Companies are using data and artificial intelligence to develop scalable solutions — but they’re also increasing their possible financial, legislative, and legal risks. For instance, Los Angeles is suing IBM for reportedly defrauding data collected with its popular weather app. Regulators are investigating Optum for developing an algorithm that allegedly recommended that white patients be paid more attention to by doctors and nurses than sicker black patients.

Today, the world’s largest tech companies, Microsoft, Facebook, Twitter, Google, and more, bring together fast-growing committees to resolve the ethical issues that emerge from the massive collection, analysis. Also, in the use of enormous data troves, mainly when that information is used to train machine learning models, aka AI.

How Data and AI Ethics are Operational

A data and AI ethics policy must be customized to the unique market and regulatory needs critical to the enterprise, considering the differing standards of firms across thousands of industries. However, there are seven steps toward creating a personalized, operationalized, flexible, and sustainable software of knowledge and AI ethics.

#1 Find Existing Infrastructure that Data can Leverage

For example, the strength and power of current infrastructure, a data governance Council that convenes to address safety, cyber, comply, and other data-related threats, are the secret to effectively developing data and an IT ethics scheme. If such an entity does not exist, businesses should build it with ethics-adjacent staff, such as cyber, risk and security, privacy, and analytics, such as an ethics council or committee. It can also be proposed that external experts, including ethicians, should be included.

#2 Build a Data and AI Ethical Risk Framework

At a minimum, a robust system requires an articulation of the company’s ethical values, including the moral fears, an identification of the various appropriate stakeholders, a proposed governance structure, and an indication of how to sustain the system in the face of shifting workers and circumstances. To evaluate the continued efficacy of the strategies that enforce your plan, it is necessary to develop KPIs and a quality assurance program.

#3 Change your Approach to Building Ethics

Often senior leaders characterize ethics in general as “squishy” or “fuzzy,” and data and AI ethics in particular, and claim it is not “tangible” enough to be actionable. Leaders should draw lessons from health care, an industry that has been systematically based since the 1970s on ethical risk reduction. Medical ethicists, health care professionals, policymakers, and attorneys have extensively discussed crucial questions regarding what constitutes anonymity, self-determination, and informed consent, for instance.

#4 Optimize Guidance and Tools for Product Managers

Although the system gives high-level guidance, guidance must be granular at the product level. Take, for example, the often-lauded significance of AI description, a highly appreciated feature of ML models that is likely to be part of your system. Standard machine-learning algorithms are too unwieldy for humans to understand pattern recognition. Product managers need to learn how to make the tradeoff and help product managers make those choices; personalized tools should be created.

It’s not a straightforward job to operationalize data and AI ethics. Buy-in from top management and cross-functional coordination is needed. However, businesses that commit will see mitigated risks and implement the technology they need to forge ahead more effectively. And hopefully, they’ll be just what their suppliers, suppliers, and staff are waiting for: efficient.