From the State Government edition of October 3, 2026
StateTech (Sep 29): NJ CIO Cole cites GenAI production metrics—30% faster UI responses, weeks-to-minutes EDA reviews, 15% more callers
New Jersey Innovation Authority / Office of the Chief Innovation Officer (Dave Cole), as interviewed by StateTech Magazine · State innovation / digital services / workforce GenAI · United States — New Jersey (statewide employee GenAI; Labor UI notices; EDA document review; Taxation call-center menu AI cited by Cole)
- Publisher
- StateTech Magazine
- Original publication
- 2026-09-29
- Source retrieved
- 2026-10-04
- Event date
- 2026-09-29
What happened
In a Sep 29, 2026 StateTech Q&A, New Jersey Chief Innovation Officer Dave Cole described GenAI moved beyond pilots: Labor used generative AI prompts to rewrite unemployment-insurance transactional email templates, and recipients responded to notices 30% more quickly, speeding claims/benefits; EDA document review for small-business applications fell from weeks to minutes with AI issue-spotting; in one large call center, AI transcript analysis for better menus helped agents handle 15% more callers. Cole also described the state-hosted NJ AI Assistant (on state IT infrastructure with training) as the practical guardrail alternative to shadow IT, with surveyed users estimating hours saved each week.
Why it matters
Strongest unarchived measured GenAI production interview after the 2026-10-01 Nevada/Maryland nomination edition. Why now: Oct 2 package was rejected on timestamp validation and never published, so Cole’s Sep 29 metrics remain eligible; pairs with Oct 3 Midwest frontier/chatbot statute reporting as the implementation counterweight.
Evidence and measured results
Opened the StateTech HTTPS article via curl (UA SLED-research-bot/1.0) and text extraction; confirmed datePublished 2026-09-29T19:30:02Z in page metadata; captured Cole's 30% UI response, weeks-to-minutes EDA, and 15% more callers statements verbatim in context. innovation.nj.gov AI Assistant page returned HTTP 403 to curl in this environment (omitted as inaccessible primary corroboration).
Limitations and uncertainty
Secondary journalism quoting the CIO; 30% and 15% lack sampling detail in the article; EDA weeks-to-minutes is qualitative compression. NJIA project page was HTTP 403 to this bot.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-03; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- States need peer production GenAI metrics for resident notices and intake—not only survey percentages.
- Stakeholders
- State CIO/Chief Innovation Officer
- Labor/UI communications
- EDA or licensing intake leads
- Contact-center ops
- Privacy/procurement
- Discovery questions
- What baseline period produced the 30% response claim?
- Who releases rewritten notices?
- Is the employee assistant state-hosted?
- Which queue is next for menu or intake AI?
- Value hypothesis
- Prompt-assisted plain-language notices and intake issue-spotting can compress resident response and review cycles if baselines and human release exist.
- Potential engagement
- 60–90 day notice/intake measurement design on one benefits or business-application queue.
- Unsupported claims
- Guaranteed 30% faster responses
- NJ results transfer unchanged
- Interview equals independent audit
- Any specific commercial chatbot is required
Pre-sales engineering
Role takeaway
- Fit
- State-hosted or non-training GenAI for employee drafting plus human-owned notice/intake workflows.
- Architecture and integration
- Governed LLM tenant to classified corpora/templates to SME review to CMS/contact-platform publish.
- Prerequisites
- AI policy and training
- Data classification
- SME owners
- Baseline metrics
- Constraints
- Secondary interview evidence
- No sampling detail for 30%/15%
- NJIA page HTTP 403 in this run
- Security
- No case PII in unauthorized tools; prefer state-hosted or contractual non-training; audit logs for prompts.
- Proof of value
- Before/after notice response rates
- Intake cycle-time sample
- SME acceptance rate of AI drafts
Delivery
Role takeaway
- Work
- Select one notice rewrite and one intake/menu pilot; instrument baselines; enforce human release; inventory the assistant.
- Dependencies
- Innovation/CIO office
- Program SME
- OIT
- Privacy
- Contact-center vendor if any
- Ownership
- Program agency owns resident-facing content; CIO sets enterprise guardrails.
- Skills and adoption
- Workforce GenAI literacy paired with tool access; reviewers trained to catch hallucinations in notices.
- Governance checkpoints
- Pre-pilot risk review
- Mid-pilot metric check
- Go-live SME sign-off
- Acceptance criteria
- Baseline and post metrics recorded
- Human release documented
- Inventory updated
- Risks
- Metric theater
- PII in prompts
- Skipping SME validation
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Pattern is state-hosted employee GenAI plus prompt-assisted plain-language rewrite and intake issue-spotting—not autonomous eligibility decisions. Keep models on state infrastructure or government-friendly ToS; wire outputs into human-owned notice templates and call-menu changes.
Governance
Who approves, reviews and stays accountable for outcomes?
Cole frames training + authorized tool as the control against shadow IT; metrics are operator-attributed in an interview, not an independent audit. Require published baselines and sample methods before treating 30%/15% as transferable ROI.
Security and privacy
What data, permissions and controls need testing?
State-hosted assistant reduces training-on-vendor-data risk versus public ChatGPT; still need classification gates so UI/EDA documents with PII are not pasted into unauthorized tools.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Plain-language UI notices and call-menu improvements target resident comprehension and access; workforce literacy training is paired with the tool so employees know when not to use GenAI.
Procurement
What should contracts, pricing and exit terms secure?
Prefer procurements that allow state-hosted or non-training ToS GenAI and require measurement plans for notice-response and contact-handle rates—not seat-count alone.
Operating model
Which teams own the service once it runs?
Human-centered problem definition first (Cole); GenAI accelerates routine rewrite/review while agencies retain ownership of notices, eligibility, and menu changes.
Publication history
- 2026-10-03State Government · Issue 275 resources
Stable resource ID: new-jersey-statetech-cole-genai-metrics-20260929