Better AI policies start with employee input

Employee participation is crucial for creating trusted AI policies

Only 48% of organizations have a formal AI policy, according to Dice’s 2026 Tech Sentiment Report. Nearly one-quarter of technology professionals surveyed said they have already used AI without manager approval. This creates an immediate governance problem: employees are adopting AI faster than many companies are defining how it should be used.

The practical response is to involve employees in policy design. Workers know which tasks consume time, which processes could benefit from automation, and where AI creates new risks. Their input can expose issues that a technical deployment team may miss.

IT teams naturally focus on integration, security, access, and system capabilities. Amy Loomis, group vice president of Workplace Solutions at IDC, argues that these questions differ from asking employees where they waste time on low-value work and where AI could improve their day. Both perspectives matter. An effective policy connects technical controls with actual workflows.

J&Y Law took this approach. Monica Washington Rothbaum, COO and senior attorney at J&Y Law, said the firm first surveyed employees through SurveyMonkey to understand which AI tools they were already using and why. It then worked with individual departments and involved operations, IT, HR, and leadership. The process examined potential uses alongside potential risks.

For executives, the key constraint is policy legitimacy at the point of use. A technically sound policy has limited value if employees find it impractical and work around it. Unauthorized AI use can expose confidential data, bypass security controls, and make oversight more difficult. Employee participation helps management design rules that people can follow during real work.

Paul Farnsworth, president of Dice, said the firm’s 2026 research shows that technology professionals generally are seeking clarity about how AI will affect their jobs, careers, and workplace decisions. His conclusion is important for management: “The research suggests that employee buy-in comes from making AI feel like something being done with workers rather than to them.”

Participation should therefore continue beyond the initial policy draft. Employees should have a role in pilots, feedback processes, and policy updates. Management retains accountability for the final rules. Employee input supplies the operational evidence needed to make those rules effective.

Employee anxiety about AI extends beyond job replacement

Three-quarters of technology professionals surveyed by Dice believe junior-level workers are most at risk of AI-driven displacement. A majority of non-AI technology professionals also believe AI eliminates more jobs than it creates. These concerns can directly affect adoption because employees judge AI partly through its expected impact on their careers.

Job loss is one concern. Employees can also worry about AI systems used to screen job applications, monitor performance, or provide information that influences raises and promotions. These uses carry higher stakes than routine productivity tools because they can affect income and career progression.

Executives also need to examine what happens after employees start using generative AI successfully. Generating an output is only part of the work. Employees may have to add missing context, verify facts, correct errors, review confidential information, and repeat prompts until the result meets the required standard. This oversight burden is sometimes called “botsitting.”

The same issue can spread between employees. Poorly checked AI-generated material can become “workslop”: output that looks complete but transfers verification and correction work to colleagues. AI can therefore shift effort between people even when it reduces the time required for the original task.

Workload is another management issue. Higher individual productivity can lead to higher expectations and more assigned work. Heavy users may also move repeatedly between AI tools, write and refine prompts, and evaluate inconsistent outputs. This can contribute to what workers describe as “prompt fatigue” or “AI brain fry.”

The executive measure of success should therefore extend beyond AI usage rates. Leaders need to understand whether AI reduces total task time after review and correction, whether output quality improves, and whether workload remains sustainable. These measures reveal whether AI is producing operational gains rather than moving effort into less visible activities.

Employment concerns also require specific communication. Employees need to know how AI will affect their responsibilities, how performance will be assessed, and where human judgment remains required. Clear governance around AI-assisted employment decisions can reduce uncertainty while preserving management accountability.

This matters for workforce relations as well as technology adoption. Technology professionals have shown growing interest in unionization, with workplace AI among the factors driving that interest. Companies that address employment effects, workload, monitoring, and employee input early have a stronger basis for maintaining trust while expanding AI use.

Organizations need specific protections to address employment fears

Three-quarters of technology professionals surveyed for Dice’s 2026 Tech Sentiment Report believe junior-level workers face the greatest risk of AI-driven displacement. A majority of non-AI technology professionals surveyed also believe AI eliminates more jobs than it creates. Executives should treat this concern as a governance issue because uncertainty about employment can weaken trust and slow adoption.

General promises about job security have limited value when employees can see roles, teams, or reporting structures changing. Amy Loomis, group vice president of Workplace Solutions at IDC, said: “Reassurances that no jobs will be lost ring hollow when workers can see reorganizations happening around them.” She argues that organizations sustain trust by explaining what is changing, how those changes affect individual roles, and what the employer will commit to in return.

