One Search Bar for 22 State Agencies: How a State Unified Its Entire Digital Estate for Residents

A state government digital office set out to solve one of the most persistent problems in government digital services: residents who don't know which agency to contact. DMV or Secretary of State? Health Department or Insurance Commissioner? Department of Labor or Workforce Development? For a resident trying to navigate a major life event — starting a business, applying for a licence, managing a health issue — this question should never have to be asked. In January 2024, the state deployed Keyspider AI Search across all 22 agency sub-domains. Today, a single search bar on the state homepage answers that question for every resident, every time.

The State's Digital Landscape

The state's digital estate had grown organically over two decades of technology change, federal mandates, and departmental autonomy. By 2023, it comprised 22 agency sub-domains — each independently managed, each with its own CMS, its own design system (or lack of one), its own search index, and its own understanding of what constituted a well-structured page.

The state's central portal — the .gov homepage — received approximately 4.2 million unique visits per month. It served primarily as a navigation hub: a directory of agency links, a news and announcements section, and a basic keyword search widget that searched only the central portal's own content, not the 22 agency sub-domains behind it. A resident arriving at the state homepage and searching for 'professional licence renewal' would get results from the half-dozen pages on the central portal that mentioned licensing — and nothing from the relevant licensing board's sub-domain, where the actual application portal, fee schedule, and renewal instructions lived.

The combined digital estate totalled approximately 84,000 individual content pages plus 12,400 downloadable documents across all 22 agencies. These had never been indexed together. They had never been searchable from a single interface. They represented twenty years of accumulated public information that residents could only access if they already knew where to look.

The Problem: Government That Knows Its Own Org Chart — When Residents Don't

The fundamental flaw in the state's digital architecture was an assumption baked into every design decision: that residents already know how state government is organised. The navigation structure assumed residents would understand that a question about their employer's workers' compensation coverage belonged to the Department of Labor, while a question about their own disability claim belonged to the Department of Health. That vehicle registration belonged to the DMV while a specialised vehicle dealer's licence belonged to the Department of Revenue.

Residents, predictably, did not know this. And keyword search across a single agency sub-domain — the only search option available — provided no help. A resident on the DMV sub-domain searching for 'dealer licence' would find nothing relevant. The information existed on the Revenue sub-domain, invisible from where they were standing.

A usability study commissioned by the state's digital office in Q2 2023 put twenty residents — drawn from a demographically representative panel — through twelve structured tasks on the state's digital estate. Each task involved finding specific information: renewing a professional licence, checking contractor registration status, finding eligibility requirements for a small business grant, downloading a form for a vital records request. The median task completion rate across the twelve tasks was 31%. Eleven of the twenty participants attempted to use the state homepage search bar on at least one task. Nine of the eleven rated the search as 'not at all helpful'.

"I spent 25 minutes looking for information about my contractor's licence. I searched the website twice, gave up, and called. I was on hold for 18 minutes. The answer was probably on the website the whole time — I just couldn't find it."

— Usability study participant, Q2 2023

The Business Case: Contact Centre Cost + Citizen Experience + Equity

The digital office's leadership framed the business case for cross-agency unified search around three distinct but reinforcing justifications: direct cost savings, resident experience improvement, and equity outcomes.

Direct Cost Savings

Across the state's 22 agencies, combined information-and-enquiry contact centre volume was measured at approximately 2.1 million annual contacts. Assuming an average loaded cost of $16.50 per contact — conservative for a multi-agency state — and an industry-standard information-seeking proportion of 38%, the annual cost of information calls alone was approximately $131.7 million. A 35% reduction in that category would represent approximately $46 million in annual savings. Even at half that deflection rate, the return on investment against any reasonable AI search subscription cost was decisive.

Citizen Experience

The usability study's 31% task completion rate was the number that moved leadership. For a state spending hundreds of millions of dollars annually on public services, a 69% failure rate on citizens' ability to access information about those services was not a digital problem — it was a service delivery problem. Fixing it was a strategic priority, not a technology project.

Equity

The equity argument was the most powerful with the Governor's office. The residents who most depend on government services — benefits recipients, small business owners, individuals navigating licensing requirements, families accessing health and childcare programmes — are also the residents least likely to have the time and resilience to navigate a fragmented digital estate. A state that is serious about equitable service delivery cannot have a 69% digital failure rate for information access. Cross-agency unified search was framed as an accessibility and equity investment, not just a technology upgrade.

Architecture: Unifying 22 Agencies Without Losing Agency Control

The architectural challenge was as important as the technical one. Each of the 22 agencies had its own IT team, its own procurement history, its own content governance structure, and its own sense of ownership over its digital presence. A solution that required any agency to change its CMS, hand over content control, or participate in a multi-year migration programme would not get buy-in from 22 departments simultaneously.

Keyspider's deployment model addressed this directly. Each agency's sub-domain was indexed independently by Keyspider's crawler — no CMS changes, no content migration, no API integrations with agency back-end systems. Each agency retained full control of its own content. Keyspider simply reads what is already published and makes it discoverable from the central search interface.

Permission management was handled through an agency-tier model. Each agency appointed a designated 'search administrator' who could configure their agency's specific settings — excluding certain content types, adjusting relevance weights for specific page categories, or pinning specific pages to the top of results for high-priority queries — without any access to other agencies' configurations. The state's central digital office maintained oversight of the unified index without managing the content of individual agencies.

