Skip Nav

Contact the Author

To contact the author, please fill out and send the form below.

    Thank You!

    Your message has been sent.

    Oops, message not sent.

    Please make sure fields are complete.

     

    AI Workplace Policies That Could Help Prevent A Crisis For Companies


    Commentary from crisis management expert Edward Segal, author of The Crisis Casebook: Lessons in Crisis Management from the World’s Leading Brands and the bestselling and award-winning Crisis Ahead: 101 Ways to Prepare for and Bounce Back from Disasters, Scandals, and Other Emergencies. He is a Leadership Strategies Senior Contributor for Forbes.com.

    As AI continues to evolve, it’s increasingly important for companies to have policies in place to help ensure that employees are using the technology the right way and for the right reason. Otherwise, instead of helping workers to be more productive, AI models could inadvertently be used to create a for their organizations. Worst-case scenarios include disclosure of client files and information, the release of fabricated statistics, and questionable advice and decisions.

    The Right AI Governance Policies

    A good first step in helping to prevent such scenarios from becoming reality is creating AI-related before allowing AI in the workplace. “Strong governance is the place you have to start, before the rollout. We only allow licensed tools. Non-licensed products like personal ChatGPT accounts are blocked at our firewall. Our governance process maps out expectations and builds automatic controls. Every AI automation is recorded in a register before it’s used, and every run is logged. Passwords never go into an AI tool, even the approved ones,” Chris Cluff, president of Paper Trails, a payroll and HR services company, advised in an email interview.

    The company established an in-house AI training course so that employees could learn the best practices and sign a copy of corporate policies for using Claude before they are allowed to access it, according to Cluff. In addition to restricting the use of AI to certain employees, Paper Trails monitors how Claude is used. “So far, we’ve had good success. The new automation governance process worked on day one. It stopped our tax manager from running Claude inside our payroll software and making changes to client tax settings,” he recalled.

    The AI model enables the company to see what the team was doing, which led to conversations with employees about the best use of the technology and how the policies were enforced. “ We definitely tightened too hard, too fast. But it caught some things, which is good. Now we’re slowly peeling back some rules. We review the actual tools and build controls around them before we approve anything that touches sensitive data,” Cluff explained.

    It’s better to introduce and enforce strict controls on the use of AI and relax them when warranted, rather than tightening policies and doubling down on enforcement after a crisis that could have been prevented.

    Plan Now For An AI-Related Crisis

    In June, professional services firm KPMG had to pull a report about the use of AI when it was discovered by several organizations that the document included fabricated information.

    “Research group GPTZero identified a number of inaccuracies in the report, which was published in October 2025. GPTZero told the Financial Times that the inaccuracies stemmed from AI hallucinations. In other words, the professional services firm appears to have used AI to help write a report about AI,” according to TechCrunch.

    The company removed the report from its websites and conducted an investigation into the matter. “We expect all our people to follow our guidelines on the responsible use of AI, including human oversight to validate content and verify independent sources,” a KPMG spokesperson said.

    Companies should address violations of AI policies in crisis management plans and responses to hypothetical situations. Typical scenario would be when an employee violates the policy by allowing AI to disclose confidential information, issue an inaccurate press release, or engage in illegal activity. How would safeguards prevent such scenarios from becoming realities? How would such violations be caught? Who would be responsible for responding to them? How companies answer these and other questions in the scenarios can help guide them in preparing more effective policies and procedures.

    Having a blanket AI policy may not work for everyone who uses the technology. The policies should “distinguish low-risk uses, such as brainstorming or summarizing public information, from consequential uses involving confidential data, financial reporting, hiring, customer decisions, or transactions. Employees should know which tools are approved, what information cannot be entered, when AI use must be disclosed, how important outputs must be checked, and who remains responsible for the final decision,” Martin Mulyadi, the YHB Endowed Chair in Accounting and an accounting professor at Shenandoah University, told me in an email message.

    Workplace policies should specify why and how employees can be given assignments that require the use of AI, not just which tools they can access. “Organizations should also keep a simple process for reporting errors, escalating uncertain results, and reviewing permissions when tools or workflows change. Human review is meaningful only when the reviewer has the knowledge, time, and authority to challenge the output,” according to Mulyadi.

    Blaming AI when something goes wrong is not the same thing as holding someone accountable for an incident. And such an excuse is unlikely to satisfy customers, regulators, or the media, who will want to know how and why it happened. That’s another reason why humans, not AI models, need to be seen as in charge and responsible for the technology.

    Don’t Leave AI Policies To Chance

    All AI policies should be in writing, discussed and revisited with employees regularly, and reviewed and signed by new hires.

    Trustible, an AI governance company, prohibits workers from using AI to do anything they could not do on their own. “We don’t want our HR team writing code, or our engineering team defining new hiring guidelines. This ensures the people overseeing AI have the actual capabilities to do so,” Andrew Gamino-Cheong, chief technology officer and co-founder of the company, told me via email.

    Before people are assigned to oversee the technology, they need to fully understand and appreciate what it can do and the problems it could create. It’s difficult, if not impossible, for people to spot mistakes in AI-generated work unless they understand what it generates.

    “Companies should avoid writing AI policies around a specific tool or today’s capabilities. Instead, policies should be grounded in durable principles such as responsible use, data protection, human accountability, appropriate oversight and risk. That gives organizations a foundation that can remain relevant even as the technology changes,” Kathleen Walch, director of AI engagement and community at the Project Management Institute, told me in an email interview.

    To help employees understand and follow AI policies, provide case studies and examples of how the rules apply to them. For example, show how the controls relate to the unauthorized release of data, protecting confidentiality and privacy, and safeguarding proprietary information. “AI governance needs to be treated as an ongoing practice, not a one-time policy exercise or an afterthought after new AI tools are introduced. Organizations need to continually look at how AI is actually being used, what new risks or use cases are emerging, what employees are encountering, and whether existing guardrails are still working as intended. The technology will keep changing, so organizations need a way to learn and adapt with it,” she counseled.

    To that end, the Project Management Institute prepared The Standard for Artificial Intelligence in Portfolio, Program and Project Management. It includes eight guiding principles that shape responsible AI behaviors and decisions, practices for reviewing and acting on AI-generated work, and ethical and legal guardrails.

    The standards place ultimate responsibility on the shoulders of those who use AI models. “Employees are still responsible for evaluating the accuracy, appropriateness, and consequences of the work they produce with it. Employees should understand that using AI does not transfer responsibility for the accuracy or outcome of their work to the technology,” Walch concluded.

    Don’t leave compliance with policies and guardrails to chance. Business executives should assign one or more people to oversee the proper use of the technology, conduct reviews when AI models are updated or replaced, ensure that the policies are easy for workers to understand and follow, and encourage feedback to help strengthen policies and training. Consider whether and how AI should be scaled back in the workplace if the technology is abused, and who should be contacted if an AI-generated mistake becomes public.

    Organizations should review their AI policies at least once a year and whenever new tools are added, upgraded or replaced. To be as prepared as possible when a crisis strikes, companies should have an approved statement template that can be customized and issued immediately when an AI-generated mistake becomes public knowledge.

    Skip to content