Webinar: Put your data to work
Kevin Hsu
Affinity runs a webinar series called Put Your Data to Work covering how private market firms are using AI across the deal cycle. Earlier episodes covered sourcing and diligence. They asked us to finish the series with the stage that starts at close and never really ends: portfolio monitoring.
Everything we showcased ran on data from Lumonic and PitchBook. Portfolio company financials from Lumonic. Comparable and market data from PitchBook.
Two kinds of data that rarely share a system
The most useful workflows in private markets sit between what your companies report and what the market says around them. The first lives in monthly financials and compliance certificates. The second lives in comparables, firmographic data, and index levels. Every quarter, analysts bridge the two by hand in Excel, usually under deadline. It's part of why 60% of the private credit firms we meet still run monitoring in spreadsheets, and about half of private equity firms too.
Every demo in the episode leverages MCP, the Model Context Protocol, an open standard that lets a model like Claude read from and write to systems directly. Connect the Lumonic and PitchBook MCPs and the model can pull your portfolio's reported financials and the market context around them into the same artifact, with every number still traceable to where it came from.
A truly auditable valuation model
Every private fund marks its portfolio quarterly, and those marks get audited. My co-founders spent years at Carta pulling valuations into software, and their conclusion after building that career is that valuations should live in Excel. An auditor is going to ask how you got to that number, and they want to trace every dependency down to the source.
We asked Claude to build a valuation model in Excel from Lumonic and PitchBook MCP inputs. It wrote Lumonic formulas for the financials and PitchBook formulas for the comps, blending market multiples with a discounted cash flow model. The finished file has a $47 million figure, summed from monthly values to last twelve months, that traces back to the exact cell in the financials the portfolio company submitted. One click, without leaving Excel.
Portfolio review as a prompt
The quarterly review is the same ritual everywhere: who's performing, who isn't, who needs help, who doesn't. The meeting itself runs about two hours. The prep is the expensive part. We have customers running reviews across 200 to 400 portfolio companies.
We generated company one-pagers as HTML files, a file type showing up at asset managers of every size. We've noticed this trend spike in the last year, customers are actually writing their software now, whether they meant to or not. We help by building custom style guides in their firm's colors and they hook up the MCP. After that, generating the review pack is a prompt. Every number in the one-pager clicks back to its source document, same as in Excel.
A watchlist across 200 positions
For scaled managers the question narrows to where to look first. We showed an exposure dashboard with roll-ups from Lumonic and sector and index data from PitchBook. The watchlist already knows who breached a covenant and who's growing too slowly. What you have in the ground and what's happening around it, in one view.
Customers are already building past us
The part I didn't expect to be talking about this year: large asset managers showing us interfaces they built themselves. "I vibe-coded this" is a sentence we now hear from private equity and private credit funds. If the MCP is powerful enough, the interface layer becomes yours to design, down to the hex code on a chart. We build our templates with that direction in mind.