Skip to content

Cabrera & Co put its methods to work in one private AI workspace.

Cabrera & Co uses CabrerAI, a private Astraed workspace. Custom models draw on approved proposals, methods, and contracts to support client work.

November 2024–August 2026 · Figures from the firm’s platform database

Warm timber meeting space used as the case-study cover
849working conversations
12,757files processed
>90%of chats through custom models

−75%

reported reduction in project-development team size—from 20 people to 5

“We went from 20 people developing projects to the 5 most important ones—the ones AI can’t replace. Cost and time efficiencies, better quality, and client expectations met more fully than before.”
Francisco Cabrera

Francisco Cabrera
CEO, Cabrera & Co

One private place for models, firm information, and client work.

Cabrera & Co first needed one controlled place for proposals, contracts, transcripts, and project material. The firm decided who could access that information before adding custom models.

The workspace kept work organized by client and gave the team one place to use approved models and tools. Consultants still reviewed the output and remained responsible for the final work.

Optional custom models and tools.

Astraed and the firm built tools for proposals, quotes, diagnostics, workshops, marketing, executive advice, recruiting, and business writing. More than 90% of the firm’s conversations used a custom model.

Across 4,676 user messages, the most common pattern was simple: the team added a meeting transcript or client summary and prepared a proposal, diagnosis, quote, or workshop plan. Files were attached to 551 messages; another 286 included long transcripts or documents.

The team reused successful work instead of starting over.

The team reused 29 successful proposal conversations as starting points and shared 32 conversations by link. Usage grew from about 20 conversations a month at launch to about 70 a month by mid-2026.

A private AI workspace can support tools built for the firm.

Start by deciding where data goes and who can access it. Add custom models only when they produce a useful, measurable result.

Talk to our team ↗