ERP vs APS: Why Your ERP Was Never Built to Plan

Walk into the planning team of almost any mid-sized or large Australian business and you will find the same thing. There is a multi-million-dollar ERP running the company, whether it is SAP, Oracle, or Microsoft Dynamics, and the people responsible for planning are quietly doing the actual work in Excel. The forecast lives in a spreadsheet. The safety stock calculation lives in another. The replenishment plan gets exported, massaged, and re-keyed back in. The ERP, for all its cost and reach, has become the place decisions get entered rather than the place they get made. That gap is the whole substance of the ERP vs APS question, and it is the reason advanced planning systems exist.

The question matters because the answer is rarely obvious from the inside. An ERP is a serious, expensive, business-critical system, and it is natural to assume that something so central must be capable of planning the supply chain it runs. It usually is not, at least not well, and the confusion between what an ERP does and what an advanced planning system, or APS, does costs Australian businesses a great deal in working capital, service, and wasted planner time. This article sets out the real difference between the two, which ERP planning modules businesses most commonly outgrow, what an APS genuinely adds, and, just as importantly, when your ERP is actually enough and you should not spend a dollar on anything more.

Why this question is landing on more desks right now

There is a practical reason the ERP vs APS conversation has become more common over the past two years, and it is worth naming because it changes the stakes. A large wave of ERP modernisation is underway, driven hardest by SAP. SAP has announced the end of mainstream support for ECC by December 31, 2027, after which no more security patches, compliance updates, or fixes will be delivered. Extended maintenance is available until 2030, at a significant cost premium. Yet the migration is running slowly. A Gartner report showed that in seven years licences for SAP S/4HANA have only been purchased by a third of SAP ECC customers, and 18 to 36 months is typical for a migration, longer in complex environments. Kellton + 3

This matters for planning because an ERP migration is the moment the planning gap becomes impossible to ignore. When a business re-platforms its core system, someone finally asks what the new ERP will do for demand forecasting, inventory optimisation, and replenishment, and discovers that the answer is "not much that it did not do before." The migration surfaces a decision that was always there but easy to defer: whether to keep planning in spreadsheets bolted onto a transactional system, or to put a purpose-built planning layer in place. Getting that decision right at the point of an ERP refresh is far cheaper than retrofitting it later, which is why so many Australian businesses are confronting ERP vs APS now rather than in a few years' time.

The cost of getting it wrong is not abstract. For most businesses, carrying costs typically range between 20% and 30% of total average inventory value per year, according to the Institute for Supply Management. Planning quality is the single biggest lever on how much of the wrong stock you hold, and a transactional system that cannot optimise leaves most of that lever untouched. Fishbowl Inventory

The core distinction: a system of record versus a system of decision

The cleanest way to understand ERP vs APS is to recognise that they are built to do fundamentally different jobs, and being excellent at one tells you almost nothing about being good at the other.

An ERP, or enterprise resource planning system, is a system of record. Its job is to transact and to hold the single, authoritative version of what is true: the orders, the inventory positions, the financials, the bills of material, the supplier records. It executes processes reliably at scale, it keeps the data consistent, and it is the backbone every other system relies on. SAP, Oracle, Microsoft Dynamics, and the rest are extraordinarily good at this. When you place a purchase order, receive stock, or close the month, the ERP is doing exactly what it was designed to do.

An advanced planning system is a system of decision. Its job is not to record what happened but to work out what should happen next: how much to forecast, how much buffer to hold and where, when and how much to reorder, how to sequence production, and where to position inventory across the network. It does this through optimisation, heuristics, and increasingly machine learning, generating plans, simulating trade-offs, and recommending the decision that best balances cost and service against the constraints of the business. Where the ERP answers "what is true," the APS answers "what is the best thing to do about it."

That distinction is the whole game, and it explains the most common misunderstanding in the market. The ERP planning modules disappoint not because they are badly built, but because they were designed to execute a plan, not to work out what the plan should be. Asking your ERP to optimise your supply chain is like asking your general ledger to set your pricing strategy. It holds the numbers faithfully. It was never meant to decide them.

