Agentic AI and the new rules of responsible use

Agentic AI represents a leap forward in how AI systems operate. Its ability to act independently opens up new possibilities for productivity and new questions for responsible use.

These systems can independently decide on approaches and pursue objectives purposefully. They learn from experiences, analyze options, adapt to new situations and choose the right direction to achieve business objectives.

The autonomous nature of agentic AI means we’re giving technology significantly more decision-making authority. The technical, business, and reputational risks grow alongside the capabilities. Your data needs stronger safeguards. Your governance needs tighter frameworks. Responsible AI principles are how you capture the benefits of agentic AI while protecting what matters most.

The principles of responsible AI

With responsible AI, you commit to the principles of fairness, security, accountability, inclusivity, sustainability and transparency. By monitoring what goes into AI models (the inputs) and what comes out (the outputs), you manage risk while harnessing agentic AI to manage and modernize applications, data, and cloud environments.

These principles apply to all AI types. The “why” to prioritize responsible AI remains the same, including satisfying government regulations and internal guidelines governing AI use. However, the risk is increasing as AI is integrated into more tools. Even if you’ve been using a piece of software for years, you will need to do an AI risk assessment if new AI capabilities are introduced.

A mindset shift toward more responsibility

Since Unisys shared reasons to prioritize responsible AI, momentum has grown in the broader tech community to adhere to data privacy regulations. A colleague calls this public shift "awareness and wariness." More headlines come out about major companies that had AI models that put customer data at risk of exposure. Deepfakes evolve, deepening interest in robust governance. More organizations are seeking AI risk assessments so employees can use new technology securely and safely.

The Artificial Intelligence Act, a European Union regulation that establishes a regulatory and legal framework for the technology, was rolled out in 2024. Although it appears no organizations have faced penalties yet, it’s a matter of time before one will. Globally, copyright infringement cases related to AI use are on the rise, and the US government is considering more stringent regulations for AI use. AI governance platforms and technology solutions are emerging to provide you with more visibility into the regulatory landscape and potential risks.

For more responsible AI, you need the right processes, training and technology to ensure people behave in the right way, as well as the right technology to establish checks and balances in situations where people may make an error. Ultimately, people are responsible for AI’s output, so it’s crucial to train them on different levels of risk, from the higher risk of using people’s personal information for an HR implementation to the lower risk of using generative AI to write an email. Developers require faith in their AI tools as they use them to build applications.

Strategies toward more responsible AI

Responsible AI is about what you do daily to promote safe and ethical AI for the good of society and your customers, employees and partners.

These are among the most effective strategies for implementing responsible AI in your organization and gaining more peace of mind regarding AI use:

• Establish responsible AI usage guidelines: Include responsible AI guidelines in your company policies to minimize risk. You can start with the National Institute of Standards and Technology (NIST) frameworks, such as the AI Risk Management Framework, and utilize playbooks like the NIST AI RMF Playbook for those in the U.S. or the ISO/IEC 42001 and IMDA for global organizations. Take it a step further by encouraging your partners and vendors to use AI responsibly and by publicly committing to responsible AI on your website.
• Prioritize human oversight: Although agentic AI has humanlike awareness and reasoning powers, people are the technology’s architects, and human oversight is important. Human-in-the-loop involves a human operator overseeing final decision-making, observability, explainability, bias detection and other AI output components.
• Encourage employees to use AI responsibly: Everyone is responsible for responsible AI, from C-suite executives strategizing your organization’s AI approach to the professionals operationalizing its use. Train your employees on what responsible AI is and why it’s crucial. Include responsible AI practices in change management efforts as you encourage employees to use new AI tools.

Most employees don’t intend to harm when they use AI, and training can decrease their unintentional errors, whether that’s intellectual property sharing, copyright infringement or the consequences of using the “wrong” data. It also helps to explain how guidelines protect the company.

Promote responsible use of agentic AI

Agentic AI introduces new risks, but responsible AI guidelines, practices, and policies act as business guardrails. Exploring its uses can reveal opportunities. Agentic AI platforms can scan workloads for adjacent or connected workloads required in a business continuity plan, identifying business continuity risks due to incomplete replicas of data and applications. Agentic AI can also perform targeted, well-defined tasks such as identifying data, remediating or transforming data, and placing data into “data lakehouses” in preparation for a wider agentic platform-driven business intelligence deployment.

For strategies to support your AI implementation, download our data-readiness guide and reach out if you’re ready to explore AI solutions from Unisys.