The Private Research Agent: Information Advantage at Scale

Preparation is Timeless

At Mæstery, we operate under a core philosophy: Preparation is Timeless. We have engineered our Research Mæster. Her mission: ensure the investor arrives at the table with a sharper view of intrinsic value than the seller's own advisor.

The Problem with Generic AI Agents

In institutional underwriting, a disciplined, conviction-driven approach requires unparalleled access to high-signal data. However, the current approach of giving AI agents generic web access — effectively telling them to "just Google it" — leads to agents that hallucinate searches, skim a single article, or cite junk sources. Right data, wrong logic. Right logic, wrong format. Neither survives an IC deck. This is fundamentally misaligned with the rigor of our industry.

The Research Mæster

To solve this, we engineered our Research Mæster. She is a company research analyst with an investor mental model for evaluating companies and a writing style encoded into the system prompt (much more effective than a user prompt).

Investment Process & Governance: Forced Context Injection

Instead of "voluntary Googling," our Research Mæster's research process utilizes forced context injection. We curate what the researcher can see before our researcher begins her analysis. She scans only highly relevant content from pre-filtered domains. Not keyword matching — semantic search. Think RAG, but the corpus is the entire internet, via an integration with Exa. Trusted sources include company websites, reliable news sources, government agencies, and top-tier consulting & asset management firms.

Powered by Semantic Search via Integration with Exa

Trusted Sources

Private Assets in a Public Context

Her edge: she reads private companies through the lens of their public peers.

Sector depth is the alpha. When tasked with evaluating a private company, she evaluates it through public comparables and the company's positioning within its sector.

Example

Encoding the Investor Mental Model

For each major section of an investment committee report, our researcher operates on preset templates that encode our proprietary investor mental model. She doesn't just scrape text; she conducts deep, recursive research across the following core pillars:

Autonomously selects templates:

Knows when to skip the playbook

Simple questions get a fast, single-pass scan — no recursive deep-dive, just an efficient search from trusted sources.

Encoded Behaviour: The Diligent Analyst

A star analyst knows her limitations. Our researcher's intellectual honesty is backed by directives hard-coded in her system prompt.

No data? No paragraph. She'd rather leave a gap than fill it with fluff. She never invents data — if a metric can't be cited or calculated, you get "N/A", not a guess. When her confidence in a number is low, she explicitly marks it as an [EST] estimate.

Operational Alpha: Chat History is Private by Default

Your diligence stays yours. No model trains on your conversations. Every thread lives on your cloud — not theirs.