Case Study:
Enabling Advanced LLM Functionality Protected by a Human-in-the-Loop
Introduction
Trustgrid offers a software-defined-networking (SDN) product to SaaS providers who connect their cloud applications to customer on-premise environments at scale. Trustgrid’s product replaces legacy VPN and MPLS technologies. They specialize in the banking, healthcare, energy and telco industries. As Trustgrid developed Respond, an LLM based technology to automatically diagnose and repair the network, they needed a human-in-the-loop that could review and approve the automated actions before they were executed. This would prevent Respond from making critical mistakes.
The Challenge
Existing change management processes did not anticipate a fully automated system with capabilities of such as configuration changes, reboots, updates and more. Respond would need a new system to enable human oversight of the AI’s activities.
Exploring Solutions
Trustgrid’s security team evaluated several authentication platforms to address the requirement. The identified several key requirements:
API based authentication to easily integrate with Trustgrid Respond
Group based authentication that would allow any member of the authorized group to approve the action
Group based authentication that would require a specified minimum number of members to approve for high risk actions
After reviewing available options, Trustgrid discovered SendAuth. What stood out was SendAuth's specialization in injecting human-in-the-loop oversight into automated/AI actions. It’s API first approach made it powerful and easy to integrate.
Implementation
Trustgrid partnered with SendAuth in August of 2025. The process began with:
Onboarding and Integration
Trustgrid handled all API integration work using SendAuth’s API documentation.
Customization
Trustgrid defined permission levels for different request types (e.g., triage, customer communication, disruptive actions, etc). They set up auth groups for threshold-based approvals, where sensitive requests needed confirmation from multiple authorized users.
Testing and Go-Live
SendAuth was integrated with Trustgrid Respond in its early testing phases which allowed both products to be refined for specific use cases.
The Results
Confidence in the Product
By deploying SendAuth, Trustgrid was able to ensure the inherent risks in the AI approach were mitigated by human oversight.
Operational Efficiency
The human-in-the-loop features were easy to use allowing for staff to approve or reject the requests in seconds and without being tied to a desk.
Customer Feedback
Customers that were reluctant to embrace the AI system were easily convinced by the SendAuth safeguards.