MY VIEW - Billions of dollars down the drain or in selected pockets? That is the question. How can Los Angeles not have a reliable accounting system to track spending on the unhoused?
Mayor Karen Bass has announced that she is stepping down from the Los Angeles Homeless Services Authority Commission, to which she appointed herself in 2023. She is also “unavailable” on September 15, the date Congress asked her to explain what happened to the more than $1 billion in federal funding LAHSA has received since 2021. Karen is hoping to dodge the consequences of years of mismanagement. Congress should look at Nithya Raman, CD-4, the former chair of the failed Committee on Housing and Homelessness.
Call me cynical, but I suspect Karen will do everything possible to run out the clock until after the election. If Congress compels her to appear before then, what can she say? “Who knew?” “Ask LAHSA.” “Talk to the nonprofit contractor.” Or, if the questions become legally risky, invoke the Fifth Amendment. None of those answers would look good before an election, especially from a mayor who appointed herself to the LAHSA Governing Commission and then missed 13 of 21 meetings during the most recent year examined by LAist and 25 of the 47 meetings since joining the Commission.
Whether through deliberate design, bureaucratic neglect, or sheer incompetence, Los Angeles has created a system that rewards spending while demanding little evidence of results. It is time to use AI to continuously audit spending, match invoices to actual services, and identify which operators produce results and which NGOs merely consume public funds.
Machine intelligence can help expose the grift, suspicious billing, conflicts of interest, and, where the evidence establishes it, outright theft. Los Angeles has spent years claiming that its homelessness machinery is too complex to follow. AI can remove that excuse.
AI cannot cure schizophrenia, reverse a fentanyl overdose, or persuade a methamphetamine addict to enter treatment. It cannot replace psychiatric hospitals, detoxification centers, judges, doctors, police officers, or competent leadership. But it could do what the Los Angeles government has repeatedly failed to do: connect every public dollar to a contractor, a service, an individual, and a measurable result.
Mayor Bass, the City Council, and the Board of Supervisors have little incentive to support ongoing AI auditing of homeless spending. It could expose the entire power-and-patronage structure. Where would they be then? Club Fed?
By calling all these people “homeless,” Los Angeles focuses on where they sleep rather than why they are there. It lets officials claim that almost every failure can be fixed by spending more money on housing. There isn’t enough money in California to build all the $600,000 apartments, assuming the street addicts wanted these apartments for anything other than a stash house. AI could force the system to distinguish among the people it is supposed to serve and match the intervention to the actual condition.
The most immediate opportunity, however, is financial. Los Angeles needs an independent AI spending auditor that can examine every homelessness-related contract, appropriation, invoice, payment, subcontract, payroll record, bed-night, and claimed client outcome.
That system could read millions of pages, connect nonprofit boards, contractor addresses, and subcontracting relationships, and identify duplicate invoices, unsupported payments, unusual price increases, and organizations billing for the same person or service.
This is no hypothetical concern. The federal-court-ordered Alvarez & Marsal assessment of Los Angeles homelessness programs examined approximately $2.4 billion in funding and found a fragmented system in which spending, services, and outcomes could not be reliably reconciled.
The California State Auditor reached a similarly devastating conclusion about the state’s homelessness programs. Despite spending approximately $24 billion over five years, the state still lacked adequate information on the ongoing costs and results of many of those programs.
When the government cannot link spending to outcomes, everyone can claim success, except the taxpayers. Politicians cite appropriations, LAHSA cites contracts, contractors cite funding, and advocates cite motel rooms. Yet the tents remain, sidewalks remain dangerous, and the same impaired people cycle through emergency rooms, shelters, jails, and encampments.
Los Angeles measures money spent because it is easier to count gross dollars than to measure success. An AI accountability system would replace that shell game with a single chain of evidence: money appropriated, money paid, service delivered, person served, and result achieved.
AI could also reveal the true cost of the homeless bureaucracy. Taxpayers should be able to see how much of every dollar reaches the person supposedly being helped and how much is absorbed by administration, consultants, executive compensation, outreach, security, legal expenses, and layers of subcontractors. We should not have to guess. A nonprofit should not be able to hide a large payment to a for-profit affiliate behind multiple layers of contracting. AI is particularly effective at uncovering precisely those relationships.
