Chicago's Next Mayor? George Cardenas on Algorithms, Assessments, and Who Owns AI in City Hall

Chicago's own AI roadmap has recommended naming an AI lead since 2024. Nobody has been named, and an ordinance requiring the city to report publicly on its AI programs has sat in committee since April of that year. George Cardenas — Cook County Board of Review chairman, Navy veteran, and career auditor — becomes the first candidate in the 2027 race to answer that with an org chart and a hundred-day deadline.

When we recorded this episode, the phrase "artificial intelligence" did not appear anywhere on George Cardenas's campaign site — not on the platform page, not on any of his nine policy pillars. On August 29, 2026, a policy paper titled "Chicago Technology Reengineering" was posted to the campaign's policy hub, covering municipal technology modernization, legacy-system accountability, public-safety technology, artificial intelligence, procurement, and taxpayer protection. It is the first document on the site to carry the phrase. The pillar pages have not been updated to reference it.

That sequence is worth stating plainly, because the substance of this conversation is what a mayor does about AI when the platform hasn't said. George is the second guest in the AI in Chicago Mayoral Candidate Series, in which every declared candidate answers the same six core questions, sent in advance. He is also the only candidate in the field on record committing to put artificial intelligence inside a government decision that determines what people owe — his own office's property tax appeals.

The Accountant's Path: Discipline, Auditing, and 1.8 Million Parcels

George came to Chicago from Durango, Mexico as a boy, graduated from Lane Tech, and enlisted in the U.S. Navy to pay for his education — a trade he still describes as the best decision he ever made, because it also covered the master's degree he did not know was included.

"Without discipline, I was not going to get to the things that I wanted to accomplish," he said. The military, he added, was full of temptations for a young man: "But you need to be disciplined and you need to have goals and apply that to where you want to go."

His first job out of college was auditing subsidiaries of Centel Corporation across nine states. "We look at the revenue cycle, we look at the expense cycle as well, and find holes, and say — why is your inventory at this level, and why are you spending so much money on this particular commodity?" He spent more than a decade in accounting, auditing and consulting roles at Fortune 500 companies including Tenneco and Ameritech, then nearly twenty years as alderman of the 12th Ward.

Since 2022 he has served on the Cook County Board of Review, the three-member body that adjudicates property tax appeals across more than 1.8 million parcels. The board received a record 273,907 appeals in tax year 2024.

That résumé matters here for a specific reason: it is the closest thing in this race to an executive who has actually operated a decision-making system at scale and been accountable for its errors.

The Roadmap Nobody Executed

The City of Chicago's technology department published a Roadmap for AI. It makes three recommendations: establish an AI lead to develop policy and governance, immediately issue a short usage policy every employee can follow, and provide use-case examples so departments share a vocabulary.

There is no public evidence that any of the three has happened. No AI lead has been named. No citywide employee policy has been published. An ordinance from Ald. Gilbert Villegas that would require semi-annual public reporting on city AI programs was introduced on April 17, 2024 and has had no recorded action since.

Peer cities have moved. San José built a ten-week AI upskilling program with San José State University, launched it in September 2024, and had trained more than a thousand employees — roughly fifteen percent of its workforce — by June 2026. Boston issued generative AI guidelines for all city agencies in May 2023 and replaced them with a fuller policy in December 2025.

The gap is not a shortage of intent. Roadmaps recommend; assignment is a separate act, performed by whoever controls reporting lines. In Chicago's case, that is the mayor.

A Chief of AI in the First Hundred Days

Asked whether he would name an AI lead in his first hundred days and where that person would sit, George answered with an org chart rather than a principle.

"Within my first hundred days, I would designate a chief of AI and a digital transformation officer, that would be working with the current CIO and corporation counsel and the inspector general and procurement and operating departments — and ultimately to be accountable to the mayor. To me."

Then he separated the functions. "The CIO handles technology infrastructure, corporation counsel handles the legality, the DI officer the governance and implementation of AI, and the mayor handles the accountability part."

Three of those four seats already exist in Chicago. There is a CIO. The Department of Law runs an Office of Privacy and AI Compliance. The inspector general has audited a city algorithm before — its 2020 advisory on the Strategic Subject List remains the most rigorous public document any Chicago institution has produced on the subject.

The fourth seat is the one that has been empty, and it is the one that converts capability into accountability. Infrastructure, legal review and audit are functions. None of their occupants can be held to a deployment decision they did not make. What George adds is an officer who answers to the chief executive for implementation.

Any executive can run the same test internally. Count the AI-adjacent functions that already exist — security review, privacy, procurement, internal audit, a center of excellence — then ask which one will be asked, by name, why a system shipped.

Compliance Is the Floor, Trust Is the Standard

George named the governing statute unprompted, and named it correctly. Illinois law makes it a civil rights violation to use AI that discriminates against protected classes in employment decisions, bars zip codes as a proxy for a protected class, and requires notice to workers in covered uses. It took effect January 1, 2026. The city is a covered employer regardless of headcount, because the statute's definition of employer includes any municipal corporation "without regard to the number of employees."

"Compliance is going to be the floor," he said, "and trust is going to be the standard."

The floor is incomplete. The Illinois Department of Human Rights published implementing rules on May 15, 2026 and postponed them on June 2, saying the delay was necessary to allow continued collaboration with other state agencies. No revised timeline has been issued. The statutory duties are live; the mechanics of notice — when, in what form, through what channel — are not settled. An employer of roughly 35,760 full-time equivalents owes notice now, against rules that do not yet exist.