Why the ERP planning module disappoints

Most ERPs ship with planning functionality, usually centred on material requirements planning, and businesses understandably assume this covers them. In practice the limitations show up quickly, and they are structural rather than something a better configuration will fix.

The foundation of ERP planning is MRP, which works backwards from a production or demand plan to determine what materials are needed and when. MRP is genuinely useful and it has been around for decades, but on its own it assumes infinite capacity. It will happily generate a plan that requires more output than your factory or your suppliers can actually deliver, because checking feasibility against real constraints is not what it does. The result is a schedule that looks tidy on screen and falls apart in the real world, which is why planners learn not to trust it and revert to manual workarounds.

The deeper issue is that ERP planning is rules-based, not optimisation-based. It applies the parameters you give it: a fixed lead time in a field, a reorder point someone set two years ago, a flat number of weeks of cover across the catalogue. It does not search across thousands of possible plans to find the one that minimises total cost at your target service level, because it has no optimisation engine to do so. This is the difference between a calculator and a strategist. The ERP will faithfully compute the consequences of the policy you entered. It will not tell you that the policy is wrong, or find you a better one.

Three specific gaps follow from this, and they are exactly where the money sits. ERP planning treats lead time as a single fixed number rather than the variable, shifting distribution it actually is, which forces businesses to either guess high and carry excess or guess low and run short. It sets safety stock one location at a time, with no ability to optimise inventory across a multi-tier network, so every warehouse buffers against the same uncertainty and the business pays for the same protection several times over. And it produces a single forecast number rather than a probabilistic range, which is far less honest about uncertainty and far less useful for setting buffers intelligently. None of these are configuration problems. They are the boundaries of what a transactional system was built to do.

This is the same reason spreadsheets eventually fail, and it is no coincidence that businesses outgrowing their ERP planning are usually drowning in spreadsheets at the same time. The ERP cannot do the sophisticated planning maths, so planners export the data and attempt it manually, which does not scale, breaks when the person who built the model leaves, and produces the familiar situation where two reports disagree and nobody can explain why. The spreadsheet is not the disease. It is the symptom of an ERP being asked to do a job it was never designed for.

Which ERP planning systems businesses outgrow

The ERP vs APS question looks slightly different depending on which platform you run, but the pattern is the same across all of them: the native planning is MRP and reorder-point logic, supplemented heavily by spreadsheets, and it is that combination businesses outgrow. These are the systems we most often see prospects planning in before they move to an APS.

SAP (ECC and S/4HANA). The core planning is MRP and production planning. Larger SAP sites often added SAP APO, the Advanced Planner and Optimizer, which was SAP's APS application, the first version of which launched in 1998. APO is now being retired, and how SAP has handled that retirement actually proves the point of this article. APO's Demand Planning and Supply Network Planning are now covered by SAP IBP, while Production Planning and Detailed Scheduling and Global Available-to-Promise live on as embedded functionalities in SAP S/4HANA. SAP mainstream maintenance for APO ends in 2027. The successor, SAP IBP, introduced in 2015, is a cloud planning suite and is itself an APS. In other words, SAP's own roadmap keeps the planning layer separate from the ERP core, which is exactly the distinction we are describing. SAP shops weighing this are really choosing between adopting IBP and going best-of-breed with a platform such as Kinaxis, o9, or Blue Yonder. Implement + 3

Oracle (E-Business Suite, JD Edwards, and Fusion Cloud SCM). Oracle's ERPs carry native planning that is commonly outgrown once multi-echelon optimisation, demand sensing, or genuine scenario modelling are needed. Oracle offers its own supply chain planning cloud, and many Oracle customers also evaluate independent APS platforms alongside it.

Microsoft Dynamics 365 (Finance and Supply Chain Management, formerly AX). Dynamics uses MRP-based master planning, which is adequate for simpler operations and very commonly outgrown as range, network complexity, and demand volatility grow. It is one of the platforms we most often see prospects supplementing heavily with spreadsheets.

