Where the work
gets stuck
A useful project usually starts with one concrete problem:
- Analysts repeatedly gather the same filings, transcripts, notes, and portfolio data.
- Useful internal research exists but is difficult to find or reuse.
- A recurring review — a new deal, quarterly earnings, or a portfolio screen — still depends on manual steps that vary by person.
- People use AI tools individually, but the work is not shared, connected to firm data, or checked consistently.
I choose one workflow with the team and trace it from source material to final output. Then I separate what AI can assist with, what ordinary software should handle, and what still requires an analyst's judgment.
The first build is the smallest version that can be tested on real work. If the useful answer is a process change or ordinary software rather than AI, that is what I recommend.
What I've built
In an ongoing engagement, I am building screening, report-generation, and connected research tools around an investment team's existing workflow. I also built a connector that makes the data and tools available through interfaces the team already uses.
The tools gather, organize, draft, and check research. Analysts retain judgment. The work continues as the team tests and adds workflows.
I have also spoken with teams at other investment firms about where their processes differ. Those conversations inform my perspective, but I only present work I have actually built as client work.
Featured conversation: I joined Brett Caughran and Khe Hy on Invest with AI to discuss how investment teams move from individual experiments to shared workflows — and why the model itself is rarely the hard part. Listen to the episode →
How engagements
work
Most build engagements are monthly retainers. A scoped assessment is also possible.
If you have engineers, I work alongside them and document the system for handoff. If you do not, I can handle the technical work.
This is a good fit when you have a real workflow, access to the people doing the work, and a willingness to test with live examples. It is not a good fit for a generic AI roadmap or an attempt to automate investment judgment.
Start with one workflow
If your team has one research workflow worth improving, tell me how it works today, where it slows down, and what you have already tried. A few sentences is enough.
Get in touch