Hazeldenes: AI Agents Redefining What a Lean Finance Team Can Do

Hazeldenes is a PE-backed agricultural, poultry, and Fast-Moving Consumer Goods (FMCG) group whose brands are stocked across major Australian retailers.With a clear mandate from the board to find operating leverage through technology, the finance leadership set out to maintain reporting quality with lower overhead while improving the working experience of the team.

‍

Rather than framing the initiative around cost-cutting the focus was the people whose time gets freed up. FloQast Transform was selected because it enables them to build and control their own AI agents end to end, with auditability at every step. Multiple agents are now live, a roadmap of further agents  is underway, and the team is already shifting from data preparation to strategic review and business partnering.

‍

The Challenge: Delivering More with Less

The problem was familiar to any controller. Skilled accountants were spending significant portions of their week on manual data work: pulling inputs, manipulating spreadsheets, producing journal entries, and reconciling sub-ledgers. These were structured, repeatable tasks running  on weekly or monthly cycles, leaving little room  for the higher-value work the business actually needed, such as analysis, business partnering, and addressing control findings that had been deprioritised while the team was occupied keeping BAU running.

‍

The framing within the finance leadership team was deliberate. The hours matter, but what matters more was what the team could do with those hours back. When AI handles the repetitive, rule-based work, the accountants move up the value chain. Preparers become reviewers, and the quality of financial reporting improves because the people closest to the numbers finally have time to think about what those numbers mean.

‍

{emphasize}

"It takes our senior accountants out of the weekly data work - the hours spent manipulating Excel to produce journals - and turns them into reviewers of that work instead."

— Ben Cirillo, Group Financial Controller, Hazeldenes

{emphasize|

‍

The Solution: AI Agents Built by Accountants, for Accountants

The evaluation criteria were practical. The solution needed to automate real accounting processes, not simply add dashboards or workflow visibility, and it had to be something the team could own. No black box, no consulting project, and a governance model that could withstand board scrutiny and audit review.

‍

FloQast Transform met those requirements. The platform is ISO 42001 certified, and the human stays in the loop throughout with the accountant remaining the owner and reviewer of the work.

‍

The rollout prioritised the processes where AI agents would deliver the most immediate ROI, the most frequently repeated and consistently structured tasks. Automating something performed weekly compounds faster than automating something performed monthly or quarterly, so the highest-frequency, most rules-based workflows were targeted first.

‍

{emphasize}

"We started with the low-hanging fruit — routine tasks repeated weekly or monthly, where the process was consistent and was currently being done by an accountant. And to get the most ROI, we targeted the most frequently repeated ones first."

— Ben Cirillo, Group Financial Controller, Hazeldenes

{emphasize}

‍

We partnered with a FloQast Forward Deployed Accountant (FDA) to co-build the five agents, defining the workflow logic, mapping each step, and testing agent outputs against real data. The accountants owned the process knowledge while FloQast brought the platform expertise. That co-building model meant the team understood exactly what each agent was doing from day one with no handoff gap, no black-box implementation.

‍

Every agent includes human review steps. A preparer still takes ownership, and a separate reviewer validates the output, preserving segregation of duties. The controls baked into each workflow provide full traceability and keep the accountant in the driver's seat at all times.

‍

Strategic Impact: A Team Transformed

The live agents are saving Hazeldenes approximately 600 hours annually, covering allocation, reconciliation, and journal entry workflows across both weekly and monthly cadences. Further agents are planned to extend the model into month-end processes, where time pressure is most acute. The team is reclaiming meaningful hours across the close cycle and has been able to absorb natural attrition without backfilling, in line with the board's mandate to deliver more with less while maintaining output quality.

‍

‍

An additional benefit was the impact to the processes themselves. Building an agent requires prescriptive documentation of every step, which forced the team to examine workflows that had been running on autopilot for years. Legacy steps were removed, inputs that other teams had been provided were rationalised, and handoffs across the organization were simplified. 

‍

As a next step, the team is now working to layer analytical checks into the agents themselves that will flag values outside expected ranges and surface accounts with unexpected zero balances, catching potential omissions before they post. These checks will give reviewers a second layer of confidence and effectively turn each agent into a first-pass reviewer.

‍

{emphasize}

"It helped us look at our processes from first principles and rebuild them from the ground up. We actually reduced some inputs that other teams had to provide us, which made things easier for everyone."

— Ben Cirillo, Group Financial Controller, Hazeldenes

{emphasize}

‍

Where the change really shows up is in how the team spends its time. Senior accountants are no longer buried in weekly data manipulation. They're reviewing agent output, partnering with the business, and working through control findings that had previously been deprioritised. The roadmap for the next phase is already in development with a particular focus on month-end processes, and there is appetite to extend the model beyond finance into other functions where routine, structured workflows exist.