NetSuite. Oracle's cloud ERP is ubiquitous among growing Australian businesses, and its demand planning is deliberately light. It is usually the first system a scaling business outgrows on the planning side, well before it outgrows the ERP itself.

Pronto Xi. A long-standing Australian ERP common in distribution, manufacturing, and retail, with reorder-point planning that is typically run alongside extensive Excel models.

TechnologyOne. The major Australian ERP across government, local councils, universities, and health, where its supply chain and planning capability is thin and planning generally lives outside the system entirely. For public-sector and asset-intensive organisations this is a frequent starting point.

Infor, Epicor, Sage, QAD, IFS, and MYOB Advance. The mid-market field, used widely across Australian manufacturing and distribution. Each carries some planning functionality, and all of it is commonly outgrown for the same structural reasons set out above.

What businesses move to splits into two groups. The first is best-of-breed advanced planning platforms such as Kinaxis, o9, Blue Yonder, RELEX, ToolsGroup, OMP, Logility, and GAINS, which handle optimisation, multi-echelon inventory, and demand sensing at scale. The second, at the smaller and more inventory-focused end, is lighter tools such as Netstock and Slim4. The right destination depends entirely on the size and complexity of the supply chain and the maturity of the planning process behind it.

What an APS actually does that your ERP cannot

An advanced planning system sits above the ERP, drawing on its data as the single source of truth and feeding decisions back into it for execution. It is not a replacement for the ERP and it does not compete with it. It does the planning the ERP was never built to do, and the modern generation of these platforms has moved well beyond the rules-based calculation that ERP planning offers.

The defining capability is optimisation. An APS can search across an enormous number of possible plans and find the one that best balances service and cost against the real constraints of capacity, minimum order quantities, shelf life, and supplier limits. This is the step-change, because it shifts planning from "compute the result of my assumptions" to "find me the best decision." On top of optimisation, the leading platforms layer heuristics and machine learning to handle problems that are too large or too uncertain for classical methods, and to incorporate the variables, such as promotions, pricing, and weather, that statistical forecasting alone struggles with.

Several specific capabilities follow, and they map directly onto the ERP's gaps. An APS performs multi-echelon inventory optimisation, looking at the whole network at once and deciding where it is cheapest and most effective to hold buffer, which typically delivers the same or better service from materially less stock. It supports demand sensing, reading short-term signals to adjust the near-term forecast so the plan reflects what is happening this week rather than what a model assumed last quarter. The more capable platforms can predict lead times rather than assuming them, turning the same machine learning techniques used in demand forecasting onto the inbound side. And critically, an APS enables scenario modelling and exception-based planning, so the business can test "what if" before committing, and planners can spend their time on the genuine exceptions rather than re-keying thousands of routine orders that a system should generate automatically.

The combined effect is significant and measurable. On Trace's own planning and operations work, we typically see forecast accuracy improvements in the range of 20 to 40 per cent where businesses move from spreadsheet-driven planning to a structured process supported by an advanced planning system, and inventory carrying cost reductions of up to 30 per cent off the back of better demand and inventory planning. Those two outcomes are linked, because the inventory reduction is largely a consequence of the accuracy and optimisation gain. We explore the underlying mechanics of this in our piece on how advanced planning systems transform supply chain planning and on demand, inventory, and replenishment as a source of competitive advantage.

The honest part: when your ERP is actually enough

Not every business needs an APS, and one of the more useful things a genuinely independent adviser can tell you is when the answer is no. The ERP vs APS decision is not a question of sophistication for its own sake, and buying planning software you do not need is its own kind of waste.

