Responsible AI for a state transportation agency

The Context

A state transportation agency runs on data the public depends on and regulations the public demands. When enterprise AI arrived, VDOT faced the same pressure every large organization did — do something with AI — but with stakes most enterprises don’t carry: statewide mission-critical platforms, regulated data, and public trust. The failure mode wasn’t just a bad pilot; it was a headline.

Readiness Before Rollout

The strategy started where responsible adoption has to start: an honest readiness assessment. Before any use case was funded, I assessed the agency’s data and technology readiness — where the data actually lived, what condition it was in, and which platforms could carry AI workloads inside the agency’s security and compliance boundaries. In parallel, I evaluated Azure AI platforms against those constraints, so the recommendation was grounded in what the agency could operate, not what a vendor deck promised.

Governance That Enables

The governance work reframed a common trap. AI governance in the public sector usually shows up as a wall of no; here it was built as the thing that makes yes possible:

• Usage guardrails: clear boundaries on what AI could touch, which data could reach which models, and where human review was mandatory.
• Evaluation criteria: every proposed use case measured the same way — value, risk, data sensitivity, and operability — so prioritization was a decision process, not a popularity contest.
• Architecture standards: reference patterns for how AI capabilities integrate with agency platforms, keeping adoption aligned with security, compliance, and regulatory requirements from the first design review rather than retrofitted after.

Because the guardrails were explicit, product lines could move inside them with confidence — governance as pavement, not speed bump.

Finding the Value

Strategy documents don’t ship anything. The adoption work happened in direct partnership with product line leaders, walking their operations to identify and prioritize high-value AI use cases against the shared evaluation criteria. That work fed enterprise architecture research that influenced the statewide digital strategy and supported the approval and launch of three innovation initiatives — funded because the readiness, governance, and value questions had already been answered.

The Impact

VDOT’s AI posture moved from ad-hoc curiosity to a governed adoption pipeline: a readiness baseline, an evaluated platform direction on Azure, guardrails that satisfy security and compliance by construction, and a prioritized portfolio with three initiatives through approval. It’s the same discipline I bring to every architecture problem — the technology is the easy half; the operating model around it is what determines whether it survives contact with reality.