A placement workbench built around agents, for the brokers who decide

Why we built it

Most of a transaction risk broker’s week goes into reading: the purchase agreement for its reps and qualifiers, the schedules for what is disclosed, every diligence report for what it covers, each quote for its exclusions, the binder for what changed. That reading is exactly what a language model does well — as long as it is held to the page and kept inside the broker’s authority.

So InsightDealDesk starts from the agents. Each deal is worked by eleven of them, each with one job; their work is shown first in the Deal Workspace with every finding cited; and what they prepare waits for a broker to accept, change or reject. The record keeps who decided what, on what.

What we hold ourselves to

• AI first, brokers decide. The agents reason with the model by default; their rule-based findings are the grounding and the fallback.
• Every figure from the engines. A model sentence with a number the facts don’t hold is discarded.
• Placement actions stay with people. Sending, approaching markets, selecting, accepting a binder and signing off are capped at ask-first.
• Honest demo data. Every deal, market, person and figure on this site is fictional, and labelled so. We do not name customers here.

The broker suite

InsightDealDesk places the cover; InsightDueDiligence reviews the target’s insurance program and hands its findings to the placement. Both share the Insight platform with InsightPulse for company intelligence and InsightXtract for reading documents: the same identity, agent framework, autonomy dial, Rules Studio and decision record.

VerticalServe

InsightDealDesk is built by VerticalServe, part of the InsightLake enterprise AI suite.