Compare contractors using standardized metrics: cost per person, permanent placements, treatment completion, returns to the street, and the number still housed after six, twelve, and twenty-four months.
A person who moved from a sidewalk to a motel room for three weeks has not been permanently housed. A tent cleared from one block and reconstructed two blocks away has not been eliminated. A person “contacted” by an outreach worker has not been treated.
The second application would be a unified, by-name system for the most impaired street dwellers. With proper safeguards, it could connect information scattered among LAHSA, mental-health officials, hospitals, paramedics, drug-treatment programs, shelters, outreach teams, and the criminal-justice system.
Consider a mentally ill addict encountered by five agencies. One knows about repeated overdoses; another knows about psychiatric holds, arrests, or abandoned housing. Because the records remain separated, the next outreach worker may know almost none of this history and offer yet another voluntary referral.
AI could recognize the full pattern. If a person has had 14 emergency-room visits, six overdoses, eight psychiatric interventions, three arrests for assault, and four failed placements in the past year, the system should stop treating him as a routine housing case. It should immediately alert a multidisciplinary team and identify the appropriate clinical and legal options.
Los Angeles must confront the distinction between compassion and abandonment. Allowing a psychotic person to live in filth, use lethal drugs, and terrorize a neighborhood is not compassionate. Nor is it compassionate to wait for him to request treatment when his illness prevents him from recognizing that he needs it.
AI cannot order someone into treatment, nor should it. Any involuntary intervention requires human judgment, medical evidence, and legal due process. But AI can compile the facts that doctors, judges, and treatment teams need to make an informed decision. It can also show how many times the system failed before the tragedy occurred.
A third application would manage beds and facilities in real time. Los Angeles should know every morning what treatment, psychiatric, shelter, and supportive-housing capacity is available—not just an empty bed, but the right bed for the person’s condition.
A psychotic methamphetamine user does not simply need a motel room. A frail elderly person does not belong in a chaotic shelter, nor should a mother with children compete with a violent street addict for the same placement. AI could match needs with capacity and expose costly delays.
The system could determine whether encampments are being resolved. Combining sanitation records, outreach reports, 311 complaints, and verified field observations would distinguish genuine reductions from moving people to another sidewalk or jurisdiction.
LAHSA and the contractors being scrutinized cannot control or audit this system. That would be akin to allowing an MS-13 gang member to be his own judge and jury. Oversight should begin with incorruptible programming and a public dashboard tracking the money, contracts, agencies, and results. An independent agency or federal court monitor should control it.
There must also be consequences. Contracts should include financial holdbacks tied to verified outcomes. Providers with persistently poor results should lose funding. Unsupported invoices should not be paid. Refer suspected fraud immediately for investigation.
One enormous limitation remains: AI can reveal failure, but it cannot force officials to respond. It cannot build treatment facilities that politicians refuse to approve. It cannot impose consequences on dishonest contractors if prosecutors decline to act. It cannot compel treatment when elected officials insist that an obviously incapacitated person has an unlimited right to destroy himself in public.
AI will also fail if it is fed dishonest information. But even there it can help. It can detect inconsistent dates, impossible service volumes, repeated language, missing documentation, and outcome claims that conflict with other records. Garbage data would no longer disappear quietly into a government filing cabinet.
We don’t need another homelessness task force, another public-relations campaign, or another expensive consultant producing recommendations nobody is required to follow. It needs an independent system that continuously asks four simple questions: Who was paid? What was delivered? Who received it? Did it work?
The technology is available. ChatGPT assures me that, with access to the underlying records, it could build the system to track payments and outcomes. What Los Angeles lacks is leadership willing to spend our money wisely.
Certainly, Los Angeles does not want to remain the homeless capital of America, or does it?
(Eliot Cohen is a longtime civic advocate who has served on the Neighborhood Council, the Van Nuys Airport Citizens Advisory Council, and the Board of Homeowners of Encino, where he was president of HOME for over seven years. A retired Wall Street executive with a 35-year career, Eliot brings a sharp eye to local governance. He critiques the bureaucratic missteps of City, County, and State officials. Eliot and his wife split their time between Los Angeles and Baja Norte, Mexico.)
