Human-led. AI-assisted. Accountable by design.

Human Accountability

A human professional remains accountable for PEC work.

AI systems are tools. They are not accountable professional actors.

PEC does not delegate ultimate responsibility for professional judgment, client recommendations, organizational diagnoses, personnel decisions, strategic conclusions, ethical decisions, validation of important findings, or acceptance of final deliverables to an AI system.

Where AI contributes substantively to professional work, a qualified human must evaluate the resulting output before it becomes part of PEC’s final work.

Human oversight must include the ability to question, correct, reject, or replace AI-generated information and recommendations.

The individual responsible for consequential work must be identifiable.

Purposeful Use

PEC uses AI where it improves the work, not merely because AI is available.

Before introducing AI into a process, we consider:

1. Q1What problem are we solving?
2. Q2What value could AI add?
3. Q3Is AI the appropriate method?
4. Q4What risks might its use introduce?
5. Q5What human capability should remain involved?
6. Q6What should the resulting process allow people to do better?
7. Q7Could the intended outcome be achieved effectively without AI?

Efficiency is one consideration. It is not the only consideration.

We evaluate AI use in relation to quality, accuracy, judgment, learning, confidentiality, organizational performance, and the people affected by the work.

The ability to automate a task does not, by itself, establish that automation is the appropriate decision.

Appropriate Human Oversight

The level of human oversight should correspond to the sensitivity, complexity, uncertainty, and potential consequences of the work.

Routine applications such as formatting, editing, brainstorming with non-sensitive information, and document organization generally require less oversight than work involving strategic recommendations, employee information, organizational diagnostics, financial implications, employment decisions, or other consequential outcomes.

PEC applies greater review, verification, documentation, and human intervention as risk and consequence increase.

For consequential applications, oversight must occur at appropriate stages of the work, rather than relying exclusively on final approval.

Human reviewers must have sufficient information, competence, and authority to challenge the output and determine whether its use is appropriate.

Where meaningful human oversight cannot be provided, AI use must be restricted, redesigned, or declined.

Verification and Evidentiary Integrity

AI-generated output is not automatically treated as fact, evidence, or professional analysis.

PEC independently verifies material factual claims, calculations, sources, interpretations, and analytical conclusions when they contribute to consequential findings, recommendations, or deliverables.

Verification is proportionate to the significance and potential impact of the work. Depending on the application, verification may involve:

• →Consulting authoritative or primary sources
• →Examining original documents or datasets
• →Testing calculations and analytical methods
• →Reviewing assumptions and alternative explanations
• →Evaluating the quality and representativeness of information
• →Comparing findings against established professional knowledge
• →Assessing uncertainty and limitations
• →Obtaining additional qualified review

AI-generated citations must be verified before they are relied upon as supporting evidence.

Outputs that cannot be adequately substantiated will not knowingly be presented as established fact.

Reviewing one AI system’s output with another AI system may support analysis, but does not automatically constitute independent verification.

PEC distinguishes between evidence, professional interpretation, assumptions, hypotheses, and AI-generated suggestions.

Confidentiality, Privacy, and Data Stewardship

Client confidentiality remains a governing obligation regardless of the technology used.

PEC evaluates information sensitivity, client permissions, contractual obligations, platform safeguards, and relevant legal requirements before processing client information through AI systems.

Special care is required for personally identifiable information, employee records, personnel matters, assessment results, protected or health-related information, proprietary materials, intellectual property, privileged communications, confidential strategy, and sensitive organizational data.

PEC’s intended safeguards include:

• →Using approved AI platforms and configurations appropriate to the information involved
• →Limiting information provided to what is reasonably necessary
• →Avoiding unnecessary identifiable or sensitive information
• →Applying de-identification or aggregation where appropriate
• →Respecting contractual restrictions and client-specific AI requirements
• →Evaluating provider data-use, retention, access, and security terms
• →Obtaining required authorization before materially sensitive processing
• →Managing AI-related work products and records according to applicable retention obligations

Authorization to perform consulting work does not automatically establish authorization to process all client information using external AI providers.

Where adequate safeguards, permissions, or lawful grounds are unavailable, the information must not be processed through the proposed AI system.

PEC does not assume that consumer and enterprise versions of the same AI platform offer equivalent protections.

PEC does not use client information to train AI models, and uses paid platforms with model-training settings configured to exclude client information.

Transparency and Disclosure

PEC does not intentionally misrepresent materially AI-generated or AI-assisted professional work as exclusively human-produced when that representation would reasonably affect a client’s understanding of how the work was developed.

Transparency is proportional to the nature of AI involvement, the information processed, the consequences of the work, and the client’s expectations or contractual requirements.

Not every use of grammar correction, formatting assistance, administrative automation, or incidental AI support requires individual disclosure. PEC distinguishes between:

• →Incidental production assistance
• →Substantive professional development support
• →AI-supported analysis
• →Sensitive or consequential AI use

Material AI involvement is disclosed at the appropriate engagement, process, or deliverable level when required by the significance of its contribution, client agreement, or applicable obligations.

Disclosure should be accurate and sufficiently specific to explain how AI contributed without exposing confidential methodologies, privileged information, or unnecessary technical details.

Where advance client authorization is required, disclosure after the work has been completed is not a substitute.

