How New Jersey Is Adopting AI: State-Level Results and Lessons for Small Businesses

Updated October 15, 2026. This is an independent summary and business analysis of Mickey McCarter’s September 29 StateTech interview with Dave Cole and the official NJ AI Assistant project page. The featured image is an editorial illustration, not a screenshot of the state tool.

When people hear “government AI,” they often picture a chatbot answering every resident’s question. New Jersey’s story is more practical: help public employees get routine work done, reduce avoidable friction, and keep people responsible for the results. That matters to residents waiting for help—and to business owners waiting for an application to move forward.

In StateTech’s interview, New Jersey Chief Innovation Officer Dave Cole explains how the state moves from an identified problem to research, testing, and repeated improvements. Generative AI supports that process; it does not replace understanding the people using a service. Below is what the sources actually report, what they do not establish, and how New Jersey businesses can apply the same discipline.

New Jersey’s AI strategy: five key takeaways

• Start with the problem. Cole emphasizes understanding who is affected before designing a solution.
• Give staff a safer place to work. The NJ AI Assistant is an internal state-employee tool, not a public chatbot.
• Pair access with education. The tool launched alongside training on privacy, safe use, and bias.
• Improve the unglamorous tasks. Document checks, clearer messages, and call analysis can improve service without handing over decisions.
• Measure what changes for people. Response speed, document-review time, and resolved calls are more useful than a tool-launch announcement.

What is the NJ AI Assistant—and who can use it?

The NJ AI Assistant is a generative AI platform that offers New Jersey state employees a secure way to improve government services and operations. The official project page credits work with the NJ State Office of Information Technology and lists the NJ State Civil Service Commission as another key partner.

It launched in July 2024. Enhancements in March 2025 included simultaneous file uploads, larger file sizes, and additional file types. In March 2026, it was relaunched with more robust infrastructure informed by user research, an updated model, in-app responsible-use resources, and more ways to get help. The project page says the upgrade leveraged the open-source community using LibreChat.

Access is restricted to state employees. A restaurant owner in Newark, an HVAC operator in Monmouth County, or a dental practice in Jersey City should not mistake the project page for a public AI service they can join. It explains a government program; it does not provide business owners with an assistant account or an official answer about permits, taxes, or benefits.

For the most accurate description and any later updates, go directly to the New Jersey Innovation Authority’s AI Assistant page.

What results has New Jersey reported?

The interview and the official project page describe useful outcomes, but their figures are not identical. Keeping the source and measurement attached to each number prevents a neat-looking summary from becoming misleading.

Why do some percentages differ? The interview reports 30% faster responses; the project page reports 35%. It also describes a different call-center measure from the ANCHOR result. Neither source reconciles those figures. We therefore present them separately rather than averaging them, treating them as contradictory proof, or claiming a statewide performance improvement.

The official page also reports satisfaction above 80% and an in-house cost of about $1 per user per month, compared with about $20 for off-the-shelf commercial licenses. It estimates first-year savings in the millions. This is the state’s reported comparison—not a small-business price quote, a complete independent cost audit, or proof that every company should build its own platform.

Three real applications worth understanding

1. Finding document problems before they become long delays

Cole describes small businesses uploading documents for New Jersey Economic Development Authority programs. When the wrong file is submitted, manual review can mean days or weeks before the application is returned. Staff using AI to look for common document issues can shorten that part of the process.

There is an important distinction: feedback to applicants at upload is described as a future iteration in the interview. It should not be advertised as a feature already available on every NJEDA application. Nor is faster document review the same as faster final approval. Applicants should follow the current instructions for their specific program on NJEDA’s official website.

2. Turning confusing notices into clearer communication

With the Department of Labor, generative AI helps rewrite unemployment-insurance email templates. The goal is not fancier writing: it is making it easier for a recipient to understand what to do next. Cole links faster responses to faster processing and issuance of benefits.

The transferable idea is simple. If customers routinely misunderstand a request, first improve the request. For a local business, that might mean a clearer estimate follow-up, appointment reminder, or explanation of the documents needed before a visit. The factual details, deadlines, and obligations still need human checking.

3. Learning from call transcripts instead of guessing

New Jersey uses AI to analyze transcripts and improve call menus. The official project page describes ANCHOR property-tax callers finding what they need through better self-service options rather than waiting for an agent. The StateTech interview separately reports agents handling more callers in one large call center.

That is different from saying an autonomous AI answered every call. An owner can borrow the analysis approach: identify recurring questions, confusing instructions, and missing information, then improve the website, menu, or staff script. Use only transcripts you are permitted to process, with suitable privacy and recording safeguards.

How New Jersey balances AI adoption with safeguards

Cole recognizes that employees will encounter AI through familiar apps and sales pitches whether an organization has a plan or not. Pairing practical training with an authorized tool helps reduce unauthorized use—often called “shadow IT”—rather than merely telling people not to experiment.

