Start with a Model
The Equitable AI Playbook encourages organizations to consider a hub-and-spoke model for their Equitable AI initiative. In a Hub-and-spoke model, a central group (“Hub”) is led by C-Level and establishes standards, processes, and policies. “Spokes” are business unit or function teams that oversee execution of the policies and processes by implementation teams.
This type of model has typically worked well for implementing other organization-wide priorities (e.g., privacy, security, and accessibility) and can enable collaboration among the following organizational components and roles. The existence and specific names of these components and roles will vary based on your organization’s size, maturity, and other factors.
Establish a Central Hub
An Equitable AI Hub should be established to serve as the heart of your organization’s Equitable AI Initiative and provide three key functions. First, the Hub can set policies, develop resources to help implement policies, put in place governance mechanisms, and monitor the overall Initiative. Second, the Hub can engage a set of cross-functional in-house experts who can help identify issues and define technologies, processes, and promising practices. Third, the Hub can develop communications, training, and professional development components to embed Equitable AI practices across the whole organization. The Hub should also:
- Monitor industry challenges, technologies, processes, and best practices.
- Draw on internal expertise as well as the broader community of AI researchers to translate research into actionable components of your initiative.
- Build and regularly review and update the organization’s Equitable AI business case to ensure executive support.
- Develop both internal and external policies that will communicate the organization’s commitment to Equitable AI.
- Advise and make recommendations to senior leadership on Equitable AI issues, technologies, processes, and best practices, calling special attention to sensitive use cases.
Define Your Stakeholders
There are four stakeholder categories to consider when putting in place your Equitable AI Initiative: Advisors, Champions, Implementers and Impacted End Users. Use these stakeholder categories to implement Play 3: Identify and Categorize Stakeholders. Each are described below.
Advisors are senior executives, business managers, and technology leaders—including any internal subject matter experts, and external consultants or researchers—who will build, lead, and advise your organization’s Equitable AI initiative. Advisors can help shape your initiative, connect it to other strategic programs, determine a governance strategy, review outcomes, and support effective cultural change. Advisors can include:
- Your Board of Directors or Advisory Committee should set an organization-wide commitment to Equitable AI in line with your organization’s Corporate Social Responsibility, Diversity/Equity/Inclusion, and Accessibility Programs. They should also ensure that Equitable AI Principles are adopted and communicated, that the CEO and entire C-Level team are informed and engaged, and that progress in implementing Equitable AI is being monitored to mitigate and reduce risk.
- Your CEO or President should ensure that internal or external human and material resources are in place to enable planning, implementation and monitoring of the organization’s overall progress in implementing Equitable AI. This includes appointing a Chief AI Ethics Officer who has voting power on the Board or Advisory Committee.
- A Chief AI Ethics Officer should lead your centralized Equitable AI Hub and keep all other senior leaders—including the CEO and Board—informed of the initiative’s progress. Titles for this role will vary (e.g., “Chief Responsible AI Officer”, “AI Ethics Lead”, etc.), and in some organizations the role might be combined with roles related to data, privacy, security, or risk.
Champions are senior leaders and influencers who will promote the Equitable AI initiative across your organization. Champions will lend their voice and serve as “spokespeople” of your initiative, helping employees understand and adopt practices required to implement Equitable AI. Champions can include:
- Employee Resource Groups (ERGs) should be engaged to look at the impact of AI workplace technologies on people of varying ages, gender identities, racial and ethnic backgrounds, sexual orientations, and people with sensory, physical, mental health, and cognitive disabilities. ERGs can also help co-design the AI implementation process, review challenges, and ensure equitable use of AI to mitigate bias related to factors such as race, gender, class, age, and disability.
- Disability ERGs can help define specific benefits and risks of AI technologies for people with disabilities, help determine if user interfaces/training/support are accessible and provide input to reasonable accommodation practices for people with disabilities. Don’t have an ERG for employees with disabilities and allies? Start a new one.
- Diversity, Equity, Inclusion and Accessibility (DEIA) Initiatives should be engaged in Equitable AI, as they can help with building more diverse teams, defining requirements, performing audits, and assessing the fairness of outcomes of AI. There is a growing recognition that DEI initiatives should be involved AI development and implementation.
- Digital Accessibility Programs should be engaged by the Hub to help determine how accessible AI-enabled workplace technologies conform to established digital accessibility standards, provide input on accessibility factors, and evaluate how usable they are for people with disabilities.
- HR, Training & Employee Development should help the organization acquire, train, and develop employees so they can understand and adopt practices required to implement Equitable AI. An employee professional development program should be put in place in collaboration with the Equitable AI Hub so that Implementers can become Champions as they grow their expertise in Equitable AI.
- Procurement should follow Equitable AI governance and approval procedures and work with implementation teams to assess risk and perform Equitable AI risk audits in collaboration with vendors. When Equitable AI risk audits are conducted, they should outline intended uses for the technology, anticipated benefits, and an analysis of specific risks—including data and privacy risks, as well as risks of discrimination and bias—for people with disabilities and other protected classes. Results of Equitable AI risk audits should be reviewed with the Hub before proceeding with purchases and implementations. The Hub and Chief AI Ethics Officer should have the authority to delay or halt AI implementations that have an outsized risk.
- Legal, Risk & Compliance staff should ensure all legal risks are assessed and addressed in AI workplace technology implementations. They should also ensure the organization complies with all applicable laws and put in place a compliance recordkeeping system.
- Other Departments and Business Units should identify and groom Implementers who will serve as Champions in distinct areas where AI is being implemented. For example, HR might have a Champion who specializes in Equitable AI for hiring or employee development technologies, and Customer Support might have a Champion who specializes in Equitable AI for help desk technologies.
Implements are people responsible for executing your Equitable AI processes and procedures on a day-to-day basis. This includes people who will plan, procure, design, configure, integrate, monitor, and support AI technologies your organization deploys. Implementors may be internal employees, contractors, or external consultants helping with AI implementations. Implementors are not necessarily the end users of an AI tool, though some might be. For example, an Implementor tasked with building a machine learning model could also be one of the end users who interprets the model’s results. Implementers may include:
- Internal staff and contractors
- External consultants, such as:
- Vendors, partners, and suppliers
- AI Ethics & Equity researchers
- DEIA & Accessibility Consultants
Impacted End Users
Impacted end users are people who ultimately use (or are intended to use) AI implemented by your organization. End users can ultimately benefit from your organization’s Equitable AI implementations —whether by using them directly (e.g., communicating with a chatbot) or indirectly (e.g., receiving highlights from automated data analytics). It is important to gather input from these groups of end users and ensure their needs are considered when planning or implementing Equitable AI. For example, some Advisors, Champions, and Implementers should be end users with a tangible grasp on how AI tools work in practice. End users might include:
- Prospective employees
- Business partners