Caddi

Product Introduction

1. Definition: Caddi is an AI-powered operations automation platform, specifically categorized as an "Automation by Demonstration" or "Agentic Process Automation" tool. It uses a combination of AI reasoning and deterministic code generation to create software agents that execute back-office workflows.
2. Core Value Proposition: Caddi exists to automate repetitive, high-volume back-office tasks in professional services firms without requiring traditional coding, complex RPA (Robotic Process Automation) setup, or workflow builders. Its primary value is in discovering, learning, and automating existing manual processes across a company's real software ecosystem, governed by leadership with full auditability.

Main Features

1. Process Discovery Engine: Caddi's AI analyzes connected systems (like Outlook, Docusign, NetDocuments) to identify and quantify repetitive work. It ranks processes by frequency, complexity, and automation potential, providing a data-driven backlog of tasks worth automating. This is powered by system integrations and data analysis algorithms.
2. Automation by Demonstration (Narrated Screenshare Training): The core teaching mechanism. Users train Caddi agents by screensharing a task while narrating their actions and decision logic. The platform's AI observes the UI interactions, listens to the narration, and asks clarifying questions about edge cases (e.g., "What if only one party has signed?"). It translates this demonstration into a deterministic automation script.
3. Deterministic Execution with AI Guardrails: Unlike pure AI agents that "reason" from scratch each time, Caddi generates and runs deterministic code for the core workflow. AI is used only for judgment calls (like classifying an email) within the predefined script. This ensures reliable, repeatable, and predictable execution where the same input always produces the same path and output.
4. Comprehensive Governance & Audit Trail: Every agent run is logged in detail. Leaders can see a replay of each decision, the data involved, and the specific permissions used. Runs can be paused, and processes can only be updated with human approval. This provides transparency and control, crucial for regulated industries.
5. Loop Studio & Pre-built Integrations: The interface for managing automations ("Loops"). It provides a visual diagram of the running process and allows for updates via plain English chat. Caddi offers 150+ native integrations with core professional services tools like Microsoft 365, Salesforce, Docusign, NetDocuments, iManage, Clio, Tamarac, and QuickBooks, enabling cross-platform automation.

Problems Solved

1. Pain Point: Manual, repetitive back-office work consumes significant employee time in law firms, RIAs, accounting firms, and insurance carriers. This includes email triage, document filing, client intake, payment reconciliation, and matter setup, which are error-prone and scale poorly.
2. Target Audience: Operations leaders (COOs, Directors of Operations), practice managers, and administrative teams in regulated professional services firms. Key personas include Legal Operations Managers, RIA Compliance Officers, Accounting Firm Partners, and Insurance Back-Office Supervisors who need reliable, auditable automation.
3. Use Cases:
   • Legal: Automating the triage of inbound client emails, filing executed agreements from Docusign to document management systems (NetDocuments/iManage), and updating matter status in Salesforce.
   • Financial Services (RIA): Reconciling trust account payments, automating client onboarding data entry, and syncing data between portfolio management (Tamarc/Addepar) and CRM systems.
   • Accounting: Processing inbound remittance advice emails, posting to payable in QuickBooks, and managing engagement letter workflows.
   • Insurance: Intake triage for new claims, document classification, and updating carrier systems.

Unique Advantages

1. Differentiation vs. Traditional RPA/Workflow Builders: Unlike RPA (e.g., UiPath) which requires technical developers to map processes step-by-step, or low-code workflow tools that need scoping and building, Caddi discovers the process and learns it by demonstration. It drastically reduces the time-to-automation and technical barrier to entry.
2. Differentiation vs. Pure AI Agents: Unlike AI agents that operate in a "black box" and can drift or hallucinate, Caddi's hybrid model uses deterministic code for execution. This provides the reliability and cost predictability of traditional automation with the adaptive learning of AI, avoiding unpredictable token costs and output inconsistencies.
3. Key Innovation: The "Agent that builds agents" paradigm. Caddi's "Loopy" AI assistant doesn't just perform tasks; it analyzes systems to propose automation opportunities, then guides the user through a collaborative training session to build a robust, production-ready software agent. This shifts the paradigm from building automations to teaching them.

Frequently Asked Questions (FAQ)

1. How does Caddi ensure data security and compliance for law firms and financial advisors? Caddi is SOC 2 Type II certified and operates on a strict data governance model. Content from your connected systems is never used to train public AI models (using zero-retention terms with providers like OpenAI). Access is scoped to the user's existing permissions, every action is logged for audit, and you retain full ownership of your data.
2. What happens when a website or application I use changes its interface? Caddi's deterministic automation scripts are built to be robust, but if a UI change breaks a step, the platform's monitoring will flag the failed run. Users are notified and can quickly retrain that specific step through the chat interface in Loop Studio, without rebuilding the entire automation.
3. Can Caddi handle complex, non-routine decisions or only simple tasks? Caddi is designed for structured, repetitive processes. It excels at tasks with clear rules and steps. For complex, novel decisions requiring deep expertise, it is not a replacement. Its AI is applied to bounded judgment calls within a known process (e.g., "is this email a contract or a payment?"), not open-ended problem-solving.
4. How is Caddi priced compared to RPA platforms or other AI automation tools? Caddi offers a predictable pricing model, not based on the volatile consumption of AI tokens. This provides cost certainty for scaling automation. It offers a free Individual plan and tiered plans for teams, contrasting with the high developer costs of RPA or the unpredictable run costs of pure AI agent platforms.
5. What kind of technical setup or IT support is required to deploy Caddi? Setup is designed for operational users, not IT departments. It involves installing a browser extension for the recorder and connecting existing cloud applications via OAuth. No code, servers, or complex API integrations are required by the customer, significantly reducing the deployment burden.