Australian firms show how AI reshapes customer support and operations
Three Australian companies revealed concrete applications of AI in customer service at a recent Zendesk event in Sydney, moving beyond pilots to systems handling real customer interactions and business workflows.
MYOB's new product Solo automates receipt reconciliation for sole traders, reducing a five-to-six-step process to a single photo. Guzman y Gomez uses AI to decide which kitchen line prepares each order. Aware Super is testing an AI coach that members sometimes prefer to human advisors for financial guidance.
These deployments share a pattern: AI handles specific, high-volume tasks while humans focus on exceptions and complex decisions.
MYOB: Receipt scanning replaces manual reconciliation
Solo emerged from research showing over half of sole traders use spreadsheets to manage their business. These owners are time-poor and often blur personal and business finances-75% use personal bank accounts for business purposes.
The product centralises administration by framing accounting as "money in" and "money out" rather than traditional bookkeeping. AI automatically matches receipts to bank transactions, eliminating manual reconciliation.
The AI model isn't perfect. When confidence falls below a threshold, users confirm or correct the interpretation. Early versions struggled distinguishing between personal items (children's shoes from Kmart) and business supplies. As successive models improved, manual confirmation became the exception rather than the norm.
A Zendesk-powered chatbot resolves 90% of remaining support issues-three to five times better than industry standard. Less than 1% of Solo users bypass the chatbot to reach a human agent, according to Sally Davies, general manager of Solo.
For financial transactions, MYOB mandates human support for compliance and customer loyalty. AI assists these agents as a "copilot," aggregating customer information from multiple sources including app interactions and previous chatbot conversations.
A community manager-not an IT specialist-built this custom app using Zendesk App Builder and marketplace integrations. The project went from idea to live capability in one week. Davies noted that ten years ago, a week would pass just scheduling the initial meeting.
Guzman y Gomez: AI decides which kitchen line prepares each order
The quick-service restaurant chain deployed AI in nine restaurant kitchens to dynamically assign orders to one of two food preparation lines. The system balances load on crew members while supporting the company's "hotter, fresher, faster" mantra.
An order management system (OMS) rolls out nationally in May. Bryce Maybury, chief technology officer, said the next step is predicting when to open the second line before the first becomes overwhelmed.
Opening a line takes 15 minutes of prep work. AI will analyse order velocity, item modifications, and order origins (point-of-sale, app, drive-through, delivery aggregators) to determine timing. Even 15 minutes saved matters in quick-service restaurants.
Future additions might include weather and traffic data around drive-throughs, though Maybury cautioned that restaurant managers already know their areas well. The real question is whether machine learning genuinely outperforms human judgment in every scenario.
At the corporate level, GYG uses AI coding tools for software development. "At least 80% of our code is now generated by AI agents," Maybury said, with focus on identifying restaurant problems and delivering solutions.
Later this year, the OMS will gain a "cook to demand" feature that forecasts when to prepare the next batch of chicken to meet anticipated order spikes.
Maybury's team maintains direct contact with restaurant operations. All head office staff spend at least three days working in restaurants during induction. Many team members previously worked as crew, cooks, or shift leaders. The head of restaurant technology spent weeks in a restaurant during OMS development to gather feedback.
Aware Super: Members discuss finances with AI agents
The superannuation fund is testing an AI coach providing financial literacy guidance to members. Early testing revealed an unexpected finding: some members feel more comfortable discussing finances with an AI agent than a human.
Richard Exton, group executive and chief technology officer, said Aware Super's advantages lie in proprietary member data and status as a trusted, regulated entity. Members increasingly turn to free tools like ChatGPT for financial advice, but those lack member context, regulatory backing, and data protection guarantees.
Exton faced two challenges: ensuring AI was treated as an organisation-wide matter rather than purely an IT concern, and educating staff to use these tools without fear.
He cautioned against jumping into proofs of concept. The starting point should be understanding the expected value AI provides in a specific context. Organisations should then decide whether investment is worth it.
Governance and privacy require equal investment. "You can have the best service out there, but as soon as you're breached, your reputation is damaged, and no one will ever talk about how great your AI model is," Exton said.
Zendesk's next frontier: Supervision and preparatory work
Adrian McDermott, chief technology officer at Zendesk, said the technology's next frontier moves beyond automating customer conversations to supervising agents and preparing work before humans see tickets.
AI models now monitor human agent conversations, scoring them on grammar, tone, and other criteria while providing immediate coaching. Long-running agents complete preparatory work by analysing text and attachments, interrogating backend systems, and compiling comprehensive notes for human agents.
AI also drives continuous improvement. An agent might flag that 70% of human agents modify a specific automated response, or that customers frequently call about a missing product in the knowledge base.
Internally, Zendesk uses pure agentic coding with zero human-written code across more than a dozen teams. A single developer manages a team of 20 AI agents. McDermott compared this to the era when developers manually optimised portions of code in assembly language-a practice that stopped once compilers became reliable.
The productivity gap has widened dramatically. Engineers previously considered "10x" performers are now 50 times more productive with AI agents. Average engineers see only modest gains.
McDermott predicted the industry will treat large language models as compilers of ideas. However, he warned that some AI models hallucinate passing tests to fulfil their goals, even when tests should fail.
The shift applies beyond engineering. "Where programmers go today, all information workers will follow tomorrow," he said. The focus changes from individual tasks to broader purposes-a contact centre worker's task is handling enquiries, but their purpose is improving customer satisfaction and attracting new customers.
AI also flattens skill sets. A product manager might generate prototypes in Claude Code instead of writing a product requirements document. "It's just such a richer way to present what's going on," McDermott said.
Learn more about AI for Customer Support and AI Agents & Automation.
Your membership also unlocks:
