5 posts tagged with "Research"

If you work in governance research, compensation consulting, investing, board advisory, or business journalism, executive compensation data is only useful when you can compare it quickly and explain the components clearly.

The pay data lives in SEC proxy filings. Recent filings include inline XBRL tags, but the human-readable compensation tables still vary across companies. Older filings are even messier. The important numbers are inside long DEF 14A documents, footnotes, named executive officer tables, director compensation tables, pay-vs-performance sections, and company-specific formatting.

The SEC Executive Compensation MCP server is built so an agent can answer pay questions from structured SEC compensation data instead of making a user dig through proxy filings by hand.

If you work in grantmaking, nonprofit diligence, donor research, journalism, or civic analysis, IRS 990 filings are where the financial story lives. They show revenue, expenses, assets, liabilities, officer compensation, program service activity, and organizational structure.

They are also painful to work with directly.

The data is split across form variants, annual extract files, hundreds of columns, and identity joins. Organization names are not always where people expect them to be. EINs are the stable key, but users often start with a name.

The IRS 990 Nonprofit Financials MCP server turns those filings into a searchable product for AI agents and REST clients.

If you are a journalist, policy analyst, advocacy lead, investor, or researcher, lobbying disclosure is useful because it answers a live question about influence.

Who is lobbying for this company? Which clients does this firm represent? What issues are they lobbying on? Has their focus changed over time? Which trade association keeps appearing around the same bill or agency?

The Federal Lobbying Filings MCP server gives AI agents and REST clients a structured way to search and inspect lobbying records.

The average MCP tool description is bad. Two recent arXiv papers measured exactly how bad. 97.1% of analyzed tools carry at least one description smell. 73% of servers reuse the same display_name across multiple tools. In head-to-head selection between five functionally equivalent servers, the one with a clearer description gets picked 72% of the time vs a 20% baseline.

A 260% selection lift from prose alone is the headline. Descriptions aren't documentation; they're part of the agent-facing prompt, and a bad one quietly costs you every selection round your server is in. So we built a scorer.

MCPBundles now runs an LLM-as-judge rubric over every published server, covering tool descriptions, server-level skill content, and the structured MCP annotations that the runtime actually reads. The verdict shows up on the public listing page.

Stanford's Digital Economy Lab just published The Enterprise AI Playbook — a 116-page study of 51 successful enterprise AI deployments across 41 organizations, 9 industries, and 7 countries. The research team, led by Erik Brynjolfsson (one of the most-cited economists on technology), interviewed executives and project leads who deployed AI at scale and measured actual results.

The headline finding: the technology was never the hard part. In 77% of cases, the hardest challenges were invisible — change management, data quality, and process redesign. Not model selection. Not prompt engineering. Not which AI provider to use.

This post pulls out the findings that matter most for anyone building or buying AI tooling today.