By Futurist Thomas Frey
America spends approximately $50-70 billion annually on homelessness through federal, state, and local programs, nonprofits, emergency services, law enforcement, and healthcare. Major cities spend $40,000-80,000 per homeless person annually. Yet homelessness has increased in most major urban areas over the past decade. Tent cities expand. Encampments grow. The crisis seems perpetual despite massive spending.
AI analysis of homelessness spending, service delivery, outcomes, and alternative approaches is revealing something deeply troubling: a system that has evolved to manage homelessness rather than solve it. Where service providers are financially incentivized to maintain client populations rather than reduce them. Where coordination failures create massive duplication and waste. Where the most effective interventions are systematically underfunded while ineffective traditional approaches consume the majority of resources.
The awakening in homelessness services isn’t about whether we should help people experiencing homelessness—we obviously should. It’s about revealing that the systems we’ve built to address homelessness often serve the institutions providing services more than the people they’re supposed to help. And AI can now track individuals through multiple service systems, analyze spending versus outcomes, and compare approaches to reveal what actually works versus what perpetuates the problem it claims to solve.
The Shelter Industrial Complex
Emergency shelters are supposed to be temporary—a safe place while people transition to housing. AI analysis reveals that shelter systems have become semi-permanent warehousing with minimal pathway to housing and enormous per-person costs.
Here’s what AI discovered: Major cities operate shelters costing $50-150 per person per night—$18,000-55,000 annually. Yet individuals often stay in shelters for months or years rather than weeks. AI analysis tracking individuals through shelter systems shows average stays of 6-18 months, with substantial populations staying multiple years.
The economics are perverse. For $50,000 annually, a city could provide an apartment, supportive services, and intensive case management—the “Housing First” model proven to be more effective. Instead, that same $50,000 goes to a shelter bed with minimal services, no privacy, restrictive rules that many people avoid, and no pathway to permanent housing.
Even worse: AI has revealed that shelter operators are paid per bed per night, creating financial incentive to maintain occupancy rather than help people exit to housing. Empty beds mean lost revenue. Success—people moving to housing—threatens the business model. Shelter operators optimize for bed utilization, not housing outcomes.
AI analysis of shelter contracts shows that very few include performance metrics tied to housing placement. Contracts specify bed capacity and occupancy requirements, but rarely require measuring or improving housing outcomes. Operators are paid for providing beds, not for solving homelessness.
One analysis found that a typical 200-bed shelter costs $8-12 million annually to operate but houses fewer than 50 people annually (the rest are repeat users or long-term residents). The per-exit cost: $160,000-240,000 per person successfully housed. Meanwhile, Housing First programs house people at costs of $15,000-25,000 per person including rental assistance and services.
One estimate: if cities redirected shelter spending to Housing First approaches, they could house 2-3 times as many people at the same cost while providing superior outcomes. The shelter system persists not because it’s effective but because it’s established.
The Service Provider Fragmentation
Dozens of organizations in each city provide homelessness services—shelters, meals, healthcare, case management, job training, mental health, substance abuse treatment. AI analysis reveals that fragmentation creates massive duplication, coordination failures, and gaps where people fall through cracks.
Here’s the dysfunction: A homeless individual might interact with 10-20 different service providers. Each has separate intake, separate eligibility requirements, separate documentation, separate databases. None share information effectively. Each operates independently, optimizing for their own metrics rather than overall system effectiveness.
AI analysis tracking individuals across multiple service systems shows:
- People complete intake paperwork 5-10 times with different providers asking identical questions
- Individuals receive duplicate services from multiple providers (3 different case managers, 2 mental health providers) while missing critical services
- Providers don’t know what other providers are doing—no coordinated care plans
- Individuals navigate complex eligibility requirements, appointments, and locations creating barriers that many can’t overcome
- Small organizations with 80-90% overhead (only 10-20% going to direct services) persist because there’s no mechanism to consolidate or eliminate ineffective providers
Even worse: AI has revealed that providers actively protect their turf rather than collaborate. They compete for the same funding sources, resist sharing information (citing privacy concerns that often aren’t legally required), and duplicate rather than coordinate services because collaboration might threaten their organizational existence.
