AWS team briefs Wake County board on generative AI, urges data-first, use-case approach

Amazon Web Services staff delivered a briefing on artificial intelligence and machine learning to the Wake County Board of Education during the Sept. 16 work session, framing generative AI as a rapidly adopted technology that boards should approach with clear use cases, robust data practices and ethical guardrails.

"My name is Ann Marie Lehner and I'm from Amazon Web Services. And I'm the K‑12 strategy leader at AWS," said Ann Marie Lehner, introducing the AWS team and its educational specialists before turning the presentation to Mary Strain, AWS's AI and machine learning strategist for education.

Strain summarized how generative AI differs from earlier AI work and emphasized its reliance on very large, unstructured data sets and foundation models. "Artificial intelligence has reflected terrible bias," she said, adding that those biases typically come from the data used to train models. She told board members that generative AI can produce useful automation and productivity gains — but that districts must plan for hallucinations, privacy and equity concerns.

Why it matters: presenters said generative AI adoption has outpaced other consumer technologies and that districts face choices about embedding vendor products, building custom tools, or combining both approaches. Mary Strain urged Wake County Public School System (WCPSS) leaders to begin with problems to be solved rather than technology for its own sake, and to prioritize data access, security and flexible tooling.

Key points from the briefing: - Pace of adoption: Strain noted ChatGPT reached 100 million users in 2 months; by contrast, many other widely used platforms took years. She said dozens of new foundation models are published monthly, and districts should plan for rapid change. - Data and knowledge graphs: Strain described knowledge graphs (she cited a University of Delaware example) as a way to link disparate student, course and operational data to generate insights that a human would not easily see. She emphasized that good outcomes depend on accessible, well-governed data. - Technical choices and guardrails: AWS presenters recommended flexibility to swap or combine foundation models, use secure customization to improve accuracy, and adopt guardrails for bias mitigation, privacy, transparency and explainability. - Pragmatic use cases: presenters outlined three broad buckets of k‑12 use: teaching and learning (tutor bots, personalized professional learning, automated classroom observation support), operations and community engagement (translation, meeting summaries, enhanced access), and administration (HR onboarding, document processing, enrollment automation). Strain highlighted intelligent document processing — for example, extracting data from registration PDFs — as a high‑ROI, "boring AI" use case. - Agents and workflows: the team described agents (semi‑autonomous software that completes tasks) as the next wave of value beyond single chat interactions: agents can ingest uploaded documents, extract data, route items to staff, and complete end‑to‑end workflows such as registration or IEP document handling.

Presenters also identified risks and mitigations. They listed common failure modes — hallucinations and inaccurate outputs, gaps in language coverage (five primary language groups dominate many models), copyright and intellectual property disputes, and data security concerns for student information. They recommended alignment with responsible‑AI frameworks (presenters cited EU frameworks, the UN commission guidance and U.S. Department of Education guidance) and suggested the Council of the Great City Schools' generative AI assessment as a self‑diagnostic tool.

Board members asked how the district should balance adoption and classroom integrity. Board member Rice said she supports AI but stressed the district must teach students to use it responsibly. Superintendent and other board members raised concerns about customer service and the effect of automation on human interactions.

Next steps presented: AWS offered follow‑up sessions focused on technical implementation and data security, and recommended that WCPSS (1) identify a small number of high‑value, measurable pilot use cases, (2) inventory and modernize data sources needed for those pilots, and (3) create governance aligned to district values.

The presentation closed with an offer of tools and open‑source pilots AWS and partners use with districts and higher education institutions. The AWS team said they will continue the conversation in subsequent sessions already scheduled with district staff.

Ending: presenters and board members said the topic will inform forthcoming policy work. No formal action or policy adoption occurred at the session; the briefing was informational.