
Fix the Data First: Why AI Strategy Starts With Your Data Foundation, with Taylor Lowe of Metal
In this episode of the Rainmaker Podcast, host Gui Costin is joined by Taylor Lowe, CEO and co-founder of Metal, a New York-based AI company building the context layer for private capital. The show usually features heads of distribution, but this conversation takes on a question every distribution leader is facing: what to do about AI, LLMs, internal data, and external data sources while compliance, data integrity, and other stakeholders weigh in on every decision.
Taylor traces his path from Washington, DC, to San Francisco, where he started as an SDR at Copper (then ProsperWorks) before moving into account executive and product manager roles. He then joined Kustomer in New York, a CRM that was later sold to Meta, and met his co-founders there. After the acquisition, they launched Metal out of Y Combinator in early 2023, first as horizontal infrastructure for large language models and soon after focused on private markets. He credits YC with teaching the team to ask the right questions and stay committed to what is true in the market rather than chasing hype.
The conversation then turns to the core problem. A firm’s most valuable asset is its internal data, including its judgment, expertise, and hard-won lessons, but that data sits in unstructured documents across CRMs, file repositories, email, and chat, with no single system of record. Taylor explains that LLMs finally make it possible to connect and normalize this information, but only when paired with software that classifies document types, extracts the right data, and ranks sources by provenance. Metal is also model-agnostic, so customers can reach the same contextualized data through Claude, GPT, or Gemini over MCP, in whatever tools they already use.
A major theme is starting with the data foundation instead of the workflow. Firms that rush to automate a single use case, such as analyzing expert calls, often hit a wall as soon as they add a second source like the CRM or file repository. Gui compares this to outlining a presentation on a yellow pad before ever opening PowerPoint, since design is only as good as the data underneath it.
Gui and Taylor also dig into why data decays. Gui argues that acquiring data is not the hard part. Enriching, organizing, and continually updating it is, and bragging about how much data you have is really admitting an updating problem. Taylor likens bad data to salt in cooking, easy to add and hard to remove, and shares how Metal’s own agents keep its CRM so clean that he no longer logs into it directly. With a trusted core, the platform compounds in value over time, which Gui compares to a golf course thoughtfully taking down trees to improve the health of the whole course.
Once internal data is clean and maintained, external sources and benchmarks become far more useful. Firms can compare their own performance to third-party data, spot patterns in the deals and LPs they have worked well with, and keep everything traceable and auditable. Taylor closes by naming education as a key challenge: as the models evolve, he works to help firms keep pace and understand why the data foundation matters. He adds that his main competition is firms building in-house on the foundation labs, where the hard part isn’t the first workflow but maintaining the data layer underneath it. Gui wraps up by congratulating Taylor on solving a very large problem in a trustworthy way.
More episodes from "Dakota Rainmaker Podcast"



Don't miss an episode of “Dakota Rainmaker Podcast” and subscribe to it in the GetPodcast app.








