Research, operating lessons, and sharp takes on what it actually costs to put AI to work inside real businesses.
Each post is designed like a working note: a clear thesis, grounded evidence, and a direct path into the full argument.
This post makes the case that a knowledge graph is becoming basic infrastructure for private equity, the way a CRM was twenty years ago: it connects the people, companies, and conversations already scattered across a firm's tools so deal sourcing stops depending on who happens to remember what.
18 Jun 2026 · 8 min read
A knowledge graph links everything a fund cares about (documents, emails, meetings, the CRM) around people and companies, so you follow the threads instead of digging for them.
One firm reported that after putting this in, the number of deals it could screen in a year went from roughly 500 to roughly 5,000.
Keep it concrete: show the banker-deck workflow, explain why sourcing comes first, and address the 'don't we already have Copilot?' objection head on.
The strategic point isn't cost savings. In a high-margin business, the win is better decisions: more targets screened means better investments reach the table.
Proprietary deal sourcing used to be about building relationships. Increasingly it's about remembering and reusing them better than everyone else. Here's why funds are treating knowledge graphs as basic infrastructure.
The knowledge is already in the firm's files. The problem is that it gets harder to access every year.
A data-led look at UK tech compensation and AI investment, with a projection for when AI spend per employee may converge with average salary cost.
Salary growth trends remain linear. AI tooling spend appears to be compounding much faster.
The important question is not whether AI is getting cheaper. It is what cost structure gets rewritten when it does.
The same discipline behind these research notes is how we design production AI systems for operators, analysts, and revenue teams.