One city analyzed by AI had 47 separate nonprofits providing homelessness services. Collectively they had approximately $200 million in revenue and employed 1,500+ staff. But coordination was minimal. AI analysis showed that consolidating to 8-10 integrated providers could serve the same population with 30-40% fewer administrative staff while improving coordination and outcomes.
The resistance to consolidation is fierce. Each organization has a board, executive director, staff, and donor base invested in maintaining separate existence. No individual organization will voluntarily dissolve even when consolidation would clearly improve services.
One estimate: fragmentation and coordination failures waste approximately $15-25 billion annually nationwide—money that goes to duplicated overhead, redundant intake systems, and lack of coordination rather than to actual services or housing.
The Criminalization Cost Spiral
Cities respond to visible homelessness with enforcement—sweeps of encampments, citations for camping, arrests for trespassing. AI analysis reveals that criminalization creates enormous costs while solving nothing and often making homelessness harder to escape.
Here’s the economics: Arrest and booking costs approximately $500-1,000. A night in jail costs $100-300. Court processing costs $500-1,500. People experiencing homelessness are arrested repeatedly for survival behaviors—sleeping in parks, sitting on sidewalks, public urination (when no bathrooms are accessible).
AI analysis of jail populations shows that approximately 15-30% of jail bookings in major cities involve people experiencing homelessness, often for low-level offenses directly related to homelessness. Many are arrested repeatedly—monthly or more frequently—creating revolving door cycling between streets and jail.
The costs compound: arrests make it harder to get housing (criminal records), harder to get jobs, harder to access services (missing appointments while incarcerated), and traumatize people already experiencing crisis. Criminalization doesn’t reduce homelessness—it perpetuates it while generating enormous costs.
Even worse: AI has revealed that cities often spend more on enforcement than they would spend housing people. One analysis found a city spending $50 million annually on homeless-related enforcement (police, courts, jails) while spending $30 million on homeless services and housing. The enforcement spending made problems worse while housing spending could have solved them if adequately funded.
AI analysis also shows that encampment sweeps simply disperse people—they move to different locations rather than into services or housing. Sweep operations cost $10,000-50,000+ per sweep (police time, waste disposal, storage of belongings) while producing no positive outcomes. Most people return to the same locations within days or weeks.
One estimate: cities spend approximately $10-20 billion annually criminalizing homelessness through enforcement that doesn’t reduce homelessness but does make it harder for people to escape while consuming resources that could fund actual solutions.
The Emergency Room Revolving Door
People experiencing homelessness use emergency services at extremely high rates. AI analysis reveals that a small number of individuals account for enormous costs—and that housing them would cost far less than continued emergency service use.
Here’s the pattern: AI analysis of hospital records identifies “super-utilizers”—individuals with dozens or hundreds of ER visits annually. A single person might generate $100,000-500,000 in annual ER costs, often for conditions that aren’t medical emergencies but for which ER is the only accessible care.
These super-utilizers are disproportionately people experiencing homelessness with untreated mental health or substance use issues. They use ERs for primary care, mental health crises, substance withdrawal, exposure-related issues, and sometimes just for a safe warm place to sleep.
AI analysis shows that housing super-utilizers with supportive services reduces their ER use by 60-80%, saving far more than the housing costs. One study found that providing supportive housing to 100 super-utilizers cost $2 million annually but saved $5 million in reduced ER costs—a net savings of $3 million while dramatically improving outcomes.
Even worse: AI has revealed that hospitals and healthcare systems rarely invest in housing solutions despite clear evidence they would save money. Why? The savings accrue to the system (lower overall costs) but individual hospitals lose revenue from reduced ER visits. The misaligned incentives prevent investment in solutions that would benefit everyone.
