The Boundary Call: What an Indiana State Agency Got Right About AI in 2026

What Indiana built.

Money Quest replaces the outdated state website materials the securities division had been running on financial education. The platform is responsive to the user's device and adaptive to the user's persona. Six personas are wired in: K-12 students, young adults, working adults, people nearing retirement, seniors, and educators. Each persona gets a different learning pathway through the same core curriculum.

The build partner was Venturit, a Michigan software firm. The underlying AI system is Google Vertex AI. Phase 1 went from initiation to completion in about four weeks at a cost of roughly fifty thousand dollars. Fulk's exact testimony in the StateScoop piece was: "We literally did this in about four weeks to completion. … We would never be able to do that without AI." On cost: "We literally probably did this phase one of this thing for fifty grand, and this would easily [have] been a quarter million dollar project."

The line Fulk drew.

The load-bearing detail in the StateScoop piece is the boundary call. Inside Money Quest, AI handles a specific list of functions: graphics, audio, video production, generation of short modules and adaptive media from existing curriculum, and device-responsive content adaptation. AI does not handle the financial substance. Core financial guidance stays on the human side of the line, along with the investment-advice and fraud-prevention decisions adjacent to that guidance.

That distinction is what makes the project worth a careful read. The securities division of a state Secretary of State office is, by statute, a guardian against investment fraud. They are also a public-facing source of financial information for residents who do not have a financial advisor on retainer. The cost of an AI hallucinating into investment-advice territory inside a platform with that agency's name on it is meaningful. Fulk's office took the risk seriously enough to engineer the boundary on the front end, before turning AI loose on the production side at all.

Why production-yes, guidance-no is the right call here.

The two functions AI handles well in 2026 and the two it handles poorly map cleanly onto Fulk's split.

Production work, including video generation, graphic design, audio production, and content adaptation across formats and personas, is solidly inside the capability envelope of current AI systems. The work is creative-but-bounded. The output can be quality-checked by a human reviewer before publication. Errors are usually obvious and correctable. The cost and speed advantages over traditional production studios are real.

Authority-bearing financial guidance is in the opposite category. The output has to be correct in substance, not just well-produced. The errors that current AI systems are most prone to (confident-sounding fabrication, plausible-but-wrong specifics, drift from underlying source material) are the exact errors that a regulated public agency cannot ship inside investment advice. A wrong answer on the production side costs a re-render. A wrong answer on the guidance side could put someone's retirement at risk and put the agency's name on the harm.

Fulk's call separates the two cleanly: AI takes the production budget while humans hold authority over the substance.

The boundary work was done first. The four-week build came second.

Read it through The 7 Levels of AI Proficiency.

The Money Quest decisions are operating decisions a Level 4 Commander makes day to day, with a Level 5 Architectural Strategist read showing up in the broader portfolio thinking the office is signaling.

A Level 4 Commander knows which functions translate cleanly to AI-directed execution and which functions require human authority. The boundary-call discipline is the L4 move. It is the leader who can look at a portfolio of work and say, with conviction, "AI handles this category, humans hold this category, here is how we know the difference, and here is the operating system that keeps the line clean." Fulk's office did exactly that work before the four-week build started.

A Level 5 Architectural Strategist runs the same boundary-call discipline at the portfolio level. The Phase 2 plans Fulk described, which include connecting K-12 and collegiate educators to the platform and updating the notary licensing workflow with AI-assisted application review, signal an office that is treating AI as a portfolio question across operations, not a one-off project. The projected sixty to seventy percent workload reduction in the back-office notary process is the kind of compounding return that L5 thinking generates when applied to multiple workflows at once.

What an Indiana CEO should take from this.

The Money Quest story does not require an Indiana CEO to do anything urgent in the next 30 days. It does invite a careful look at the active AI projects already underway inside the company, with one specific question in mind: where exactly is the line between functions AI is allowed to handle and functions humans must hold authority over? If the answer to that question lives only in someone's head, the line is not yet operational. If the answer lives in a written policy that the team can point to, it is.

Indiana operating CEOs running 200 to 5,000 person companies will face this design question across multiple workflows in 2026, including customer-facing chat, internal sales analysis, contract review, hiring funnel triage, financial reporting drafts, and regulatory filing preparation. Each of those workflows has a production component AI can usefully accelerate alongside an authority component humans must keep. The Fulk pattern (engineer the boundary first, then turn AI loose on the production side) is the version that protects the company from worst-case errors while still capturing the cost and speed lever.

The companies that do this design work deliberately end up with saner answers than the companies that arrive at the inquiry reactively, whether after a competitor has already executed against it or, worse, after a public AI mistake has put the company's name on a harm.

A caution.

Money Quest is brand new. Adoption rates, learning effectiveness, and fraud-prevention impact have not been measured at publish. The fifty-thousand-dollar cost and four-week timeline are Fulk's testimony from one project; whether the cost holds when Phase 2 ships and the system has to scale is unknown. The boundary line between production and financial guidance has only been live for a short window; whether the line holds operationally over time, especially as users push the system to answer questions adjacent to investment advice, is also unknown. Anyone telling you Money Quest is the model for AI implementation in state government is moving faster than the data supports.

What the Money Quest announcement provides is a real, named, specific Indiana example of operator-discipline applied to a public-sector AI implementation, with the boundary call doing the load-bearing work, the cost and speed claims providing useful supporting context, and the unmeasured outcomes setting the honest limit on what the story can say in May 2026.

Related reading: The 7 Levels of AI Proficiency.

Harrison Painter

AI Business Strategist. Founder, LaunchReady.ai and AI Law Tracker.

Harrison helps teams build AI systems that cut cost and grow revenue. Nearly 20 years of business experience. 2.8M YouTube views. Founder of LaunchReady.ai and the 7 Levels of AI framework.

Connect on LinkedIn