Discovery reads the tools a firm already uses, finds the work its people keep repeating, ranks each finding by the hours it could give back, and hands it to Caddi Studio to build. Available today in every Caddi workspace.
Caddi Blog
Latest from Caddi Blog - Page 2
Excel to Salesforce
How to load a spreadsheet into Salesforce properly: external IDs, upsert, dependency order, and the format traps that fail silently.
CRM Data Hygiene
The five recurring checks that keep a CRM trustworthy in a law firm or RIA, how often to run them, and why annual cleanups never hold.
RIA Back Office Automation: The Six Workflows Advisory Firms Should Stop Doing by Hand
The six jobs that grow with the book, in the order they give hours back, with two of them shown end to end.
Data Import Wizard vs Data Loader
The real limits of Salesforce’s Data Import Wizard and Data Loader, when to use each, and the three jobs neither of them does.
Caddi's launch announcement is live on AP News
The Aug 24 launch release went out over PR Newswire and is now carried on AP News: an AI agent that discovers a firm’s repetitive work, is trained on each process the way you would train a new hire, and…
Caddi in Artificial Lawyer: legal AI works, so why is the return so hard to find?
Our CEO Alejandro Castellano wrote for Artificial Lawyer on legal tech’s most uncomfortable conversation: the tools work and the P&L still can’t see them. The return is real, it just lives in the business of law, the non-billable work no…
Caddi Launches an AI Agent That Builds and Governs a Firm's Back-Office Agents
The Aug 24 launch release. The agent finds the repetitive work running inside a firm’s own systems, is trained on each process the way a new hire would be, presses on the edge cases until it has them right, and…
Caddi is a 2026 XYPN AdviceTech Competition finalist
Out of 60 applicants, six companies were selected to demo live at XYPN LIVE 2026 in San Diego, and Caddi is one of them. The finalists compete on the Expo Stage September 30 for Best in Show and Advicer’s Choice.
Why AI Agents Fail on Long-Running Tasks
On the leading real-world benchmark the best agents finish about 30% of professional tasks, and finance and HR score worse than that. Reliability decays exponentially with length and starts degrading at step two. The math, why a bigger model does…