AI has also identified that many people cycle between ERs, jails, and streets repeatedly—the “tri-morbidity” of homelessness, incarceration, and emergency healthcare. One individual might generate $200,000+ annually across these systems, yet none of them invest in housing that would break the cycle because each operates independently optimizing for their own metrics.
One estimate: providing supportive housing to the approximately 100,000-150,000 highest-cost homeless individuals nationally would cost approximately $2-3 billion annually but would save approximately $5-8 billion in reduced emergency services, law enforcement, and healthcare costs—a net savings of $3-5 billion while ending homelessness for those most in crisis.
The Housing First Evidence Versus Traditional Approaches
Decades of research show “Housing First”—providing housing without preconditions, then wrapping services around people—is more effective and often cheaper than traditional “treatment first” approaches. Yet AI analysis shows most funding still goes to traditional approaches despite evidence they don’t work.
Here’s the evidence: Housing First programs show housing retention rates of 80-90% after one year and 70-85% after two years. Traditional programs requiring sobriety, treatment compliance, or employment before housing show retention rates of 30-50% because many people can’t maintain these requirements while homeless.
AI analysis of costs shows Housing First programs cost $15,000-35,000 per person annually (including rent, services, case management). Traditional programs cost $20,000-50,000 per person annually (shelters, transitional housing, treatment programs) while housing far fewer people successfully.
Yet funding allocations tell a different story. AI analysis of homelessness spending shows approximately 60-70% goes to traditional shelter and transitional housing approaches, with only 20-30% to Housing First despite superior evidence. Why?
The reasons AI has revealed:
- Shelters and traditional programs are established with decades of operations, boards, facilities, and political connections
- Housing First requires coordination across housing, healthcare, and social services—harder to fund through traditional categorical programs
- Voters and funders often believe homeless people should “earn” housing through compliance and sobriety rather than receiving it unconditionally
- Traditional providers resist Housing First because it threatens their funding and existence
Even worse: AI has shown that communities claim to embrace Housing First while actually implementing “housing readiness” programs that maintain preconditions—essentially rebranding traditional approaches as Housing First to access new funding while maintaining old practices.
One comprehensive analysis: if all homelessness funding were reallocated to Housing First approaches proven effective, the U.S. could house all homeless individuals (approximately 650,000 on any given night) within 2-3 years at current spending levels. The barrier isn’t funding—it’s allocation of existing funding to approaches proven not to work.
The Chronic Versus Transitional Misidentification
Homelessness is not a monolithic experience. AI analysis reveals two very different populations with different needs—but services often treat everyone the same, wasting resources on those who don’t need them while underserving those who do.
Here’s the distinction: “Transitional homelessness” affects people who lose housing temporarily due to job loss, domestic violence, eviction, or crisis—approximately 70-80% of people experiencing homelessness at some point. Most exit within 3-6 months, often with minimal assistance.
“Chronic homelessness” affects people with serious mental illness, substance use disorders, or disabilities who experience long-term or repeated homelessness—approximately 20-30% of the homeless population. They account for approximately 50-60% of shelter nights and 70-80% of costs.
AI analysis shows that services often provide intensive, expensive interventions to transitionally homeless people who would likely exit on their own with minimal help, while providing inadequate services to chronically homeless people who need intensive support but instead cycle through shelters, streets, jails, and hospitals.
The resource misallocation is enormous. AI analysis found:
- Transitionally homeless families receive months of shelter and services costing $30,000-60,000 when rapid rehousing assistance of $3,000-8,000 would have been sufficient
- Chronically homeless individuals receive shelter beds costing $40,000-60,000 annually while remaining homeless, when supportive housing costing similar amounts would end their homelessness
- Assessment tools that should differentiate populations often don’t get used, or get overridden by first-come-first-served access
Even worse: AI has revealed that service systems optimize for serving easier-to-serve transitionally homeless people because they produce better success metrics. Organizations report high exit-to-housing rates by focusing on people who were going to exit anyway, while chronically homeless people who most need services are underserved because they’re harder cases with lower success rates.
