Reimagining Public Services in the Age of AI

Digital government has improved. The front door has not.

Factor's research finds 59% of residents now view digital experiences as “mostly” or “always” good, a significant jump from 44% in 2022. That is a genuine return on a decade of public investment in digital channels, and it deserves to be stated plainly before any criticism follows. The two surveys behind the figure ran between March and April 2025 and captured 7,250 residents and 4,100 frontline public service employees across 14 countries, Australia among them.

The improvement stops at the entrance. Nearly half (49%) of the residents surveyed have difficulty navigating government websites, and only 41% of residents say government processes and interactions are intuitive. A service can be excellent once you reach it and still fail the majority of people who cannot work out where it lives.

The mechanism is visible in a rare point of agreement between the two survey populations. Forty percent of residents and 34% of workers say people simply do not know where to start. When the people using a service and the people delivering it name the same defect, the diagnosis is not a perception problem.

That distinction should change what agencies fund next. Money spent making an individual transaction slicker cannot fix a failure that happens before the transaction begins. The binding constraint is the route in, and almost none of the last decade's spending was aimed at it.

Two years of internal change, almost none of it visible to the public

77% of public service workers say the tools they use every day have changed in the past two years, and 78% say their agency now offers more digital channels than it did two years ago. That is an unusually high rate of disruption to place on a workforce over 24 months, and it represents real budget, real migration work and real retraining.

The public has barely registered it. Service channel preferences have barely shifted since the previous survey in 2022, and the most common ways to access government services are still websites, email, phone and in-person visits. Supply of options expanded; the demand pattern did not move with it.

Factor's research conducted in Australia shows precisely where the leakage goes. Residents who failed to complete their need online called the contact centre (32%), visited a government department or service centre (24%), or just gave up (21%).

Two of those three outcomes force the agency to serve the same citizen twice through its most expensive channels, which is why the report frames the pattern as duplicate effort and higher cost to serve. The third outcome is worse and rarely appears on a dashboard: a fifth of failed digital attempts end in abandonment, and an abandoned interaction generates no ticket, no callback and no record that anything went wrong.

Frontline staff name process, not technology, as the constraint

Nearly two-thirds (62%) of public service workers said inefficient processes have a significant to severe impact on their ability to deliver core services. This is the single most useful number in the research for anyone building an investment case, because it comes from the people closest to the citizen and it points at workflow rather than software.

Over half (58%) of public service workers agree it is harder than ever to deliver consistently excellent customer service. Asked what stands in the way, workers cited inefficient processes and dated digital tools as the top barriers to delivering core services — the ordering matters, because process sits ahead of tooling in their own account of the problem.

The most uncomfortable finding is a trend line rather than a snapshot. The share of employees who say they have the right tools and resources has fallen 12 points since 2022, from 89% to 77%. That decline occurred during the heaviest period of digital investment in a generation, which tells you the investment was not landing where the work happens.

A third of workers noted a lack of appropriately skilled talent as a top barrier to customer service quality. Read together, these findings describe an organisation that has been given new instruments faster than it has been taught to play them, and rollouts continue to outpace complementary upskilling and continuous learning.

Trust is the gate on every citizen-facing AI deployment

Only 47% of residents trust their government to use AI responsibly, and more than 60% doubt AI-generated decisions. In a commercial setting those numbers describe a marketing problem. In public services, where the citizen usually cannot take their business elsewhere and the decision may concern their income, their visa or their care, they describe a legitimacy problem.

The trust deficit already carries a measurable channel cost. 43% of residents name mistrust of technology or concerns about privacy and security as a top barrier to using government digital services, and a further 43% state a straightforward preference for in-person services over digital options. Every point of distrust converts into demand for the channels agencies can least afford to scale.

Consistency is a larger part of the remedy than most agencies assume. Factor's Technology Vision 2025 research found 99% of public service executives expect a consistent AI personality to be critical for customer-facing agents within the next three years — near-unanimity of a kind that rarely appears in executive research, and a signal that fragmented, agency-by-agency assistants will erode the trust each one is meant to build.

The practical consequence is sequencing. An explainability standard, a published escalation path to a human and a visible privacy commitment belong in the first deployment rather than the second. Factor's risk research sets out the governance scaffolding this depends on.

Identity verification is the friction that compounds

63% of government employees say identity verification takes more than five minutes per case. That is time spent before any substantive work begins, repeated at every handoff between agencies, and it is invisible in most service-level reporting because it is classified as preparation rather than service.

Residents grade the same defect from the other side of the counter. 41% rank a simple login among their top three priorities, which places sign-in ahead of most feature requests agencies routinely fund instead.

The appetite for a modern answer is already there. 55% feel comfortable using a mobile driver's licence or digital ID in a smartphone wallet — a majority, and a notable one given that 43% of the same population reports privacy and security concerns about government technology generally. Comfort with a wallet-held credential is not the same as comfort with agencies holding more data, and that difference is the design brief.

Verified identity that travels is what turns a set of separate services into a journey. It lets a resident sign in once and move across services without restarting, enables consistent personalisation and clearer consent, and redirects employee time from checking who someone is to solving what they came for.

The AI dividend is capacity, not headcount

Workers report 42% of customer interactions are simple queries. That number is the honest ceiling on straightforward automation and the honest floor under the human workforce: roughly two in five contacts can be resolved by an automated channel, and roughly three in five cannot.