PEC’s disclosure decisions are guided by the accompanying AI Use, Risk & Disclosure Framework.

Human Judgment at Consequential Decision Points

PEC does not believe automation should automatically equal decision authority.

Where AI contributes to consequential decisions affecting individuals or organizations, human judgment, context, challenge, and accountability must remain identifiable. Organizations should be able to answer:

1. Q1Who is responsible for this decision?
2. Q2What information informed it?
3. Q3How was AI involved?
4. Q4What assumptions or limitations were considered?
5. Q5Who can challenge, correct, or override the outcome?
6. Q6What recourse exists if the result is inaccurate or harmful?

PEC does not treat AI-generated recommendations, scores, classifications, or predictions as independently sufficient grounds for consequential personnel or organizational decisions.

Particular care is required when AI output influences access to employment, advancement, compensation, performance opportunities, or other material outcomes.

Bias, Fairness, and Organizational Context

AI systems may reproduce or amplify limitations, assumptions, biases, or gaps contained in training data, system design, prompts, measurement approaches, or supplied information.

PEC therefore treats AI output as an input to professional reasoning rather than a substitute for it.

Particular caution applies when AI informs work involving people, talent, leadership, employment, assessment, opportunity allocation, performance, organizational power, or consequential decision-making. Appropriate assessment may include consideration of:

• →Data quality and representativeness
• →Relevant employee or stakeholder populations
• →The validity of measures and underlying constructs
• →Foreseeable disparate effects
• →Accessibility and reasonable accommodation
• →Contextual and structural explanations for observed outcomes
• →The possibility that AI recommendations reinforce existing inequities

A statistically identified pattern is not automatically an explanation of its cause.

PEC considers the broader organizational system before turning a pattern into a conclusion or recommendation.

Where legally regulated assessments or employment practices are involved, applicable validation, nondiscrimination, accessibility, and other legal requirements must be considered.

Human Capability, Not Dependency

PEC seeks to use AI in ways that expand human and organizational capability rather than unnecessarily erode it.

When redesigning work, we consider not only what can be automated, but what expertise people must continue developing.

Professional judgment is frequently developed through experience, feedback, repetition, problem-solving, interpretation, and the opportunity to make and correct decisions. Some of the work that AI can perform is also work through which people historically learned their craft.

This creates an important organizational design question:

What capability might we unintentionally lose while improving efficiency?

PEC encourages organizations to evaluate how AI adoption may affect the development of expertise, particularly among early-career professionals, emerging leaders, and employees whose roles are undergoing substantial redesign. That evaluation may include:

• →Preserving meaningful opportunities for independent problem-solving
• →Maintaining access to foundational work experiences
• →Designing supervised practice and feedback
• →Developing AI fluency alongside domain expertise
• →Reinforcing critical thinking and analytical judgment
• →Ensuring that employees can recognize and challenge AI errors
• →Periodically reassessing whether automation is creating unhealthy dependence

These considerations do not imply that AI should always be restricted during employee development. They recognize that capability development must be intentionally designed when technology changes how experience is acquired.

PEC believes responsible AI adoption should improve both current organizational performance and the capacity of people to perform effectively in the future.

Responsible Experimentation, Governance, and Continuous Improvement

AI technologies, applications, and associated risks will continue to evolve. PEC therefore treats responsible AI practice as an ongoing professional discipline rather than a one-time policy decision. Our approach includes:

• →Evaluating tools and use cases before consequential adoption
• →Testing AI-assisted methods before relying on them for significant work
• →Identifying foreseeable risks and necessary safeguards
• →Maintaining appropriate records of material AI use
• →Monitoring the reliability and continued suitability of recurring applications
• →Addressing identified errors or incidents
• →Reviewing changes to platforms, capabilities, terms, or requirements
• →Strengthening AI fluency and professional competence
• →Periodically revising our methods and standards

Material errors, inappropriate disclosures, or other significant AI-related concerns must be evaluated and addressed through appropriate corrective action.

Where an AI application cannot be operated with sufficient reliability, confidentiality, oversight, or accountability, PEC will modify, suspend, or discontinue that use.

Client Choice and Professional Boundaries

Where AI use materially affects how client work is performed, processed, analyzed, or governed, client requirements and contractual commitments will be respected.

Clients may establish additional limitations regarding permitted systems, information, confidentiality, disclosure, and AI use. PEC will communicate material AI use and obtain required authorization before proceeding.

Where a client’s requirements prohibit particular forms of AI assistance, the engagement will be designed accordingly or PEC will discuss the implications before agreeing to the work.

PEC will not knowingly use AI in a manner inconsistent with agreed client restrictions.

Client authorization does not eliminate PEC’s independent responsibility to exercise professional judgment or comply with applicable obligations.

AI in the Service of Possibility

PEC’s position is ultimately human-centered.

The objective of AI implementation should not simply be to remove labor or accelerate existing processes. It should be to determine how technology can help organizations operate at greater capacity while improving:

• →How work is performed
• →How people contribute
• →How decisions are made
• →How expertise and capability develop
• →How responsibility and accountability are maintained
• →What organizations become capable of achieving

We believe the responsible introduction of AI requires organizations to decide what they are trying to accomplish, intentionally design how work should change, establish appropriate governance, and then deploy technology.