One estimate: proper targeting of resources—minimal assistance to transitional homelessness, intensive supportive housing for chronic homelessness—would house 40-60% more people at the same total cost by matching resource intensity to actual need rather than serving everyone uniformly.
The NIMBY Housing Production Blockade
Homelessness persists partly because housing production can’t keep pace with need. AI analysis reveals that NIMBY (Not In My Back Yard) opposition systematically blocks affordable housing development while cities spend billions managing homelessness that could be prevented through housing.
Here’s the dysfunction: Cities zone 70-90% of residential land for single-family homes only, prohibiting apartments, duplexes, and affordable housing. Zoning approval processes take years and cost $500,000-2,000,000 before construction begins. Height restrictions, parking requirements, and design reviews make affordable housing financially infeasible.
When affordable housing or supportive housing projects are proposed, neighborhood opposition emerges: residents fear property value impacts, increased crime, changed neighborhood character. Projects get delayed, modified to unworkability, or rejected entirely.
AI analysis shows systematic patterns:
- Affordable housing projects face neighborhood opposition at rates 3-5 times higher than market-rate housing
- Projects in higher-income neighborhoods face more opposition than identical projects in lower-income neighborhoods
- Opposition uses coded language about “neighborhood character” and “traffic concerns” while underlying motivation is often socioeconomic exclusion
- Approval processes that should take 6-12 months take 3-5 years due to discretionary reviews allowing opponents to delay indefinitely
The cost is enormous. Delayed approvals add $50,000-200,000 per unit in carrying costs, legal fees, and redesigns. Projects that get built cost far more than necessary due to restrictions. Most projects never get built at all.
Even worse: AI has revealed that the same communities blocking affordable housing then complain about homelessness and demand enforcement against visible homelessness. They prevent the solution (housing) while demanding management of the problem their policies create.
One analysis of a major California city: zoning restrictions, approval delays, and NIMBY opposition prevent approximately 4,000-6,000 affordable units annually that would otherwise be financially feasible. The city spends approximately $400 million annually managing homelessness while blocking the housing development that would cost approximately $300 million annually (spread over 30-year mortgages) to permanently house the homeless population.
One estimate: if cities eliminated exclusionary zoning and streamlined affordable housing approvals, housing production would increase 50-100%, eliminating most homelessness within 5-7 years while costing less than current spending on managing rather than solving homelessness.
The Mental Health and Substance Abuse Treatment Gap
Serious mental illness and substance use disorders contribute substantially to chronic homelessness. AI analysis reveals massive gaps between treatment need and treatment availability—with waiting lists months long while people cycle through streets, shelters, and jails.
Here’s the gap: Approximately 25-30% of people experiencing homelessness have serious mental illness. Approximately 35-45% have substance use disorders. Many have both (co-occurring disorders). These populations need intensive treatment to maintain housing stability.
But AI analysis of treatment capacity shows:
- Psychiatric hospital beds have declined 90% since the 1960s
- Community mental health services are underfunded and overwhelmed
- Substance abuse treatment programs have waiting lists of 3-6 months
- Programs that accept people with complex needs (co-occurring disorders, criminal history, no income) are extremely limited
- Insurance often doesn’t cover intensive services needed for people experiencing homelessness
The result: people with serious mental illness and substance use disorders cycle through emergency services without receiving adequate treatment. They’re stabilized in ER or jail, released, decompensate, and cycle again—a pattern costing $100,000-300,000 annually per person while never addressing underlying conditions.
Even worse: AI has revealed systematic gaps in the continuum of care. Hospital psychiatric units discharge people directly to streets or shelters because no housing with mental health services is available. Substance abuse treatment programs discharge people to homelessness after 30-90 days because residential treatment ends but housing and ongoing support doesn’t exist.