This requires concrete policy. A company can define which employment decisions require human authority. Hiring, dismissal, promotion, compensation, and other high-impact decisions are clear candidates. AI can support analysis where policy permits, while an accountable person retains responsibility for the final decision. One protection proposed is an explicit requirement that decisions affecting an individual worker’s employment be made by a human.

Executives should also communicate at the level employees can act on. Workers need to understand which tasks may be automated, which responsibilities may change, what new skills will be expected, and what support will be available. That gives employees useful information for career planning and gives managers a clearer framework for implementing AI-driven changes.

The business objective is predictable governance. Companies can pursue productivity gains while setting clear limits on AI’s role in consequential workforce decisions. Specific commitments also give HR, legal, IT, and business leaders a common standard against which deployment decisions can be assessed.

AI governance requires continuous, multi-channel communication

AI policies need regular updates because the underlying tools and their uses continue to change. New capabilities can alter which data employees enter into AI systems, how outputs are reviewed, who needs access, and which workflows require stronger controls. Communication therefore becomes an ongoing governance function.

J&Y Law provides a practical example. Monica Washington Rothbaum, COO and senior attorney at J&Y Law, said the firm communicated AI updates through company-wide meetings, email, Microsoft Teams, and department-level discussions. It also assigned a member of its marketing team to oversee communication about AI adoption, creating explicit ownership and accountability.

The range of channels matters because AI decisions affect different functions in different ways. A company-wide update can establish common rules. Department discussions can translate those rules into specific workflows. Digital channels can distribute revisions quickly and provide employees with a current reference point.

Rothbaum describes AI adoption as an operational, communication, and governance challenge. Organizations must decide what data employees can use, who receives access, how AI-generated output is reviewed, and how compliance requirements apply as the technology evolves. These are management decisions with consequences across IT, legal, security, HR, and business operations.

For C-suite leaders, ownership is the central issue. Someone must be responsible for communicating policy changes, collecting feedback, coordinating affected functions, and ensuring updated rules reach employees. Clear ownership reduces the risk of different departments developing conflicting practices.

Communication should also support two-way feedback. Employees using AI in daily workflows can identify unclear rules, unexpected failure modes, and new use cases before those issues become visible at executive level. That feedback can then inform policy revisions, training, access controls, and deployment choices.

Treating AI governance as a continuing operating process gives executives a more durable model for adoption. The policy sets expectations. Regular communication keeps those expectations aligned with changing technology, workflows, and organizational requirements.

Continuous training should be a formal part of AI policy

AI skills have a short shelf life. Tools, capabilities, workflows, and risks change quickly. Companies therefore need an ongoing training commitment tied to actual employee roles and approved AI systems.

Paul Farnsworth, president of Dice, said employees are more likely to embrace AI when they see opportunities to develop alongside it. As AI becomes embedded in daily work, he recommends investment in AI literacy, upskilling, and career development to help employees adjust to changing job requirements.

Executives should formalize that investment in AI policy. Amy Loomis, group vice president of Workplace Solutions at IDC, said this requires “treating training as a policy commitment with defined standards, timelines, and completion tracking.” That creates measurable responsibility. Leaders can establish who requires training, what competencies each role needs, when employees must complete courses, and how skills will be refreshed as systems change.

Role-specific training is important because effective AI use depends heavily on context. A lawyer reviewing confidential material has different requirements from a marketer producing draft content or an IT employee configuring an AI service. Training should cover approved tools, permitted data, output validation, escalation procedures, and the risks relevant to each workflow.

Human judgment also remains central. Loomis said effective training combines instruction in specific AI tools with judgment, critical thinking, and adaptability. Employees need the ability to identify weak output, question unsupported conclusions, detect missing context, and decide when human review or expertise is required.

This has a direct operational benefit. Poorly trained employees can spend significant time correcting AI output, repeatedly refining prompts, or passing weak AI-generated work to colleagues. Better training can help employees determine where AI improves a workflow and where its review costs reduce the expected productivity gain.

Training also supports workforce planning. Employees who understand how AI changes their roles have a clearer path to acquiring relevant skills. Management gains a structured way to prepare teams for changing responsibilities. Loomis warns that when training becomes a one-time event, adoption can stall and distrust can grow. Continuous learning provides a stronger basis for sustained adoption.