That sequence recurs across every jurisdiction now legislating AI, and it inverts how most organizations plan. Waiting for regulatory clarity before naming an owner leaves the interval governed by improvisation — and whoever inherits the function will be judged on what got improvised.

The Algorithm He Turned Down — and What the Record Says

An owner who has never declined anything is processing requests. George is the only candidate in this race who describes declining a specific system.

Asked how he thinks about AI inside property assessment, he described a tool a colleague had brought into the Board of Review — "a STAR program," an algorithm he said predicted assessment increases across parcels and returned false positives.

"Is that something that should happen in the Board of Review? And the answer for us was no," he said. "Because you need a human hand to understand a little more the intricacies of valuation. And it is simply fallacious to just go and apply a tool that automatically assigns a value to all of the PINs at the same time without looking at the detail and the particulars of that property."

He put it on fairness. "You don't want to apply AI to every particular decision that you're making, because it's not going to be a hundred percent accurate. Fairness is an issue, and I don't think AI would be a good measure of fairness when it comes to that type of review."

The public record complicates part of this, and readers should have both halves. Commissioner Samantha Steele, who lost her primary in March 2026, developed a system called STARA in District 2, so the identification is almost certainly right. But every published description of STARA characterizes it as a comparables-retrieval and standardization tool rather than a predictive one, none calls it artificial intelligence, and rather than being rejected it won a National Association of Counties award in 2025 and is described in the present tense on the board's website today. The reading that reconciles both accounts — that STARA was a district-level initiative the other commissioners declined to adopt in their own districts — is documented nowhere. No Chicago news outlet has ever covered the system, so the only public account of what it does comes from the office that built it.

What is not in dispute is the standard he applied. On what he would build instead, he was specific: a tool that surfaces comparable properties and flags discrepancies for a human analyst, not one that sets values. "AI is not gonna determine your taxes … You ultimately are going to have to make the decision, whether that's correct or not."

The Megawatt Question and the Regulator Nobody Campaigns About

Chicago has 39 data centers inside city limits. ComEd's price to compare rose 51.8 percent between August 2024 and August 2026. PJM's independent market monitor found that including 11,993 megawatts of data center load in the 2026/2027 capacity auction raised capacity revenues by $7.27 billion, and concluded that conditions in that market are "almost entirely the result of large load additions from data centers."

Asked how many megawatts of new load he would approve, George declined to name a number — and declined the opposite position too. On a hard zero of unconditioned new load: "I would not meet you there." On naming a figure: "I'm not going to pick an arbitrary number of megawatts without really having a conversation with ComEd, with PJM, with the stakeholders, and also without looking at ratepayer impact data."

His one firm commitment: "Chicago should not subsidize private hyperscale computing centers as data centers."

He then pointed past City Hall to where allocation is actually decided. "I talked about ComEd and PJM, but also the ICC. Where decisions are made."

We framed that on air as a lever no Chicago mayor had used. That was wrong, and the correction is more useful than the claim. The City of Chicago is already a party in every relevant ComEd docket, appearing by right under state law and filing through corporation counsel since August 2025. Docket 26-0364, which examines the proper allocation of large-load infrastructure costs, holds an evidentiary hearing on October 15, 2026 and carries a Commission deadline of February 26, 2027.

Chicago votes on February 23. The proceeding that determines who pays to connect data centers in northern Illinois closes three days later, and more than two months before the next mayor is sworn in. The seat is already the city's. What is open is the instruction given to the lawyer sitting in it.

Paralysis Is Not a Decision: Procurement as Governance

On gunshot detection — nine bidders, no award, $5 million sitting in the 2026 budget — George conceded the premise immediately. "You could tell the procurement system needs an overhaul. So the criticism is legitimate."

Then a standard: response times, evidence recovery, victim location, clearance rates, actionable intelligence, weighed against total lifecycle cost. "If the evidence says yes, then you procure it and you deploy it. And if the evidence is no, then you don't. But paralysis is not a decision."

The city's chief procurement officer told City Council in June 2026 that a final contract might not arrive until February 2027 — the month Chicago votes.

Why This Episode Matters for the AI Ecosystem

1. Ownership, not strategy, is the binding constraint on AI in institutions. Chicago wrote the roadmap and never assigned it. Most organizations with published AI principles are in a version of the same position: the document exists, the behavior does not, and no one can say whose responsibility that is. The diagnostic is four questions — who does the AI owner report to and what can they stop; what are they required to publish and how often; what have they declined and where is the record; and what does a person on the receiving end of an automated decision get to see. An empty answer to the third question is itself a finding.

2. Statutory duties arrive before the rulebook does. Illinois HB 3773 has been enforceable since January 1, 2026, and its implementing rules were withdrawn in June with no replacement date. Any employer waiting for regulatory clarity before naming an internal owner is leaving the enforceable interval to improvisation. Compliance as the floor only works if somebody is standing on it.

3. The decisions that set your AI cost structure are made in rooms you may already have standing in. The proceeding that will determine how large-load interconnection costs are allocated across northern Illinois closes three days after the municipal election, and the City of Chicago has been a party to it for a year. The same pattern holds in the private sector — standards bodies, regulatory dockets, vendor negotiations. Finding those venues and deciding what to argue is work that belongs to an owner, and it is the first thing a new one should be asked for.

An AI owner who reports to the chief executive is that distinction drawn as a reporting line.

Listen to the full episode on Spotify.

For more insights from the operators, builders, and thinkers scaling applied AI from the heart of the Midwest, subscribe and follow at chicagopodcast.ai.