One particularly problematic pattern: involuntary psychiatric holds (5150 in California) occur when people are dangerous to self or others. AI analysis shows these holds average 3-5 days—enough time to stabilize acute crisis but not enough to address underlying illness. People get released, decompensate within days or weeks, and cycle back. It’s expensive crisis management, not treatment.
One estimate: expanding mental health and substance abuse treatment capacity to match need would cost approximately $10-20 billion annually nationwide but would save approximately $15-30 billion in reduced emergency services, law enforcement, and healthcare while enabling approximately 100,000-200,000 people to maintain housing who currently cannot due to untreated behavioral health conditions.
The Prevention Versus Crisis Management Imbalance
Preventing homelessness costs far less than managing it once it occurs. Yet AI analysis shows approximately 90-95% of spending goes to crisis management (shelters, emergency services) with only 5-10% to prevention programs proven to be more cost-effective.
Here’s the prevention math: Eviction prevention costs $1,000-3,000 per household (rental assistance, legal representation, mediation). Family diversion (helping families avoid shelter by resolving temporary housing crises) costs $500-2,000 per family. Rapid rehousing (quickly exiting shelter to housing) costs $3,000-10,000 per household.
Compare to crisis costs: Emergency shelter costs $40,000-60,000 per person annually. Once in shelter, people often stay months generating costs far exceeding prevention. And some never exit, cycling indefinitely at enormous cost.
AI analysis shows that every $1 spent on eviction prevention saves $7-12 in avoided shelter costs. Every $1 spent on rapid rehousing saves $2-5 in avoided long-term shelter stays. Prevention is demonstrably more cost-effective than crisis management.
Yet funding allocations tell a different story. AI analysis found:
- Most homelessness funding requires people to be literally homeless to qualify—you can’t prevent homelessness with funding that requires homelessness to access
- Prevention programs are underfunded because they serve people before they become homeless (invisible) rather than people already homeless (visible)
- Crisis funding is easier to justify politically (“people are sleeping on streets”) versus prevention funding (“helping people who might become homeless”)
- Service providers prefer crisis funding because it’s more stable—prevention programs might work so well they reduce their client base
One city analyzed by AI could fund eviction prevention for approximately 3,000 households annually at the same cost as sheltering 300 people—a 10:1 ratio. But 85% of funding went to shelters with only 15% to prevention.
One estimate: reallocating just 30% of homelessness funding from crisis management to prevention would reduce new homelessness entries by 40-60% while freeing up crisis resources for those who still become homeless despite prevention efforts.
The Data System Failures
Effective homelessness response requires tracking individuals, coordinating services, and measuring outcomes. AI analysis reveals that data systems are fragmented, incompatible, and often poorly used—preventing the coordination and evaluation necessary for improvement.
Here’s the dysfunction: The federal government requires Homeless Management Information Systems (HMIS) to track people receiving services. But implementation varies wildly. Some providers don’t use HMIS. Those that do often enter minimal data. Different regions use incompatible systems. Privacy concerns (some legitimate, some overstated) prevent data sharing.
AI analysis shows:
- Many cities can’t answer basic questions: How many people are homeless? How long do people stay homeless? What services are effective? Which providers have best outcomes?
- Individuals served by multiple providers appear multiple times in systems—nobody knows if you’re seeing the same person or different people
- Outcome tracking is minimal—providers record inputs (services delivered) but not outputs (housing achieved) or outcomes (housing maintained)
- By-name lists (individual-level tracking) exist in some communities but not most—without individual tracking, you can’t coordinate care or measure success
Even worse: AI has revealed that some providers resist data systems because measurement might reveal ineffectiveness. They prefer reporting activities (“served 500 people”) rather than outcomes (“housed 150 people, 120 maintained housing one year later”) because outcomes data might threaten funding.