For a smaller business with a simple supply chain, a limited range, stable demand, and short, reliable lead times, the planning functionality in a modern ERP such as Dynamics 365 or NetSuite, supplemented by some well-built spreadsheets, can be perfectly adequate. The maths that defeats an ERP, multi-echelon optimisation across thousands of items, probabilistic forecasting, network design, only becomes necessary when the scale and complexity of the supply chain make manual approaches genuinely unworkable. If your business is not at that point, an APS is an expensive answer to a question you do not have.

There is a more important caution underneath this. An APS is not a fix for a broken planning operating model, and the businesses that get the least from these platforms are usually the ones that bought the technology hoping it would supply the discipline their organisation lacked. A planning system amplifies the quality of the process it sits on top of. Implement it on clean data, a sound segmentation, a working sales and operations process, and clear accountability, and it is transformative. Implement it on the same broken process that produced the spreadsheet chaos, and all you get is faster, more expensive chaos. If your planning problems are really process and ownership problems, no amount of ERP-versus-APS deliberation will solve them, and the right first move is to fix the operating model, a theme we return to constantly, including in our look at why S&OP so often fails in Australia.

The same logic applies to the current excitement about autonomous, agentic planning. Layering intelligent automation onto an immature data and process base produces confident, fast, wrong decisions, which is why we take a deliberately pragmatic view in our article on agentic AI in the supply chain. The sequence that works, every time, is people, then process, then technology.

The symptoms that you have outgrown ERP planning

If you are weighing ERP vs APS, the most reliable signal is not the size of your business but the behaviour of your planning function. A few patterns reliably indicate that your supply chain has outgrown what SAP, Oracle, Dynamics, or any other ERP can do for planning:

• Your planners spend most of their time maintaining spreadsheets rather than planning, exporting data out of the ERP, working it manually, and keying decisions back in.
• You set safety stock and reorder points by rule of thumb, typically a flat number of weeks of cover, because the ERP cannot calculate buffers properly against demand and lead time variability.
• You cannot model a change before you make it, so questions like "what happens to service if we consolidate two DCs" get answered by instinct rather than analysis.
• Two reports disagree and nobody can fully explain why, because several people maintain their own versions of the plan and there is no single optimised source.
• Your forecast is a single number that is regularly wrong, with no sense of the range of likely outcomes and no systematic way to improve it. If forecasting is your weak point specifically, our guide to improving demand forecasting accuracy is the place to start.
• You are carrying too much of the wrong stock and too little of the right stock at the same time, expediting freight in one category while writing down ageing inventory in another.

One or two of these can often be addressed with better process. When most of them are true at once, you have reached the limit of ERP planning, and an APS is worth serious evaluation, provided the process and data foundations are sound enough to build on.

Getting the sequence right, especially during an ERP migration

For the many Australian businesses currently migrating their ERP, an SAP move to S/4HANA in particular, there is a real opportunity and a real trap. The opportunity is that a re-platform is the natural moment to design the planning layer deliberately rather than inheriting whatever the new ERP happens to offer. The trap is assuming the new ERP will close the planning gap on its own. It will be a better, more modern system of record. It will still not be a system of decision, because that is not what an ERP is for, regardless of vintage. As SAP's own split of APO into IBP and embedded S/4HANA functionality shows, even the vendor treats planning as a separate layer.

The sensible approach is to treat the ERP and the planning capability as two related but distinct decisions, and to get the foundations right first. Establish clean master data, the item attributes, lead times, costs, and supplier information that any planning system runs on, because poor data produces wrong plans no matter how good the optimisation engine is, and data readiness is a workstream in its own right rather than a box to tick. Get the demand forecast trusted and separated from the sales target. Set inventory policy deliberately against a service target. Stand up a planning and sales-and-operations process that actually makes decisions. Do that, and you will capture much of the available value before you have spent anything on an APS, and you will be in a far stronger position to choose and implement one if you decide you need it. Our broader perspective on this sits across our planning and operations and technology capabilities, and the full picture is set out in our guide to supply chain planning for Australia.

