On Friday, September 25, 2026, I spent the morning at AI Quarterly: AI in Public and Private Sectors, hosted by the MKE Tech Hub Coalition at Brooksource in downtown Milwaukee. Leaders from the City of Milwaukee, Johnson Controls, and EY walked through how AI is showing up in their day-to-day work, and then the room broke into small groups to draft recommendations for a real challenge the City is working on right now.
I left convinced of two things. AI is already doing meaningful work in both sectors, and the organizations that bring it into the open, with clear guardrails and trained people, are the ones that will stay competitive in the years ahead.
The City of Milwaukee shared how it moves carefully, because it serves the public with public dollars. The City approved an AI policy through its Common Council, rolled out a government-specific AI assistant that keeps its data out of model training, and trained early adopters to act as ambassadors across departments. From there, it has taken on targeted projects, including a virtual voice agent called Front Desk for the 286-CITY call center, a website chat and search assistant, a tool for searching historical meeting records, cameras on public works vehicles that detect potholes, a partnership with MSOE to predict and prioritize tree pruning, and CareerMap MKE, an AI-powered workforce development platform that helps residents prepare for careers with the City.
Front Desk stood out to me because of who it serves. The City's plan brings voice, text, and email together so residents can reach City services the way they prefer, and it supports more than 75 languages for the more than 125,000 Milwaukee residents who speak English as a second language. As someone who works in workforce development, I appreciated that the City measures success by response time, routing accuracy, resident satisfaction, and language access, alongside cost.
The most candid example of the morning came from the Department of Emergency Communications. Quality assurance for 911 calls used to mean sampling about 21,000 calls a year at roughly an hour each, which adds up to 21,000 hours and about ten full-time staff. Their AI tool, Comms Coach, can evaluate every call. The team also shared their current numbers openly: about 35% of AI-evaluated calls still need a manual override, often because the system mishears local street names, and about 10% of call audio still needs better pairing with dispatch records. That kind of transparency builds trust, and it shows a healthy model in which AI handles the volume and people stay responsible for the judgment.
On the private side, a data and AI leader from Johnson Controls described a very different set of pressures. Their sellers work through construction specifications that can run two to three thousand pages, and AI now takes the first pass so sellers can respond to customers faster. They hold that work to a high accuracy bar, because missing a line item worth a few hundred dollars is survivable, while missing one worth a million dollars is a serious problem. He also raised a concern I hear from business owners often: when people lean on AI for everything, newer employees can skip the years of problem-solving that build real expertise. His request to the room was to learn AI and keep sharpening your own judgment at the same time.
EY connected two of the top priorities for state technology leaders, AI and legacy modernization, both at the top of the NASCIO 2026 State CIO Top 10 Priorities. The EY speaker drew a clear line between AI that produces an output, which people review, and AI that performs a job, which people supervise the way they would supervise any other role. He also shared an example from Maryland, where AI helped turn roughly a million lines of legacy code into product requirements and a queryable understanding of the system in about a week. His point was that when building gets this fast, intent becomes the real constraint, and someone accountable still has to decide what the new system should do.
Across both sectors, the same tension kept coming up. The City described a two-part dilemma, where some employees want to use their own favorite public tools instead of the approved one, while some leaders worry enough about hallucinations to hold back entirely. Johnson Controls described employees pasting company information into public chatbots. That is shadow AI, and it is the part of this conversation I care about most.
Shadow AI is what happens when people find AI useful and the business has yet to give them a safe, approved way to use it. The work still gets done, but it happens in personal accounts, with customer details, pricing, and internal documents flowing into tools that sit outside the business's view and control. Over time, that puts the very knowledge that makes a business competitive at risk.
My takeaway is that the answer to shadow AI is embracing AI on purpose. The City of Milwaukee chose an approved tool, trained its people, and built ambassadors. Johnson Controls set accuracy thresholds before trusting AI with high-stakes work. EY is pushing for AI that gets supervised like any other role. Each of these approaches brings AI into the open, where the organization can measure it and improve it over time. For small and mid-sized businesses, the same approach works at a smaller scale. Start by choosing the tools you trust and writing down how the work should be done, then train your people on both, and keep a person accountable for every result that matters.
This is the work I do every day at Dr. Data. For teams that already use public AI tools, our AI Enablement work configures the tools a business has approved, builds prompt libraries and playbooks for each role, and trains people so the safe option becomes the easy option. For businesses ready for their own system, we start by capturing how the experts actually make decisions, using our Skill Threshold Zone method, and then build a private AI system that runs on the business's own data, with numbers computed in code and every figure traced back to its source. Through our Managed Intelligence Platform, we stay accountable after launch with monthly monitoring, evaluation, governance, and improvement, so the system keeps earning the trust of the people who rely on it. It is the same approach the City of Milwaukee, Johnson Controls, and EY described on stage, sized for small and mid-sized businesses.
The morning ended with a hands-on exercise where small groups drafted one-year recommendations for the City's 911 quality assurance challenge, from success measures to new ideas. It was a good reminder of how much Milwaukee's public and private sectors can learn from each other.
Thank you to the MKE Tech Hub Coalition for hosting, to Brooksource for the space, and to the City of Milwaukee, Johnson Controls, and EY for sharing their work so openly. I am already looking forward to the next AI Quarterly.
Want AI built for your actual job? Book a discovery call.