Direction of AI Matters
Will artificial intelligence widen the gap or help close it
By Rajib Ghosh, Founder and CEO, Health Roads
Artificial intelligence may be the most consequential technology of our time. But consequential does not automatically mean equitable. The question is not only how powerful AI becomes. It is who it is built for, which problems it targets, and who shares in the value it creates.
Bill Gates and the 2026 Goalkeepers Report put that choice plainly: AI could become a great equalizer, or it could widen the gap. The report argues that this is not a distant scenario. Decisions being made now about data, access, language, funding, and deployment will shape who benefits for years to come.
That framing deserves particular attention in healthcare and human services. These sectors carry enormous human consequences, yet fragmented data, administrative complexity, thin operating margins, and justified concerns about privacy, bias, accountability, and trust often slow progress. Skepticism around AI is not irrational. In systems serving vulnerable people, a confident error can cause real harm. A tool that cannot explain its role, protect sensitive information, or preserve human accountability should not be waved through simply because it is new.
The question behind the technology
Will AI truly lift all boats, including the organizations that previous waves of technology left behind? Or will it repeat a familiar pattern in which the strongest institutions gain more speed, the wealthiest customers receive the best tools, and everyone else is offered a weaker version later?
No law of technology guarantees shared progress. Markets reward demand backed by purchasing power. They are far less reliable at prioritizing people with the greatest need but the least economic leverage. The Goalkeepers Report identifies this as a basic market failure: those who could benefit most often have the least power to determine where innovation and investment go.
This is where the darker possibility enters. AI can concentrate knowledge, capital, and decision-making power at unprecedented speed. Human societies have repeatedly allowed transformative tools to become instruments of exclusion when the dominant objective was to capture the market, defend the moat, or pursue a winner-takes-all outcome. AI will not be immune to greed merely because its capabilities are extraordinary.
But that future is not inevitable either. Directionality matters. Intent matters. Outcomes matter.
Healthcare and human services are the real test
Healthcare is often described as a technology laggard. That description is incomplete. The sector is not short of technology. It is burdened by systems that don't communicate cleanly, workflows that ask people to bridge gaps manually, and business processes whose cost can overwhelm organizations with the fewest resources.
For community-based organizations, safety-net providers, and agencies serving Medicaid populations, administrative friction is not an abstract productivity problem. If an organization cannot confirm eligibility, document services accurately, exchange information, submit a clean claim, or resolve a denial efficiently, it may delay or lose payment. When margins are already thin, the difference between getting paid and not getting paid can be the difference between sustaining a service and withdrawing it.
Interoperability creates a similar test. Standards such as HL7 FHIR and established health information exchange transactions can make data more portable, but standards do not implement themselves. Connecting systems still requires mapping, validation, governance, security, and operational follow-through. If the work remains too expensive or too specialized, smaller organizations may be technically eligible to participate yet practically excluded.
This is where AI can be useful without pretending to be magical. It can reduce repetitive work, help staff find relevant information, identify exceptions for review, assist with coding and mapping tasks, support outreach, and make complex workflows easier to navigate. The goal should be to extend scarce human capacity, not erase the judgment of clinicians, case managers, billing specialists, program leaders, or the people receiving services.
What responsible direction looks like at Health Roads
At Health Roads, we use AI because it helps organizations accomplish more, not because the label itself creates value. We direct it toward places where lower cost matters, where administrative efficiency determines whether care delivery remains financially viable, and where the cost of interoperability can otherwise become a barrier that smaller organizations cannot cross.
That means focusing on practical outcomes: helping organizations spend less time on avoidable administrative work, improving the path from service delivery to reimbursement, making outreach more timely, and reducing the burden of connecting systems and using the data they exchange.
It also means acknowledging boundaries. In healthcare and human services, responsible AI must be designed around the work and the population it serves. People must remain accountable for consequential decisions. Sensitive data must be governed appropriately. Outputs must be tested in the actual operating context, monitored for error and bias, and constrained to the purpose for which the system was designed. A polished demonstration is not evidence of dependable performance.
The Goalkeepers Report makes the same practical point on a global scale. AI tools must work in the languages people actually speak, reflect local data and realities, and be affordable enough for communities to shape and use them. In the U.S. safety net, the details differ, but the principle is the same: technology that ignores local programs, payer rules, available services, population needs, and frontline workflows will not produce equitable outcomes.
The questions leaders should ask
The debate about AI is too often reduced to optimism versus fear. Leaders responsible for health and human services need a more useful set of questions:
Which human problem are we solving, and for whom?
Does the system lower the cost of serving people with fewer resources, or does it mainly create a premium experience for those already advantaged?
·Who benefits financially from the efficiency, and does any of that benefit reach providers, communities, or patients?
What data and local expertise shaped the tool, and whose reality may be missing?
Where does human review remain essential, and who is accountable when the system is wrong?
Can the organization measure a real outcome such as faster access, lower administrative burden, cleaner reimbursement, or more successful connection to services?
Can the customer sustain the technology after the pilot, or is affordability itself a barrier to adoption?
These questions do not slow innovation. They distinguish meaningful innovation from theater.
Measure what matters
Technology markets naturally celebrate valuations. But valuations can rise and fall like Fourth of July fireworks: bright, loud, and temporary. They tell us what investors expect. They do not tell us whether a community organization stayed open, whether a provider was paid, whether a person received help after discharge, or whether two systems finally exchanged information at a cost the participants could afford.
For AI in healthcare and human services, these measures matter. Did it lower the cost of necessary work? Did it help people act sooner? Did it improve access without compromising privacy or accountability? Did it enable an under-resourced organization to do something previously out of reach?
AI may become a transformative force that lifts more boats. But it will not do so automatically. The outcome will reflect human choices about incentives, access, safeguards, and purpose. We should be ambitious about the technology and disciplined about its direction.
The future of AI should not be judged by how much power it concentrates at the top. It should be judged by how much capability it places in the hands of people and organizations doing essential work on the front line.
Comments