Imagine a disabled veteran trying to submit an income verification through a state portal, only for the upload to freeze because their prepaid phone data ran out. On the other end, a caseworker watches the clock on a dashboard, knowing that staying on the phone to help could hurt their numbers. In the past, that worker might have had more room to use common sense and keep benefits active while the problem was fixed. Today, an automated eligibility system may only register the result: a missing document, an incomplete verification, or a missed deadline. One technical failure can become a compliance problem and push the case toward closure. Timothy Scott That is the hidden reality inside the Connecticut Department of Social Services (DSS). Beneath the language of access, efficiency, and customer service, public benefits are moving into portals, automated phone systems, eligibility platforms, data checks, dashboards, and AI-ready services. This is not simply modernization. It is a safety net being rebuilt as automated gatekeeping. Tools like ConneCT, MyDSS, document uploads, renewal portals, benefit-status checks, and self-service options are promoted as convenience. But they also shift work once handled by agency staff onto vulnerable families. Clients are expected to upload, check, renew, report, track, decode, and correct their own cases. In practice, the client becomes the agency’s unpaid data-entry clerk. For families close to the edge, digital access is not real access. It assumes a working phone, internet, data, language access, digital literacy, and time. A missed renewal, frozen upload, login problem, or confusing notice can become a benefits crisis. What looks like convenience from the top can feel like abandonment from below. Chatbot-style support and generative AI-ready tools can make digital barriers look more human than they are. Automated guides may seem helpful, but they can keep people stuck in self-service loops where system errors remain the client’s problem, putting more distance between families and human judgment. The same shift is happening to workers. Benefits communication is increasingly organized through centralized phone systems, benefits centers, automated prompts, call routing, dashboards, and digital case-management tools. These systems do not just support the work. They shape it. By the time a caseworker reaches a client, the case has often already been organized by system categories, missing documents, automated notices, deadlines, and alerts. Instead of having enough time to understand what is happening in a family’s life, the worker is pushed through a platform-managed workflow. Supervision is built into the system itself. Call times, pending tasks, processing speed, document backlogs, and case actions become performance measures. Long phone waits and abandoned calls pressure workers to clear the queue and follow the checklist. The system rewards speed over care and compliance over accuracy. Caseworkers are not simply helped by technology. Their work is being reorganized around it. At the center of this system is Connecticut’s integrated eligibility infrastructure, including ImpaCT and related workflows that organize how assistance is reviewed, renewed, interrupted, or terminated. Before a worker speaks with the person behind the file, the platform has already shaped the case through alerts, verification demands, deadlines, and exceptions. The rigidity of these automated eligibility systems is not theoretical. A civil rights lawsuit brought by Disability Rights Connecticut challenges the state’s restrictive HUSKY C Medicaid rules, exposing how the state’s infrastructure operates as a blunt financial filter for disabled adults whose monthly incomes fluctuate. The deeper issue is that when arbitrary policy limits are baked directly into automated thresholds, the system becomes structurally blind to instability, crisis, and human need. Automated data checks and external matches can turn ordinary instability into bureaucratic suspicion. A late document, conflicting wage record, address mismatch, missed notice, or frozen upload can trigger more verification or push a case toward closure. In this system, technical problems become empty refrigerators, unpaid rent, and manufactured crises. The surveillance reaches beyond paperwork. Wage records, employment information, identity checks, EBT activity, and other records subject low-income households to a constant administrative audit. When paired with predictive analytics or fraud-detection tools, poverty itself becomes treated as a risk category. Families are not simply applying for help. They are tracked, verified, flagged, sorted, and disciplined as a condition of survival. Their uploads, logins, transactions, household changes, and benefit-use patterns become data points in a system built to monitor risk before it understands need. Poor families are forced to trade privacy for survival. What is happening in Connecticut is part of a national shift toward self-service portals, automated eligibility tools, chatbot support, fraud analytics, digital case-management systems, and vendor-run infrastructure. Welfare offices are being treated less like public social-service institutions and more like enterprise technology systems. Even Connecticut’s new AI law does not fully answer this problem. Senate Bill 5 may create rules around disclosure, transparency, governance, and automated decision systems. But managing AI is not the same as protecting families from automated gatekeeping. The law does not dismantle digital barriers, guarantee lawsuits over technology failures, or confront dashboard metrics used to discipline caseworkers. It risks treating AI as a compliance problem for agencies and vendors, rather than a rights problem for people forced to live under these systems. When public assistance is rebuilt around automation, surveillance, and labor discipline, the safety net stops functioning as public care. It becomes an impersonal machine of social control. Technology should help people, not replace workers, discipline staff, burden clients, or monitor families into submission. Future technology at Connecticut DSS should require labor protections, transparency, public oversight, limits on automated decision-making, appeal rights, data minimization, privacy safeguards, and union involvement. At minimum, the state should disclose generative AI or predictive analytics use and prohibit platform analytics from replacing human judgment. Timothy Scott is an Associate Professor of Social Work at Central Connecticut State University, a tech industry researcher, and the author of the book, Schooling for Silicon Valley. ...read more read less