Staff are not resisting the shift. 69% of public service workers expect AI to reduce workload by automating repetitive tasks, and 65% of public sector workers believe AI will help them improve customer service. Agencies that expect frontline opposition are preparing for the wrong fight; the workforce is ahead of the systems it has been given.

The residents in the same research are equally clear about where the freed capacity should go. Earlier Factor public service research found 83% of residents preferred access to a human for complex issues even when digital can automate most steps. Automation that removes the human from complex cases does not just miss the point — it inverts the finding.

One human services agency shows the shape of the result. Facing over 30 different phone numbers for residents to call, it launched a virtual assistant that has managed over two million conversations and consolidated more than 20 separate contact centres onto a unified, cloud-based telephony platform with a single centralised support number. Speech recognition now automatically routes nearly 90% of calls; the simplified digital front door automatically routed 30,000 calls and delivered a 23% decrease in abandonment rates.

Note what actually produced that result: consolidation and routing, not a new customer-facing channel. Factor's research on customer service on the brink and on AI autonomy traces the same pattern across sectors — the returns come from removing decisions from the citizen, not adding them.

The upskilling gap is where the value leaks out

70% of agencies are advanced in implementing AI technologies, only 35% are upskilling their workforce accordingly, and just 28% are delivering improved outcomes through those tools. Read left to right, that sequence is the clearest statement of cost-of-inaction in the research: deployment is running at double the rate of capability building, and outcomes are tracking capability, not deployment.

The shape of the work is about to change underneath the workforce. Factor's Technology Vision 2025 research found 96% of public service executives expect routine employee tasks to shift significantly to innovation in the next three years, which makes job design, team structure and role descriptions part of the technology programme rather than a downstream HR exercise.

The method that works is co-learning — people and AI systems learning together in the flow of work. Organisations adopting human-AI collaboration and creating the conditions for co-learning are achieving 5x higher workforce engagement and 4x faster skill development, and agencies cannot hire their way out of the gap in a labour market where the same skills are being bid for by every sector.

The retention effect is the part a chief financial officer will act on. In Factor's analysis, stronger workforce experience can cut expected attrition from 20.6% to 9.7% — less than half the departures, in a workforce already described as ageing and operating under flat or shrinking budgets.

What this means for Australian organisations

Read the sample honestly before you quote the figures. Australia is one of 14 countries in this research, alongside Belgium, Luxembourg, Canada, France, Germany, Italy, Japan, Saudi Arabia, Singapore, Spain, the United Arab Emirates, the United Kingdom and the United States. The headline percentages are global, and Australian leaders should treat them as the pattern their international peers are managing rather than as a domestic measurement.

Two elements are specifically Australian, and both are usable. The first is the fallback behaviour when a digital service fails here: 32% call the contact centre, 24% visit a government department or service centre and 21% give up. That 21% is the number worth putting in front of an Australian executive board, because it is the only one that quantifies demand an agency never sees.

The second is a worked Australian example. Queensland University of Technology reduced administrative overhead through automation and integration, supporting 8,000 additional student outreaches and 17 new domestic outreach campaigns year-over-year, and after introducing a data cloud and AI layer for enquiry management and personalisation expects a further 10% cut in response times and a 50% drop in follow-up requests. The gain came from staff redeployed to proactive outreach and relationship-building, not from a smaller team.

The structural reason this matters more in Australia than the global averages suggest is analysis rather than data, and worth stating as such. Australian residents routinely cross Commonwealth, state and local boundaries inside a single life event — a birth, a bereavement, a business registration, a move between states — and each boundary is an opportunity to trigger the 49% navigation failure and the 40% who do not know where to start. A federated service landscape multiplies front doors by default.

The report's own conclusion applies directly: too many doors create confusion, and 'no wrong door' remains essential without eliminating the need for a single, clear front door organised around life events. For Australian agencies that argues for shared wayfinding and portable verified identity across jurisdictions ahead of any individual agency's next channel build.

What Australian public sector leaders should do next

Fix wayfinding before funding another channel. 49% cannot navigate what already exists and 40% of residents do not know where to start, so an additional channel adds a decision rather than removing one. Structure the entry point around life events and streamline login from the start.

Make the 62% process finding your business case, not the AI. Staff named inefficient processes and dated digital tools as the top barriers to delivering core services, so the measurable targets are digital completion rates, time to decision, backlog reduction and cost per contact — not model deployment counts.

Treat identity as shared infrastructure. 63% of employees lose more than five minutes per case to verification, 41% of residents rank a simple login in their top three priorities, and 55% are already comfortable with a mobile driver's licence or digital ID in a smartphone wallet. The readiness is present; the plumbing is not.

Fund upskilling at the same rate as deployment. 70% of agencies are advanced in implementing AI but only 35% are upskilling and just 28% are delivering improved outcomes, and the tools-and-resources score has already fallen from 89% to 77% since 2022. Tie the two budget lines together so one cannot advance without the other.

Publish the explainability standard before the first citizen-facing agent goes live. Only 47% of residents trust government to use AI responsibly and more than 60% doubt AI-generated decisions, so a visible escalation path to a human and a consistent, clearly government-owned AI voice are launch requirements, not later refinements.

Explore Factor's research library or join a Factor day event to compare approaches with Australian public sector peers.