ASK THE QUESTION. CONNECT THE SYSTEMS.

THE PROBLEM.

At Commercial Networks, we were using StackJack. I explored building our own internal alternative to connect AI assistants to ConnectWise, with more control over access and ongoing cost.

MY ROLE.

I created and tested a working prototype, catalogued more than 200 prospective Automate operations and recorded 44 passing automated tests during internal development. These figures describe the prototype, not production deployment.

HOW IT WORKED.

1. Translate intent

   Map a supported natural-language request to a defined operation through the gateway.

2. Apply boundaries

   Design authentication, read-only boundaries, secret handling and audit logging around the operation. OAuth/PKCE and Azure Key Vault were part of the authentication and secrets design.

3. Return useful context

   Query the relevant ConnectWise API and return a structured answer that an engineer can review.

MAKE IT TRUSTWORTHY.

44 automated tests passed during internal development. The prototype was prepared for a controlled internal pilot; it was not a full organisation-wide production rollout. Wider use would require validating authentication, permissions, hosting and operational ownership in that environment.

Controls that matter

• Defined operation catalogue
• Read-only access boundaries
• Authentication and secret-handling design
• Audit logging
• Controlled internal pilot scope
• No public RMM connection

WHY IT MATTERS.

The prototype demonstrated a practical route between an assistant and existing MSP systems. It also exposed the engineering work around permissions, auditability and deployment that a useful integration needs.