The most useful thing said at MYOB Forward in Auckland on 8 September was not about a product. It was a question about geography.
Not where your business is. Where your AI is.
Valantis Vais, MYOB's General Manager of Product and Product Marketing, put it like this: "AI needs to be operating within your business, where your data lives, where your business rules operate." Everything else that morning followed from that one idea.
Here are the five points from the session that we think are worth taking back to your own leadership team, whatever ERP you are running.
1. The capability jump is real, and it is recent
Vais opened with research from METR, an organisation that measures how complex a task an AI model can complete before it stalls. Plotted over time, the curve is close to vertical. Between 2025 and 2026 the complexity of tasks AI could handle increased roughly sixfold.
His practical translation: an end-to-end accounts receivable process, where AI issues the invoice, validates who needs chasing, cross-checks data between systems and then tells your team who to call, was not possible twelve months ago. It is now.
He also noted that the measurement scale itself has started breaking as newer frontier models arrive, which tells you something about the pace. If your last serious assessment of what AI could do for your business was made in 2025, it is out of date.
2. Most businesses are using AI. Very few are using it for real work
MYOB surveyed more than 1,000 mid-sized businesses across Australia and New Zealand with Oxford Economics for its Autonomous Business Report. Two findings stood out.
The first is a time number. Businesses with between 50 and 100 staff identified at least half a day per employee per week in time savings from AI capability available today. Across a team of ten, that is roughly the equivalent of adding an eleventh person.
The second is a gap. Among businesses assessed as having high AI readiness, 88% were already using AI inside core processes such as accounts payable and accounts receivable. Among low-readiness businesses, the figure was 15%.
MYOB's own team admitted they did not initially believe the 88% figure, because the underlying technology at the time was patchy. Businesses were working around the limitations anyway.
The distance between 88 and 15 is the point. It is not a gap in enthusiasm. Nearly every hand in the room went up when asked who uses an AI assistant daily. It is a gap in whether the business is set up so AI can touch anything that matters.
3. Autonomy arrives in levels, and it runs on trust
The framing device of the session was the automotive industry's autonomy scale, published back in 2014, which runs from level zero (fully manual) to level five (fully autonomous). Vais argued the value of that framework was never the plan. It was that it gave an entire industry a shared way to imagine the steps.
Mapped onto a business:
- Level 0 is spreadsheets and disconnected tools, with a person moving every step forward.
- Level 1 is basic automation. Invoices, reminders. Your ERP keeps the books, your people drive.
- Level 2 is the ERP automating steps. Dunning letters going out on time, cross-sell prompts, but a human still starts each task.
- Level 3 is AI agents processing sales orders on their own, scheduling field work, handling parts of the bank reconciliation including some exceptions.
- Levels 4 and 5 are where your people set the strategy and the guardrails, and the system runs the process.
The part we would underline is what moves a business up the levels. Vais argued it is not more AI, it is accumulated trust, and he made the case with cruise control. The first time you set it, your foot hovered over the brake. The first time lane keeping corrected your steering, you grabbed the wheel. Now you would be annoyed to drive without either. As he put it, "trust is something that is built level by level".
The same thing will happen with anomaly detection in your accounts payable. The first flag impresses you. By the tenth, you have forgotten how you coped without it.
4. Why AI beside your ERP costs more than AI inside it
This was the strongest argument of the morning, and it is one worth testing against whatever AI tooling you are currently evaluating.
For AI to process a sales order, it needs the customer's credit terms, their pricing, live stock availability and your approval limits. Every one of those rules already exists inside your ERP.
If the AI sits inside the platform, it inherits them, and you govern it the same way you govern a staff member. If it sits outside, you have to reproduce those rules externally, which means maintaining two sets of rules that will inevitably drift apart. A customer dealt with by the AI then gets subtly different treatment from one dealt with by a person.
Vais reached for cars again. You can absolutely buy a lane-detection camera for a 1998 Commodore and it will beep at you when you drift. It will never drive the car, because it cannot touch the steering wheel or the brakes.
MYOB set out three things AI needs from an ERP: it has to run securely with your data staying inside the platform and onshore, it has to be built in rather than bolted on, and it has to stay open to whatever model comes next. That last one came with an unusually candid line: "anyone who's making forecasts about AI is typically wrong. Me included." Which is a good reason to avoid locking yourself to any single model or vendor.
5. Agents are now built the way you brief a new starter
The demonstrations made this concrete. A pricing agent for a distributor was shown checking four competitors in real time, comparing against internal cost and margin rules, and returning a recommended price that was reviewed and applied in under a minute.
The reveal was the build. The agent was a single prompt written in plain English, describing the product data to look at, the retailers to check, the pricing logic to follow and how to present a recommendation. MYOB's presenter said it was built in an afternoon by someone with no development background.
We are including the demonstration footage rather than describing it further.
|
|
|
|
|
|
The line that stuck with us came from MYOB's product marketing lead closing the automation demo: the system does the running, and your people do the deciding.
What we are seeing in the market
"The businesses we talk to are well past the novelty stage with AI. Almost everyone has someone using an assistant daily. What almost nobody has yet is AI doing real work inside a core process, and that gap is where all the value sits. In our experience it is rarely an AI problem. It is a data and systems problem wearing an AI costume."
Juanita Potgieter, Group Marketing Manager, Verde Group
Where to start
MYOB closed the session with a workshop rather than a pitch, built around three questions. They are good ones to run with your own leadership team.
- What is the single greatest opportunity you see from an AI-enabled ERP? Better decisions, automated manual work, faster month-end close, one source of truth, forward visibility, automated compliance?
- What might hold you back? Leadership alignment, return on investment, organisational readiness, technology readiness, or simply not knowing where to start?
- What is your genuine best next step? Aligning the leadership team, building the business case, proving value on one use case, or agreeing a staged roadmap?
Our advice on the third question is to pick one process. Not a strategy, one process. Prove it, build the trust, then take the next level.
Avanza Solutions implements MYOB Acumatica for New Zealand businesses. If you want to work through what any of this looks like in your environment, we are happy to have that conversation.
Frequently asked questions
What is an autonomous business? An autonomous business is one where core processes run largely on their own, with people brought in for judgement calls rather than to move each step forward. MYOB uses a scale adapted from vehicle autonomy, running from level zero (fully manual) to level five (fully autonomous).
Why does AI need to run inside your ERP? Because the rules AI needs already live there: credit terms, pricing, stock availability and approval limits. AI running outside the ERP requires those rules to be duplicated and maintained separately, which creates drift between how AI and staff handle the same customer.
What is MYOB Acumatica AI Assistant? AI Assistant sits inside MYOB Acumatica and answers natural-language questions using live ERP data, under the same permissions the user already has. It was demonstrated cross-referencing revenue, outstanding balances, margin and credit limits to flag customer credit risk.
What is AI Automation in MYOB Acumatica? AI Automation allows businesses to build AI agents inside the ERP using plain-English prompts, with no developer resource required. The demonstration showed a competitor pricing agent built in a single afternoon.
How ready are mid-sized businesses for AI? MYOB's research with Oxford Economics found 88% of high-readiness businesses already use AI in core processes, compared with 15% of low-readiness businesses. Businesses of 50 to 100 staff identified at least half a day per employee per week in available time savings.