The official project page describes hosting on state infrastructure, government-friendly terms of service, no use of state data to train third-party models, filters for potentially harmful input or output, and protections against common attempts to bypass safeguards. Training addresses safe use, privacy, and reducing bias.

These controls do not make an AI response an official truth. Cole warns that generative AI can confidently produce fictitious information and reflect biased inputs. People still need to decide when to use it, verify the output, and know when not to use it. A polished answer is not evidence that a rule, citation, or eligibility conclusion is correct.

What New Jersey’s AI rollout means for local businesses

A state program is not a template to copy wholesale. A five-person auto shop does not have the same infrastructure or responsibilities as a state agency. What transfers is the discipline: choose a recurring problem, provide a safe environment, teach people how to work in it, and measure service improvements.

The following are our recommendations, not examples reported by New Jersey and not promised outcomes.

For more local examples, read our Monmouth County small-business AI guide. If your immediate problem is a missed inquiry, start with speed-to-lead workflows rather than a company-wide AI overhaul.

A practical 30-day AI pilot inspired by New Jersey

This is an independent business plan, not a state-sponsored program. It is designed to answer one question: does this workflow save time without making service less reliable?

1. Days 1–3: Pick one bottleneck. Ask staff which routine task causes repeated delays. Choose something narrow, such as rewriting an approved customer message. Record the current time per task, number of corrections, and customer questions.
2. Days 4–7: Set the boundaries. Choose an approved tool and inspect its data-retention and training terms. Decide what must never be uploaded, who reviews output, and what cases always go to a person. Use public or synthetic examples first.
3. Days 8–14: Test with a small group. Compare AI-assisted drafts with your existing process. Ask staff to flag wrong details and unclear wording. Teach them to verify answers rather than trust the tool’s confidence.
4. Days 15–21: Improve the workflow. Refine the template and checklist. Add escalation rules. If errors keep recurring or review takes longer than the original task, narrow the use case or stop.
5. Days 22–30: Decide from evidence. Compare time saved, correction rate, and customer outcomes against the baseline. Include subscription and review costs. Expand only if the benefit is real and the safeguards work.

A useful measurement is net time saved = time avoided − time spent checking and correcting. Pair it with an accuracy measure. A fast workflow that creates wrong appointments or misleading messages is not a successful pilot.

Need help choosing a tool? Our ChatGPT, Claude, Gemini, and Perplexity comparison explains everyday uses. For the broader rollout process, see the small-business AI automation guide.

What this has to do with AI search visibility

Government productivity and AI search visibility are different topics. The interview does not establish that state AI adoption improves local businesses’ Google rankings or makes chatbots recommend them. That connection should not be invented.

There is, however, a useful publishing lesson: make your information clear, current, specific, and easy to verify. Put real opening hours, service areas, policies, and answers to common questions on your website. Keep those facts consistent across customer-facing channels. This helps people understand your business and gives search and answer systems clearer source material—without guaranteeing rankings or citations.

If being found is your next priority, explore free AI visibility tools for local businesses and our answer engine optimization and generative engine optimization resources.

What not to assume from this story

• No public-access promise. The NJ AI Assistant is for state employees.
• No automatic approvals. Document screening is not a grant, loan, or eligibility decision.
• No identical savings for every business. State-level cost comparisons and service results are not your ROI forecast.
• No replacement for official guidance. Verify rules with the responsible agency and seek qualified advice when appropriate.
• No evidence that these are statewide totals. The cited outcome measures refer to particular work areas or reported usage.

The most valuable takeaway is not “use AI everywhere.” It is “use AI where it makes a real service better, and keep checking whether it does.” That is a strong starting point for a public agency in Trenton or a Main Street business in Red Bank.

Frequently asked questions about New Jersey AI adoption

Sources and further reading

• StateTech: “Q&A: New Jersey’s Chief Innovation Officer Lifts the Curtain on Launching AI Projects” — Mickey McCarter, September 29, 2026. Primary source for Cole’s approach, document-review example, and interview-specific results.
• New Jersey Innovation Authority: NJ AI Assistant — official project overview, last updated March 25, 2026. Source for access, safeguards, rollout milestones, dated usage figures, and separately reported metrics.
• Official announcement: New Jersey upgrades its AI Assistant for state workers — March 24, 2026; linked from the official project page for further reading.
• Responsible Use of Generative AI in New Jersey State Government — the training resource linked by the project page; check the current page for access and course details.
• New Jersey Economic Development Authority — consult current program-specific requirements before applying.

Editorial note: This article summarizes and attributes the original reporting rather than reproducing the interview. Business examples and the 30-day pilot are our analysis. No affiliation with or endorsement by StateTech, NJIA, NJEDA, or New Jersey state government is implied.