Call Simulator AI Governance Policy

Owner – Office of the CTO

Purpose

To ensure that AI systems are utilized safely and securely on behalf of Call Simulator, Inc. (“Call Simulator,” “Company,” “we,” “our”) customers and Call Simulator organizational needs.

Scope

Usage of AI systems on behalf of Call Simulator customers, including external usage in the context of Call Simulator applications used by customers, and internal Call Simulator usage in order to serve customer and stakeholder needs.

Policy

Call Simulator follows best practices for AI testing, research and development, sourcing, deployment, maintenance, replacement, and retirement. Call Simulator leverages the core principles of Trustworthy AI, a set of guidelines developed to direct safe AI usage in secure medical and research contexts.

The mission of Call Simulator is focused on training through use cases where no Protected Health Information (PHI), Personally-Identifiable Information (PII), or other personal, financial, or otherwise private data is necessary. As such, related AI systems used by Call Simulator on behalf of customers do not need private or sensitive data of these types for operation. Additionally, Call Simulator customers control the data inputs used for creating, running, and scoring simulations and have ability to edit data and metadata used within the Call Simulator application as part of scenario building, simulation delivery, and evaluation of learner performance. All of these controls reduce unnecessary risks or needs for Call Simulator customers to expose private data to AI systems in the course of using Call Simulator. All of these considerations and controls are communicated to Call Simulator customers in the course of onboarding and implementation of the Call Simulator application.

With the mission, use cases, and controls above in mind and in recognition that private and sensitive data is not required for usage, Call Simulator, Inc. has taken additional steps above and beyond what may be required in typical personnel training contexts in the area of AI governance by utilizing the core principles of Trustworthy AI as outlined below.

Trustworthy AI Principle – Fair / Impartial

Call Simulator uses an Ethical Sourcing policy for components of AI, including compliance with applicable laws and regulations. Call Simulator customers establish the metrics of learner / user performance and have means to see and adjust results of AI operations including elements like text generated for drafts of scenario conversations and results of AI-based scoring and applicable coaching. Call Simulator customers have the ability to export data from the Call Simulator application’s evaluation of learner performance and related measures during customer build and acceptance testing of new customer-created or customer-modified training. In this manner, Call Simulator and customers have the ability to monitor for bias and/or apply external fairness metrics towards Call Simulator’s use of AI on behalf of customers.

Trustworthy AI Principle – Transparent / Explainable

Call Simulator customers configure, test, approve, and manage scenarios and simulations. This includes the ability to see and adjust input data, such as conversation descriptions, coaching, sample “model” conversations. And this includes the ability to see and adjust scoring evaluation means, such as rubrics, skills, and criteria. And Call Simulator customers can see and adjust learner scoring evaluation results, such as data on learner performance and whether behaviors were displayed or not by learners. While some aspects of large language models (LLMs) may be fundamentally obscured by the nature of LLMs, the controls above provide Call Simulator customers with the ability to monitor and adjust AI usage within Call Simulator’s core use case of conversational training. As such, Call Simulator customer builders and trainers have visibility, tunability, explainability, and discardability of AI-based elements within Call Simulator.

Trustworthy AI Principle – Responsible / Accountable

Call Simulator regularly reviews AI systems as part of Ethical Sourcing and internal testing for applicability to customer use cases and within data usage, privacy, and AI governance policies. Where applicable, and not in violation of non-disclosure of proprietary information, Call Simulator supports customer control of and visibility to criteria evaluative logic used for AI-based simulation performance evaluation. For significant changes to Call Simulator AI processes, customer input and/or acceptance testing is included, such as through customer hands-on beta testing prior to implementing AI changes.

Trustworthy AI Principle – Safe / Secure

Call Simulator use of AI and related data falls into privacy, data usage, access control, and other data management policies and agreements with customers. In support of these controls and covenants, Call Simulator utilizes automated tools, authorized professional judgment, and regular risk reviews to support the safety and security of customer data and uses of AI. These include ongoing vulnerability scanning, code and secret analysis, and proactive consideration of elements that may lead to future risk. Additionally, Call Simulator uses content filtering and safeguards against threats, including prompt injection attacks, to keep AI systems and related data safe and secure.

Trustworthy AI Principle – Privacy

Call Simulator use of AI and related data falls into privacy, data usage, access control, and other data management policies and agreements with customers. As such, appropriate data governance is in place on behalf of customers – notably with customers having the primary control over what data is used during customer simulation training. As noted previously, Call Simulator requires no private, personal, health, financial, or other sensitive data in order to deliver the core mission and use cases around learner conversational  training. Additionally, Call Simulator customers have the ability to edit and/or remove data they may have unintentionally added to Call Simulator or that is no longer needed for Call Simulator-based training.

Trustworthy AI Principle – Robust / Reliable

Call Simulator strives for robust and reliable AI, within the acknowledgment that generative and conversational uses of AI have naturally-occurring and customer-desirable stochastic elements that allow for a level of variability in response. To shape the limits of variability, Call Simulator customers are able to choose whether to use AI processes, such as generation of sample, draft conversation text or a sample, draft scoring rubrics. Likewise, customers can choose which AI persona voices and AI persona characteristics, such as name, conversation goals, and attitudes, are to be used in a given training scenario. As such, Call Simulator customers are able to control how scripted vs. unscripted conversational training moments may flow and adjust how AI may be used for simulation evaluation. This includes choices like scoring learners on exact or similar word usage compared to a reference conversation, or scoring learners on a more behavioral basis using a rubric, and both with customer trainer overrides. 

Finally, in the case of scoring overrides, Call Simulator, Inc. staff regularly reviews scoring overrides among other channels of customer activity and feedback to monitor and tune robust and reliable AI usage within Call Simulator services to customers. These inputs along with other Trustworthy AI principles, data policies, and Ethical Sourcing policy at Call Simulator are used in ongoing AI inventory and lifecycle management.

Policy Compliance

Call Simulator will measure and verify compliance to this policy through various methods, including, but not limited to, business tool reports and both internal and external audits.

Exceptions

Requests for an exception to this policy must be submitted to the CTO for approval.

Violations & Enforcement

Any known violations of this policy should be reported to the CTO. Violations of this policy can result in immediate withdrawal or suspension of system and network privileges and/or disciplinary action in accordance with company procedures up to and including termination of employment.

Last Updated April 23, 2025