Applying artificial intelligence tools to a manager's proprietary experience and expertise captured in its data is a clear route to leveraging a firm's edge. But how do firms create the data layer required to generate AI-assisted insights, efficiencies and speed? Taylor Lowe, CEO and founder of Metal and a private markets AI specialist, and Jessica Pizzo, resource partner at mid-market investor Court Square, discuss what managers should consider as they decide what technology to buy and what to build.
Lowe describes AI infrastructure in three layers. The data layer holds all of a firm's information about its deals, portfolio companies, documentation and decisions. The intelligence layer includes core AI tools such as search, agent automations and domain-specific tasks like deal scoring, benchmarking and pattern recognition. The third layer is an interface that must be accessible to both humans and AI agents.
When choosing whether to buy or build, Lowe points to time-to-value, total costs, data portability, model-locking, ongoing talent needs and opportunity cost. Many firms are still testing off-the-shelf tools such as ChatGPT and Claude, while others engage consultants for custom builds. Lowe stresses that a proof of concept differs greatly from a scaled system serving 50 to 100-plus employees, and that firms without a coherent data foundation will run into walls.
At Court Square, Pizzo explains that a custom build was ruled out because of high one-time costs, complexity and the expense of retaining talent over time. After 12 months of evaluating tools, the firm prioritized flexibility to switch frontier models and a provider with private equity expertise who understood its processes and could build workflows and agents without extensive education.
Source: PE Hub · Summarized by HeadlinesBriefing