Financial services deserves better than another pilot.
We started Bluegrass because financial services businesses were being sold AI they could not maintain, evaluate or own, while the engineers who could build it sat inside labs and big technology companies.
So we put a small, entirely senior team between the two, independent of every vendor and focused on financial services.
Since then we have worked with some of the largest banks, RIAs, fintechs and AI labs, and with dozens of insurance companies, including a top-3 P&C carrier, a top-10 P&C brokerage, a top-5 trucking brokerage, a leading specialty MGA and a leading trucking MGA.
We specialize in systems built on large language models, which read, compare and draft the documents financial services runs on. Insurance remains a core focus alongside banking, wealth management, lending and fintech.
Fifty years of AI engineering, and no junior bench.
Bluegrass is deliberately small, with multiple granted patents, Y Combinator founders and researchers and operators from the institutions below. The people you meet in the first call write the code.
Engineering
Applied AI and distributed systems engineers who have shipped machine learning into production at scale, and hold patents for some of it. They know where these systems break, because they have been on call for them.
Founding & product
Y Combinator founders who have taken products from nothing to real revenue. That is where our bias toward shipping small, measurable things quickly comes from.
Industry & strategy
Operators out of McKinsey, Goldman Sachs, Microsoft and Vista Equity Partners, who can sit with an executive team and connect a technical decision to the P&L it moves.
Four weeks in, you should already know whether to continue.
- 01
Understand the business
We sit with the underwriters, analysts, advisors and operations staff who do the work, read the actual documents and count the actual volumes, so that nothing is scoped from an org chart.
- 02
Prove it on your data
Before committing to a build, we test feasibility against a real sample and set an accuracy bar with the people who will live with the output.
- 03
Build in the open
Weekly demos on working software, your team in the repository from day one, and evaluation results published as they change.
- 04
Hand it over properly
Documentation, runbooks, on-call training and a support window. Success is your team shipping the next change without calling us.
The rules we hold ourselves to.
We bring an informed opinion, formed with clients across financial services, about what you should and should not work on.
Say the inconvenient thing
If a workflow is not worth automating, if the data is not there yet, or if a $40 tool solves it, we say so in week one. We would rather lose scope than sell you a system that quietly fails.
Earn autonomy in stages
New systems start with a human in the loop and measured accuracy. Autonomy goes up as the evidence supports it, one workflow at a time, with the numbers on a dashboard you control.
Build for the person who inherits it
Everything we build is maintainable, understandable and well documented. This means that whoever works on it can get started and iterate quickly.
Stay independent
No vendor relationship is worth a compromised recommendation. We keep our advice free of partner economics so it is worth listening to.
Bring us the workflow that frustrates you.
The best engagements begin with someone walking us through the part of their week that should not still be manual.