Kentucky state government is already using AI across more than 15 departments - here's what to do next
LOUISVILLE, Ky. - More than 15 Kentucky state departments are using artificial intelligence in some capacity. That figure comes from the Commonwealth Office of Technology (COT) in response to an Open Records Request.
COT noted that some AI features weren't turned on by special projects - they arrived inside tools the state already uses as the technology matured. That means usage may be wider than it looks, even without formal "AI projects" on the books.
Where AI is likely showing up today
- Office Tools: autocomplete, document suggestions, meeting summaries, search enhancements.
- Help desks and contact centers: ticket categorization, routing, knowledge-base answers.
- Document workflows: OCR, summarization, form extraction, translation.
- Accessibility and language services: transcription, captioning, multilingual support.
- Data work: pattern detection, anomaly flags for audits, grants, and benefits integrity.
- Developer tools: code assistants and test generation inside IDEs.
- Security operations: alert triage and incident summarization inside existing platforms.
Because many of these features are embedded in common platforms, agency use can grow quietly. That's why inventory and guardrails matter now, not later.
What this means for your cabinet
You don't need to greenlight a big program to be exposed to AI. It's already in daily workflows through existing vendors. Your job is to make that usage safe, auditable, and useful.
Focus on three things: know where it's used, protect sensitive data, and measure impact. If you do those well, you'll capture value without adding risk or noise.
Immediate actions to take this quarter
- Run an AI inventory: ask each division which tools use AI features (even "auto" or "smart" options). Capture system, use case, data types, and owners.
- Classify data allowed in AI tools: public, internal, confidential, restricted. Block restricted data from external models.
- Turn on enterprise controls: vendor admin settings, audit logs, data retention, API restrictions, and export reviews.
- Issue a plain-English AI use memo: what's OK, what's not, where to ask for help, and how to report issues.
- Stand up a lightweight AI review board: legal, COT, privacy, security, program leads. Weekly, 30 minutes, decision-driven.
- Pilot 2-3 narrow use cases with clear metrics: case note summaries, meeting notes, inbox triage. Track time saved and error rates.
- Update records retention and Open Records guidance for AI-generated content and transcripts.
- Add AI clauses to contracts: data ownership, training rights, PII handling, model changes, uptime, and deletion on exit.
- Require human-in-the-loop for decisions with legal or financial impact. Document the review step.
- Provide quick training: privacy do's/don'ts, good prompts, and how to check outputs.
Policy guardrails to finalize
- Prohibited inputs: no PII, health data, CJIS data, or confidential legal materials in external models.
- Disclosure: label AI-generated drafts; require human sign-off for public releases.
- Bias and accessibility checks for any public-facing AI feature.
- Security: vendor SOC 2/StateRAMP equivalents, breach notice timelines, and logging requirements.
- Change management: re-review when vendors flip new AI features on by default.
Procurement and vendor management
Ask vendors to confirm which features use AI, where data is processed, whether your data trains their models, and how you can opt out. Lock these terms in writing.
Require a system-of-record plan for any AI outputs used in decisions. If an AI tool influences eligibility, payment, or enforcement, you need a clear audit trail. For role-specific policy and privacy training for regulatory teams, consider the AI Learning Path for Regulatory Affairs Specialists.
Workforce and training
- Start with frontline tasks that are repetitive and low risk: summaries, drafts, routing, cleanup.
- Teach staff to verify sources, spot model errors, and keep sensitive data out of prompts.
- Set expectations: AI speeds up drafting; staff own accuracy and final judgment.
If your team needs structured upskilling for prompts, office tools, and automation, see curated options by role such as the AI Learning Path for CIOs.
Transparency and records
- Keep a public-facing summary of where your cabinet uses AI and why. Update quarterly.
- Document prompts, settings, datasets, and human review steps for any high-impact use case.
- Ensure AI-generated content is stored and searchable per retention schedules.
Useful references
Bottom line: AI is already in your stack. Treat it like electricity in the walls - map it, meter it, and make it safe. Then point it at the work that frees your people to focus on service and outcomes.
Your membership also unlocks:
