GovTech is moving beyond portals that simply display information toward intelligent workflows that actively operate, route, and govern public services. This blog explores how GovSoft’s workflow-first approach combines AI with structured governance to transform information delivery into accountable, audit-ready action.
.svg)
The Challanges
The Solution
The Results
Key Takeaways
A decade of public-sector technology investment produced the modern government portal. The resident can find the form. The contractor can submit the application. The constituent can check the status of a license, a benefit, a permit, a tax filing. The portal era did real work. It moved government from paper to digital and made information accessible at a scale that the prior architecture could not support.
The portal is also, by design, a destination. The resident arrives, retrieves what they came for, and leaves. The work the portal references happens elsewhere, in the case management systems, the eligibility platforms, and the workflow queues that the portal does not see. When the resident's case stalls, the portal cannot move it. When the application is incomplete, the portal cannot complete it. When the determination requires a new piece of evidence, the portal cannot request it. The portal informs. The action lives behind the portal, in workflows that the resident cannot see and that the technology was not built to govern.
The vendor language for what comes next is agentic, describing AI systems that act on the work rather than describe it. The operational language is more measured. What is emerging across public-sector technology, and what federal policy is now directing agencies to build toward, is the intelligent workflow: a system that does not just display information but operates the work the information describes, inside a governance structure that names accountability for every action it takes. The shift is not from portals to chatbots. It is from information delivery to governed action.
The portal-era architecture solved discovery. The resident knows what form to file, what office to contact, what status to check. What it did not solve is operation. The workflows that produce the determinations, issue the permits, process the benefits, and close the cases live in systems that the portal connects to as a display layer rather than as a participant. The portal can show the resident that their application is "under review." It cannot tell the resident what specifically is under review, what action would advance it, or when the review will conclude, because those properties are not available through the display interface.
The operational consequence is familiar to anyone who has staffed a state or municipal service line. The portal handled the easy part. The phone calls, the back-office assembly, the cross-system reconciliation, the documentation packaging for audit, the exception handling, all of it happens behind the portal in workflows that were not designed for end-to-end visibility or governed action. The portal made government services easier to find. It did not make them easier to deliver.
The federal posture on this question shifted in 2025 with unusual clarity. On April 3, 2025, the Executive Office of the President's Office of Management and Budget issued Memorandum M-25-21, "Accelerating Federal Use of AI through Innovation, Governance, and Public Trust," directing agencies to provide improved services to the public while maintaining strong safeguards for civil rights, civil liberties, and privacy, and rescinding and replacing the previous OMB Memorandum M-24-10. The White House described the revised memoranda as offering guidance on AI adoption and procurement intended to remove unnecessary bureaucratic restrictions while increasing public trust through the federal use of AI. Ponemon InstituteComply
The direction in M-25-21 is not framed as a question of whether agencies should adopt AI. It is framed as a question of how AI is governed when it is adopted. The memorandum requires Chief AI Officer appointments, AI Governance Boards, enterprise AI strategies, and risk management practices for high-impact AI, with CFO Act agencies required to publish enterprise AI strategies within 180 days of the April 3, 2025 issuance date. The companion procurement memorandum, M-25-22, applies similar governance discipline to how agencies acquire AI. Federal policy is treating governed AI deployment as a baseline expectation, with the operational question being how the governance is structured around the deployments agencies are already moving forward with.
The other direction federal policy is moving is toward measurable improvement in how government services are delivered. The Government Service Delivery Improvement Act, enacted in January 2025, requires OMB to coordinate government efforts to improve federal service delivery from high impact service providers, including establishing government-wide standards, policies, guidelines, and performance measures, and GAO has recommended that OMB set quantitative targets and time frames for the improvements expected from those providers. Cleary enforcement watch based on SEC report
The two policy directions are connected, and the connection is structural. The service delivery requirement asks agencies to demonstrate measurable improvements in how citizens experience government. The AI policy asks agencies to adopt AI in ways that are governed and accountable. The architecture that satisfies both is the same architecture: workflows that operate the work, govern the operation, and produce a record of both that is reviewable against service-delivery targets and accountability standards.
State and federal agencies operate technology stacks that were procured in different cycles, by different programs, for different purposes. Often, no one at the agency still remembers why a given platform was chosen in the first place. The people who made the decision have moved on. The institutional memory of the procurement rationale moved with them. And in the cases where someone does remember why a system was selected, the reasons have usually shifted since: the regulatory environment around the platform has changed, the program's needs have evolved, and the agency rarely has a documented process to step back and decide whether the system in place is still the right one or whether it needs to be replaced or improved. The eligibility system was built to run determinations. The case management platform was built to track cases. The permitting system was built to issue permits. The portal sits above them as a display layer. The workflows that move work between them are often manual, often unwritten, and rarely instrumented for the kind of end-to-end visibility that governed AI deployment requires.
The operational reality is that the architecture the federal direction now expects agencies to build toward does not exist in most agencies as deployed. The systems exist. The portal exists. The governance role exists, increasingly, in the form of newly appointed Chief AI Officers. What is missing is the workflow layer that connects these elements into an operating model the agency can actually govern.
An intelligent workflow, in the operational sense the federal direction implies, is a system that operates the work referenced by the portal, governs the operation through documented workflows the agency has defined, and produces a record of every action it takes that is reviewable through the governance structure the agency has established. The portal continues to do what it does well. The intelligent workflow handles what the portal does not: moving the case forward, requesting the missing evidence, routing the exception, surfacing the decision that requires human judgment, and recording each step in the form an audit or service-delivery review will examine.
The structural commitment is specific. The workflow that operates the work is defined by the agency, owned by named operational leads, and instrumented so that every action it takes produces a reviewable record. The systems the workflow connects to continue to operate as they do. The intelligent workflow sits above them as an orchestration layer that holds the case, the determination, the application, the permit, or the request as a single object, with the full operating record attached.
This is not portal replacement. The portal remains the public face. The intelligent workflow is the back-of-house operating layer that the portal could not be, because the portal was designed for display rather than operation.
AI is only practical when human governance is built into the workflow. The federal direction in M-25-21 reads as exactly this principle, with Chief AI Officers, governance boards, and risk management practices serving as the structural commitments that make AI adoption defensible. The same principle holds at the operational layer below the governance roles.
The placements where AI assists inside an intelligent workflow are specific and bounded. AI-assisted classification of inbound requests, applications, documents, and constituent communications, so the right material reaches the right workflow in a form the workflow can act on. AI-supported routing of exceptions and edge cases, surfacing the requests that depart from the standard path so a human can review them rather than have them process invisibly. AI-assisted drafting of decision rationales, status updates, and constituent communications, so the agency staff member reviews and approves rather than transcribes. AI for monitoring workflow status across the agency, surfacing where work is slowing before it becomes a service-delivery problem or an audit exposure. AI for review prioritization across queues too large for any one staff member to triage manually.
In each placement, AI supports work that does not require human judgment, inside a workflow that the agency has defined, with a human approving the outcomes that matter. The structure the federal direction in M-25-21 describes is the structure these placements operate within: the Chief AI Officer governs the deployment, the governance board reviews the risk classification, the workflow holds the operational definition, and AI assists with the work the workflow does not require human judgment to perform.
For agencies and the GovTech vendors that serve them, the question of what good governance looks like has a stable answer. The National Institute of Standards and Technology released the Artificial Intelligence Risk Management Framework, AI RMF 1.0, in January 2023, intended for voluntary use to improve the ability to incorporate trustworthiness considerations into the design, development, use, and evaluation of AI products, services, and systems, with a companion AI RMF Playbook and Trustworthy and Responsible AI Resource Center launched to facilitate implementation and international alignment. U.S. GAO
The NIST framework is structured around four functions: Govern, Map, Measure, and Manage. The intelligent workflow architecture maps to those functions directly. The Govern function is the policy and operational structure the agency has defined. The Map function is the workflow definition that names where AI is applied. The Measure function is the instrumentation the workflow produces. The Manage function is the operational discipline that responds to what the measurements surface. An intelligent workflow built to this baseline is a workflow that satisfies the M-25-21 governance expectations not because it was built specifically for compliance, but because the architecture and the standards are pointing in the same direction.
The shift from portals to intelligent workflows is observable in the policy direction, in the standards baseline, and in the agency-level implementation patterns that follow from both. Federal agencies are appointing Chief AI Officers and standing up governance boards because federal policy requires them to. The NIST AI RMF is operating as the de facto governance standard because the federal direction and the broader regulated-industry environment have converged on it. The Government Service Delivery Improvement Act is creating quantitative service-delivery targets that the portal-era architecture is not equipped to meet.
The state and local environment is moving in the same direction, often ahead of the federal timeline in specific service areas. State CIOs have placed AI and automation high in their annual priorities for several years. State digital services teams have grown. The technology investments that follow are increasingly oriented around workflow operation rather than information display, because the service-delivery targets the citizens and the legislative oversight bodies expect cannot be met by a display layer alone.
The architectural pattern that emerges is consistent. The portal continues to operate as the public-facing display layer. The intelligent workflow architecture sits behind it, operating the work, governing the operation, and producing the record. AI sits inside the workflow as an assistive layer, supporting the operational scale at which governed work has to run, without taking on the judgment that the agency staff and the governance structure are accountable for. The architecture satisfies the federal policy direction, the standards baseline, and the service-delivery expectations because it is the architecture all three are pointing toward.
Sources:
The architectural pattern, in practice, tends to converge on three properties.
The first is that the workflow definition leads the AI deployment, rather than the inverse. The agency, in coordination with its governance structure, defines the workflow that operates each category of work. The AI is applied where the workflow's design accommodates it, in placements that do not require human judgment, inside boundaries the governance structure has approved. Agencies that proceed in the other order, selecting AI capabilities first and configuring the workflow around them, tend to encounter the deployment risks that the M-25-21 governance requirements were specifically designed to mitigate.
The second is that the portal and the intelligent workflow are architectural complements, not replacements. The portal continues to serve discovery, intake, and display, which it does well. The intelligent workflow operates the work and produces the operating record, which the portal was never built to do. Agencies that approach this shift as a portal modernization project tend to underestimate the scale of the workflow architecture required behind it. Agencies that treat the portal as stable and invest in the workflow layer behind it tend to deliver service-delivery improvements within the current operating period.
The third is that the governance structure is operational, not nominal. A Chief AI Officer who has been appointed but is not connected to the workflows where AI is deployed is satisfying the appointment requirement without producing the governance effect M-25-21 is asking for. A governance board that reviews policy without examining the operational record of AI-assisted work in the agency is satisfying the structure without exercising the discipline. The agencies that produce the governance outcomes the federal direction expects are the agencies whose governance roles are connected to the workflows that the AI assists.
These three travel together. The workflow leads the AI deployment. The portal and the workflow architecture complement one another. The governance is operational rather than nominal. Agencies and the GovTech architects working with them, observing this pattern, tend to find that the federal policy direction and the operational reality converge naturally when these three conditions are in place
GovSoft works with state and local agencies, and with the businesses operating in regulated industries that serve them, on the workflow architecture this direction calls for. We layer governed workflow operation above the portals, case management systems, and program systems agencies already operate, with AI as a governed support layer inside workflows the agency has defined and the leadership has approved. The same governance principles M-25-21 establishes at the federal level inform the architecture we build at the state and local level, with no upfront fees and a structure where you pay from the operational value the work produces.
If your agency is operating a portal that informs but does not move the work behind it, or carrying workflows that the existing systems were not built to govern, GovSoft is a conversation worth having.
Learn more at govsoft.us
Secure cloud deployment is more than modernization — it’s the backbone of citizen-focused digital governance.
Your modernization briefing is on the way.
We’ll connect you with our partnership team.
We’ll follow up about your workshop.
We’ve received your info and will connect you with the right team.
.svg)
.svg)
.svg)
.svg)
.svg)