One estimate: implementing comprehensive, interoperable data systems with individual-level tracking and outcome measurement would cost approximately $500 million-1 billion nationally (one-time investment plus annual maintenance) but would enable approximately $5-10 billion annually in efficiency improvements through better coordination, outcome-based funding, and elimination of ineffective programs.
The Federal Funding Categorical Restrictions
Federal homelessness funding comes through multiple agencies with different eligibility rules, allowed uses, and reporting requirements. AI analysis reveals that categorical restrictions prevent communities from using funding flexibly to serve people effectively.
Here’s the restriction problem: HUD funding can be used for housing but has limited flexibility for services. SAMHSA funding can be used for substance abuse treatment but not housing. VA funding serves veterans only. TANF serves families with children. Each funding stream has separate applications, reporting, and compliance requirements.
Communities receive funding in categorical silos that don’t match how people actually experience homelessness. Someone with mental illness and substance use disorder needs both housing and treatment—but housing funding can’t pay for treatment and treatment funding can’t pay for housing, so services get fragmented.
AI analysis shows that communities spend enormous staff time braiding funding sources—using HUD funding for rent, SAMHSA for treatment, county mental health for case management, and private donations for incidentals. The administrative burden is enormous and some needs fall in gaps because no funding source covers them.
Even worse: AI has revealed that categorical restrictions often prevent evidence-based practices. Housing First is proven effective but requires flexibility to provide housing without preconditions and wrap services around people—flexibility that categorical funding often doesn’t allow.
One particularly problematic restriction: many federal programs prohibit using funds for people with recent criminal convictions or active substance use—exactly the populations most likely to be chronically homeless and most in need of services.
One estimate: if federal homelessness funding were consolidated into flexible block grants allowing evidence-based approaches tailored to local needs rather than categorical restrictions, communities could serve 20-30% more people at the same cost through reduced administrative burden and better-matched services to needs.
The Visible Versus Total Homelessness Response Gap
Public and political attention focuses on visible street homelessness—encampments, panhandling, sidewalk sleeping. AI analysis reveals that visible homelessness represents only 35-45% of total homelessness, with the rest in shelters, vehicles, doubled-up with others, or otherwise invisible.
Here’s the mismatch: Most funding and attention goes to managing visible homelessness through enforcement, shelter, and outreach. But most people experiencing homelessness are invisible—families in shelters, youth couch-surfing, people sleeping in cars, individuals doubled-up with family or friends.
AI analysis shows systematic neglect of invisible homelessness:
- Families in shelters receive services, but prevention that would keep them housed receives minimal funding
- Youth homelessness is undercounted because youth avoid adult shelters and systems
- Vehicle residents aren’t included in counts unless they park in specific areas during count periods
- Doubled-up individuals aren’t counted as homeless despite lacking stable housing and being at high risk
The resource allocation reflects visibility, not need. AI analysis found one city spending $60 million annually on street outreach and downtown shelter beds serving approximately 1,500 visible homeless individuals, while spending $12 million on family homelessness serving 2,500 people—$40,000 per visible homeless person versus $4,800 per invisible homeless person.
Even worse: AI has revealed that focusing on visible homelessness creates incentives to move people rather than house them. Cities sweep encampments not to house people but to make them less visible. Enforcement pushes people to different neighborhoods or adjacent cities. The problem doesn’t solve—it relocates.
One estimate: if homelessness spending were allocated based on total homelessness rather than visible homelessness, approximately 50-70% more people could be effectively served at the same cost, with particular benefits to families, youth, and others experiencing invisible homelessness who are often easier to help but receive minimal resources.
The Institutional Knowledge Versus Evidence Gap
Homelessness services are provided by professionals with deep experience and strong beliefs about what works. AI analysis reveals that these beliefs often conflict with evidence—and when institutional knowledge conflicts with data, knowledge typically wins, preventing adoption of proven approaches.
