The productivity problem hiding in plain sight

3 signs complex work structures (not people) are blocking productivity improvements, and how AI-enabled process redesign can fix it? Liquid Partner Steven Marrinan provides the answers in a new article published in The Mandarin, following release of our AI roadmap for Australia’s public service. He focuses on why leaders need to first understand where friction site and effort is consumed.

Steven Marrinan

Powered by
Liquid
On the back of a year-long ‘AI in Government’ benchmarking initiative in collaboration with The Mandarin, Liquid last week released ‘Speed. Safety. Scalability. A roadmap for responsible AI adoption in Australia’s public service’. In this article, Liquid Partner Steven Marrinan focuses on why leaders need to first understand where friction sits and effort is consumed.

Removing friction inside government
Across government and industry, leaders are grappling with the same challenge: how do we make better decisions, faster, without compromising quality, consistency, or trust?

In response, organisations put processes in place to ensure the right information is considered, the right criteria are applied, and decisions are made consistently.

But what starts out as a logical assurance process can change as systems expand, demand increases, and scrutiny intensifies. Not deliberately, but predictably.

More steps are introduced. More checks are added. More people become involved in reviewing the same information. And, over time, something subtle but significant happens.

Effort shifts away from judgement, and toward preparation. Away from decision-making, and toward assembling, validating, and re-validating information.

In the process, risk management gradually turns into administrative burden. Notably, the No.1 pain point — and hope for where AI can make a difference — identified by public servants in Australia’s first national survey into public sector attitudes towards AI, conducted by Liquid and The Mandarin.

3 signs to look for (and questions to ask)
The following questions can help leaders identify where duplication, fragmentation, and manual effort are masking deeper structural issues that impact productivity, workplace wellbeing, service delivery and compliance outcomes.

#1 Your people are buried in paper trail.
Where does effort surface as administrative burden rather than meaningful outcomes?
#2 Your organisation depends on certain individuals to “hold it together”.
Where is this happening? If a key person left tomorrow, where would decision-making capability degrade?
#3 Your organisation has a wealth of data but finding the right information requires effort.
What information are you sitting on, and how well are you using it?
Productivity is constrained by structure, not capability
What becomes clear is highly skilled workers do not lack expertise or effort. They are constrained by how their time is used.

Effort is spent locating information, stitching it together, and checking it repeatedly, often to support decisions where both the evidence base and the criteria remain unchanged. As demand grows, effort scales with volume, but assurance doesn’t improve proportionally.

At Liquid, we’re seeing this across government departments and regulatory agencies through our AI Leap approach: working with leaders and teams to collaboratively problem-solve and identify practical ways AI systems can tackle complex problems at scale.

Unlocking productivity requires a shift in how organisations think about work itself.

This means asking the right questions to understand the actual work to be done and real problems to be solved; the purpose and functions of different roles; and where effort is being consumed and why.

Incremental change and isolated automation are not enough
This is where many organisations get stuck. Faced with this challenge, they turn to familiar levers: more effort, process refinement, or targeted automation.

Isolated automation can remove friction in individual steps, but it doesn’t fundamentally change how work is performed. If the same evidence continues to be checked multiple times, and the same rules applied repeatedly, automation simply redistributes effort.

Without a shared, systemic foundation, these interventions treat symptoms rather than the cause. Work may move between teams or stages, but it continues to scale with volume, and the underlying pattern of duplicated checks and manual rule application remains intact.

A trusted, secure AI capability enables consistency at scale
Rather than optimising individual steps, leading organisations are intentionally designing enterprise AI systems that can be applied across entire classes of work, particularly high-volume, high-repetition processes.

This is not about isolated tools or experimental use cases, but building a trusted, secure, and reusable capability that becomes part of how work is performed.

At its best, this capability is defined by a few key characteristics:

It handles evidence consistently across workflows
It applies decision criteria reliably and at scale
It embeds assurance directly into the flow of work
It operates within defined controls, governance, and security boundaries
It produces outcomes that are transparent, traceable, and auditable
When this is done well, the nature of work changes. Teams no longer begin by gathering and validating information. They start with cases where relevant evidence has already been collected, structured, and validated within the workflow.

The starting point shifts from a blank page to a complete view. As a result, expert human effort moves away from preparation and toward review, oversight, and decision-making.

Perfect data isn’t required to get started
At this point, organisations raise a familiar concern: their data isn’t ready.

Years of accumulated information, inconsistent formats, and unstructured records create a perception that significant cleansing is required before meaningful progress can be made.

In practice, waiting for pristine data unnecessarily delays progress.

Modern AI systems are capable of working with imperfect, incomplete, and unstructured information. Rather than relying on fully standardised inputs upfront, they can interpret, organise, and apply structure to data as it is used within the workflow.

In doing so, they reduce the need for large-scale, manual data preparation as a prerequisite for change.

AI expands what counts as usable evidence
Much of the information organisations hold is technically available, but practically inaccessible: locked in scanned documents, free text, or legacy formats that are difficult to review at scale.

Intentionally designed AI systems can convert this into structured, reviewable inputs, making previously underutilised information part of the decision-making process. In many cases, this leads to a more complete and consistent evidence base that is less dependent on selective manual extraction.

The opportunity, however, is not just to make better use of existing data. It is to rethink the role of data within the process itself.

Many organisations collect large volumes of information to compensate for fragmented workflows and low system trust, resulting in duplication, manual handling, and, at times, data that adds limited value to decision-making.

By redesigning workflows alongside AI capability, organisations can focus on collecting the right information once and apply it consistently. This reduces unnecessary effort, improves the relevance of the evidence base, and shifts data from being a constraint to an enabler of better, faster decisions.

Assurance is embedded by design
As outlined in Liquid’s AI roadmap, many early AI adopters have been motivated by the opportunity to reduce data effort.

This is where AI-enabled process redesign can have immediate value by delivering a responsible AI system configured as a trusted, secure and reusable capability which becomes part of how work is performed.

Human-at-the-centre controls operate within the workflow itself, shaping how evidence is handled and how decisions are assessed. Trust is not placed in outputs alone, but in the way they are produced.

In AI-enabled workflows, this means designing around what the model does well, rather than relying on it to make decisions.

A well-designed AI system:

Assembles and summarises the relevant evidence
Highlights the applicable rules and criteria
Shows the reasoning steps taken
Presents this clearly to the person making the judgement
During periods of peak demand, pressure also changes. In AI-enabled workflows, repeatable work scales automatically. Evidence handling, classification, and baseline checks continue to operate consistently, regardless of volume.

Operational strain reduces precisely when demand is highest, and staff can remain focused on what matters most.

Moving from reactive operations to proactive advantage
Organisations that address structural problems create strategic advantage. Instead of responding to demand, risk, or issues after they emerge, they gain the ability to anticipate, surface, and act earlier.

Patterns become visible sooner. Risks are identified before they escalate. Decisions are informed by a continuously updated and consistently applied evidence base.

For the public sector this shift is vital: redirecting effort away from repetitive manual workload, freeing skilled people to focus on oversight and higher value decision-making that directly impacts communities.

Download Liquid’s practical AI roadmap and start a conversation with Steven Marrinan via steven@liquidinteractive.com.au

The productivity problem hiding in plain sight

Related content

Sign up to our newsletter.