable and usable, starting from transactions. 

Transaction data is one of the most valuable systems of record that has been largely untapped because of the variety of players and legacy technology in the payments ecosystem. 

I remember pitching this to QED investors and later talking about the payments graph and its implications with Stephen at Lakestar, who led our subsequent funding round. 

In the last two years there are many things we have learnt, and the world we have been building for has changed completely.

Middleware

When we first started, we were building middleware for humans. 

Our ideal customer has always been a developer at an FI who does most of their data work in the data warehouse. Then the data is moved into business tooling, from CRM-s to analytics and underwriting dashboards where business users become the secondary users of this information. 

Most of the workflows with the data would still happen in Excel, needing to be reviewed and approved by human workers. 

Market shifts

In 2023 GPT4 launched making it feasible to get to human level understanding of financial transactions. Training our in-house models became futile. We decided to focus on the infrastructure to serve the larger models reliably at a very low cost and very high speed.

Today we have achieved results that no one else on the market has come close to.

We can enrich and standardize financial transactions using the largest and most expensive LLMs, over a hundred times cheaper and faster than the next best option.

We talked about this when we revealed the numbers and technology behind Ntropy’s AI orchestration platform for finance. 

Another shift has started happening.

The data behind automating financial workflows

We are at a cusp of living in a world where AI is going to query and build with financial data much more than humans.

Today all financial workflows are majorly run by many human operators, from KYC/KYB and onboarding, to underwriting and finally financial reporting and accounting. Most of enterprise financial data lives in Excel and is operated on in Excel.

This is quickly shifting.

We are seeing new breeds of workflows and companies emerge that are building entire workflows with LLM-s vs point solutions. Today they are called agents. 

Agentic software whether it is in lending or accounting has different requirements than humans. It needs highly accurate, fast and very precise data to rely on. LLMs can parse billions of transactions and orchestrate decisions and actions with them.

Ntropy’s API is purpose built for such a scale and precision and can operate at a cost and speed no one else can. 

When we were building for humans, the difference the API was making was often incremental. Humans do not operate at this scale. Models do.

We are in awe of what is happening and are super excited to serve our customers on this journey.