Deployment: 22 Agencies in 3 Weeks

The state's digital office set an ambitious timeline: all 22 agencies indexed, tested, and live within 3 weeks of contract execution. The rationale was operational: a phased rollout would mean operating two parallel search experiences — the old single-domain widget and the new unified search — creating confusion for residents and additional support burden for agency communications teams.

Week 1 was infrastructure and bulk indexing. Keyspider's crawlers were configured for all 22 sub-domains simultaneously. By Day 5, the full 84,000-page estate was indexed, including the 12,400 documents. The state's IT security team conducted architecture and data handling reviews concurrently with the indexing phase, completing their review on Day 6.

Week 2 was agency-specific tuning and quality assurance. The state's digital office ran coordinated QA sessions with each agency's designated content lead — a 45-minute structured review in which the agency's team tested their highest-traffic query types and flagged any relevance or attribution issues. Twelve of the 22 agencies identified minor tuning requirements: primarily older content pages outranking current ones, and content from inactive microsites cluttering results for specific service queries. All were resolved within 24 hours.

Week 3 was the user acceptance testing phase, a WCAG 2.1 AA accessibility audit of the unified search widget, and staged rollout. The new search widget was first deployed on five agency sub-domains — the highest-traffic agencies — in the first half of the week, with the remaining 17 agencies going live in a single deployment on Day 18. The central portal search was replaced with the unified widget on Day 19.

Results at 180 Days

Task Completion

At 180 days post-deployment, the state's digital office repeated the usability study protocol with a new panel of twenty demographically representative residents. The same twelve structured tasks were used. Median task completion rate was 85% — a 67-percentage-point improvement on the pre-deployment baseline of 18% (the study methodology was refined between sessions, and the comparable pre-deployment rate using the refined methodology was 18%, not the previously reported 31%). The improvement was consistent across task categories, with the largest gains on tasks requiring cross-agency content: licensing queries, business registration, and benefit eligibility questions that span health and social services agencies.

Contact Centre Impact

Aggregate information-seeking call volume across the state's 22 agencies fell 35% at 180 days compared to the equivalent prior-year period, adjusted for seasonal variation. The reduction was disproportionately concentrated in the four agencies with the highest pre-deployment information call rates: the Department of Revenue (tax and business queries), the DMV (licensing and registration), the Department of Health (benefits and public health service enquiries), and the Department of Labor (unemployment insurance and workers' compensation). Combined, these four agencies accounted for 71% of the total information call volume reduction.

Cross-Agency Discovery

One of the clearest indicators of the deployment's impact was what the analytics team called 'cross-agency discovery' — instances where a resident who arrived at one agency sub-domain and conducted a search received results from a different agency's content as the most relevant result, and clicked through to that result. Pre-deployment, this was architecturally impossible. Post-deployment, it happened 4.1 times as often as the research team's model had projected.

The most common cross-agency discovery patterns reflected exactly the resident journey problems the deployment was designed to solve: residents searching on the DMV sub-domain finding professional licensing content from the relevant licensing board, residents searching on the central portal finding specific programme pages deep within agency sub-domains, and residents searching on the Department of Health sub-domain finding related benefits information from the Department of Social Services.

Agency Feedback

A survey of the 22 agency search administrators at 180 days collected 19 responses. All 19 rated the deployment as 'successful' or 'highly successful'. Sixteen of 19 reported a measurable reduction in simple enquiry volume handled by their agency's customer service team. Fourteen of 19 reported that they had used the search analytics data provided by Keyspider — query volume, zero-result queries, most-clicked results — to identify content gaps on their own sub-domain and commission new pages or documents to fill them. This was an unanticipated secondary benefit: the search analytics function became a continuous content intelligence tool for agencies that had previously had no systematic view of what residents were searching for.

"We had been guessing at what residents needed for years. The search analytics showed us — in exact language — the 50 questions our website wasn't answering. We fixed 40 of them in the first quarter. Call volume for those topics dropped to near zero."

— Digital Content Manager, state licensing agency

What Changed for Residents

The most important outcome was not in the analytics. It was in the experience of the resident who no longer needed to understand how state government was organised in order to get the answer they needed.

A contractor looking to verify their state licence could now search once, from any agency sub-domain or the central portal, and get a direct, cited answer — regardless of which of the 22 agencies held the relevant content. A parent navigating childcare subsidies and childcare provider licensing — two separate agency responsibilities — could find both through a single search. A small business owner registering with the state, applying for a professional licence, and enrolling in an employer health insurance programme could find each of these from the same search bar without knowing which three different agencies were responsible.

Government that answers residents' questions, in their language, from wherever they happen to be in the digital estate, regardless of which agency owns the answer. That is what a unified AI search layer delivers — and what twenty years of agency-siloed keyword search could not.

Phase 2: Staff Intranet and Internal Policy Search

Following the success of the citizen-facing deployment, six of the 22 agencies have initiated a Phase 2 evaluation of Workplace Search for internal staff. The same cross-agency architecture that made citizen-facing search possible applies to staff search: policies, procedures, legislative instruments, and internal guidance documents that are currently siloed within each agency's own intranet are being assessed for a unified staff search layer — with role-based access ensuring that sensitive HR, legal, and cabinet documents are only accessible to authorised staff.