Clear guardrails are required for secure and responsible AI use

Every enterprise AI policy needs explicit boundaries around data, access, output review, and acceptable use. Employees should be able to determine which tools they can use, what information they can provide to those tools, and which outputs require human verification.

Data handling is a central risk. Employees can expose confidential business information, personal data, client material, or intellectual property when entering content into an AI service. J&Y Law addresses this directly. Its policy states that confidential client, firm, or employee information may not be entered into an AI tool unless the use is explicitly authorized and approved.

Executives should translate this principle into controls employees can apply during daily work. Policies can identify approved systems and data categories, define access permissions, establish review requirements, and specify escalation procedures for uncertain or sensitive cases. Technical controls can reinforce those rules through identity management, permissions, logging, and approved enterprise AI environments.

Guardrails must also address the consequences of AI-generated output. Generative systems can produce errors, omit relevant context, or generate biased results. Human review is especially important when an output influences customers, employees, legal matters, financial decisions, or other high-impact activities. Clear accountability ensures that a named person or function owns the resulting decision.

Discrimination requires particular attention when AI affects people. AI-assisted hiring, screening, performance assessment, and other workforce processes can create legal and governance risks if models or data produce unfair outcomes. Policies should define appropriate human oversight and review standards for these uses and align them with applicable employment, privacy, and anti-discrimination requirements.

These employee-facing safeguards should sit alongside the company’s wider governance, risk, compliance, cybersecurity, and regulatory controls. This gives leadership a consistent framework from initial experimentation through production use.

Monica Washington Rothbaum, COO and senior attorney at J&Y Law, summarized the management priority clearly: “The organizations that get the most value from AI won’t be the ones that adopt it the fastest. They’ll be the ones that communicate clearly, train consistently, and build the right guardrails before they need them.”

For executives, speed of deployment is therefore a weak measure of success. Useful adoption depends on whether employees can use AI within clear boundaries while protecting company, client, and workforce information. Strong guardrails make responsible adoption easier to scale because employees know which actions are permitted and where human accountability begins.

Effective AI policies should define productive uses as clearly as restrictions

An AI policy should tell employees where AI creates business value and how they can use it safely. Clear permission matters. Employees need practical guidance on approved tools, suitable tasks, required review, and the types of information they can process.

Restrictions remain essential for security, confidentiality, compliance, and human oversight. Yet a policy centered primarily on prohibited behavior can leave employees uncertain about legitimate use. That uncertainty can suppress useful experimentation or push employees toward inconsistent practices. Executives should define approved use cases alongside the controls that apply to them.

Paul Farnsworth, president of Dice, said effective policies give employees “clear guidance on how to use AI responsibly and confidently in their work.” Dice applies that principle internally. Its AI policy encourages employees to use AI when it can improve productivity while establishing guardrails for data security, confidentiality, and human oversight.

Executives can make this guidance concrete. Policies can specify where AI may support drafting, summarization, research preparation, analysis, routine administrative work, or other approved workflows. Each use case can include clear requirements for data handling and output verification. High-impact activities may require additional human review or approval.

This approach also creates a better basis for measuring value. Leaders can evaluate specific workflows using total time saved, output quality, error rates, review effort, and employee experience. The review effort matters because apparent productivity gains can shrink when employees spend significant time checking and correcting generated content.

Policy should evolve as evidence accumulates. Successful use cases can be expanded to other teams. Weak ones can be redesigned or discontinued. New risks can trigger stronger controls. Employee feedback can identify workflows where AI reduces repetitive work and areas where the technology creates additional effort.

The executive objective is controlled adoption with measurable business value. Employees should understand where AI can help, how to use it responsibly, and when human judgment is required. A policy that provides those answers gives the workforce practical operating guidance while giving management clearer control over risk and returns.

In conclusion

AI policy is an operating decision as much as a governance decision. Employees determine how AI enters daily workflows, where it creates value, and where new risks emerge. Their input gives executives better information for setting effective rules.

The strongest approach combines employee participation with clear accountability. Define approved uses. Protect sensitive data. Keep humans responsible for consequential employment decisions. Commit to continuous training. Update policies as tools, workflows, and risks change.

Executives should also measure the full cost of AI adoption. Productivity gains matter after accounting for verification, corrections, additional workload, and employee experience. Usage alone is a weak success metric.

The goal is controlled adoption that produces measurable value. When employees understand what AI can do, where its limits sit, and how their roles will evolve, companies have a stronger foundation for scaling AI with trust and accountability.