Here’s the conflict: Shelter operators believe that structure, rules, and sobriety requirements are necessary for people to stabilize. Housing First research shows unconditional housing works better. Service providers believe people need intensive case management. Evidence shows that some populations need minimal support once housed. Providers believe treatment must precede housing. Data shows housing enables treatment compliance.
AI analysis of provider practices versus evidence-based approaches shows systematic gaps:
- Providers maintain requirements (sobriety, employment, treatment compliance) proven counterproductive
- Programs designed around provider convenience (intake hours, location, rules) rather than accessibility for people experiencing homelessness
- Resistance to harm reduction approaches despite evidence they improve engagement and outcomes
- Continued funding for approaches with decades of evidence they don’t work (transitional housing, treatment-first models)
Even worse: AI has revealed that when evidence contradicts institutional knowledge, many providers dismiss evidence as “not applicable to our population” or “not understanding the complexity of our situation.” The resistance is genuine—providers sincerely believe their experience and observations over statistical evidence.
One particularly clear example: abstinence-based treatment programs that require sobriety before or during treatment show far lower retention and success rates than harm reduction programs that meet people where they are. Yet many providers resist harm reduction based on belief that “enabling” drug use prevents recovery—despite evidence to the contrary.
One estimate: if homelessness programs adopted evidence-based practices and eliminated approaches proven ineffective, effectiveness would increase 30-50% at the same cost. The barrier isn’t funding or even knowledge availability—it’s willingness to change practices based on evidence rather than institutional preference.
What Happens Next
Homelessness is solvable. AI analysis, combined with decades of research, shows clearly what works: Housing First with appropriate services, prevention over crisis management, treatment capacity matching need, and adequate affordable housing supply. Communities that implement these approaches reduce homelessness dramatically.
Yet most communities continue ineffective approaches. They fund shelter systems that warehouse people without housing them. They criminalize survival behaviors rather than providing alternatives. They block affordable housing development while spending billions managing homelessness. They spread resources across fragmented providers rather than consolidating for efficiency.
Why? Because the current system serves providers even when it fails people experiencing homelessness. Organizations have budgets, staff, facilities, and missions tied to current approaches. Change threatens established interests. And because homelessness makes some people uncomfortable, there’s political pressure to manage visibility rather than solve problems.
But pressure is mounting. AI can now track spending versus outcomes with unprecedented clarity. Voters can see that spending increases while homelessness increases. Comparisons between communities reveal that some solve homelessness while others perpetuate it at similar cost. The patterns are becoming undeniable.
Final Thoughts
The awakening in homelessness services isn’t about whether we should help people experiencing homelessness—we obviously should. It’s about revealing that the systems we’ve built often serve providers more than people, that we spend enormous sums perpetuating rather than solving homelessness, and that effective solutions exist but are systematically underfunded in favor of ineffective traditional approaches.
AI makes visible what was always true but impossible to quantify: most homelessness spending doesn’t house people, fragmentation and coordination failures waste billions, and the populations most in need often receive least appropriate services. The crisis is perpetual not because it’s unsolvable but because solving it would disrupt established systems.
We can do better. Evidence shows that Housing First approaches, adequate treatment capacity, prevention over crisis management, and streamlined affordable housing production could end most homelessness within 5-7 years at current or lower spending levels. The choice isn’t between helping people and fiscal responsibility—it’s between perpetuating ineffective systems and adopting approaches proven to work.
The crisis is perpetual by design, not necessity. AI has revealed the design. The question is whether we’ll choose to redesign based on evidence of what works rather than institutional preference for what’s established.
In our next column: Criminal Justice—The Profit in Punishment.
Related Articles:
National Alliance to End Homelessness – Cost of Homelessness
The Lancet Public Health – Effects of Housing First Approaches on Health and Social Outcomes of Homeless People
Urban Institute – Reducing Homelessness by Reducing Evictions