How Trace Consultants can help

At Trace Consultants, we help Australian and New Zealand businesses navigate exactly this decision, and we do it without a platform to sell, which means our advice is about what your business actually needs rather than what we are trying to move. Our practitioners have built and run planning processes inside businesses and have selected and implemented planning systems across retail, FMCG, and manufacturing, so we know that the technology is the easy part and the operating model is where the value is won or lost.

We tell you honestly whether you need an APS at all. We assess where your planning capability genuinely sits today, across forecasting, inventory, lead time, and S&OP, and give you a clear-eyed view of whether your ERP and a tighter process would serve you, or whether the scale and complexity of your supply chain genuinely warrant an advanced planning system. No vendor incentive, just your numbers.

We fix the process and data before the technology. Through our planning and operations work, we get the demand forecast trusted, the inventory policy set deliberately, the master data clean enough to plan from, and the S&OP process making real decisions, so that whatever technology follows amplifies a process that already works rather than accelerating one that does not.

We select and implement the right platform, properly. When an APS is justified, we help you choose the platform that fits your situation, whether that is SAP IBP alongside an S/4HANA programme or an independent platform, and implement it on a sound foundation. We bring genuine implementation experience, including with platforms such as GAINS, and a structured methodology that treats the rollout as a capability change rather than a software install. More on our approach sits on our technology page.

We connect planning to the structural decisions. Where you hold inventory is both a planning decision and a network decision, so our strategy and network design work makes sure the two are designed together rather than in separate rooms, and our sector depth across FMCG and manufacturing and retail means the approach fits how your business actually operates.

Explore our Planning and Operations services →

Speak to an expert at Trace →

Where to begin

If you are weighing ERP versus APS, resist the urge to start with a vendor demonstration. Start by being honest about where your planning capability actually is and what is really causing the pain. A short diagnostic across forecasting, inventory, lead time, and your S&OP process will usually reveal quickly whether your problem is a technology gap that an APS would close or a process and data gap that no software will fix. For most businesses the value is not evenly spread, and two or three disciplines are quietly costing the majority of the lost value.

From there, the sequence is the same whether or not you end up buying a platform. Fix the data discipline and the process first. Get the forecast trusted and separated from the target. Set inventory policy against a service level rather than habit. Stand up a planning process that makes decisions. If, having done that, the scale and complexity of your supply chain still defeat what your ERP can do, then evaluate an APS from a position of strength, with a clean foundation for it to optimise. That is the difference between a planning system that transforms the business and one that becomes an expensive disappointment.

Your ERP is the system that runs your business, and a good one is worth every dollar. It was simply never built to plan your supply chain. Knowing the difference, and acting on it in the right order, is what separates the businesses that take cost out and free up capital from the ones that keep firefighting the symptoms with the most expensive tools money can buy.

Frequently asked questions about ERP vs APS

What is the difference between an ERP and an APS?

An ERP, or enterprise resource planning system, is a system of record. It transacts and holds the authoritative version of your orders, inventory, financials, and master data, and it executes processes reliably at scale. An advanced planning system, or APS, is a system of decision. It works out what should happen next, the forecast, the inventory buffers, the replenishment, the production sequence, using optimisation, heuristics, and machine learning. The ERP answers what is true. The APS answers what is the best thing to do about it.

Can SAP, Oracle, or Microsoft Dynamics handle supply chain planning?

They can all do basic planning, usually through MRP, but they were designed to execute a plan rather than to work out what the plan should be. ERP planning is rules-based rather than optimisation-based, assumes infinite capacity, treats lead time as a fixed number, sets safety stock one location at a time, and produces a single forecast rather than a probabilistic range. For a small, simple supply chain that can be enough. For a complex one it leaves most of the value on the table, which is why businesses on SAP, Oracle, and Dynamics so often end up doing the real planning in spreadsheets.

What happens to SAP APO, and do I need SAP IBP or a third-party APS?

SAP APO is being retired, with mainstream maintenance ending in 2027. Its demand and supply network planning move to SAP IBP, and its detailed scheduling embeds into S/4HANA. SAP IBP is a cloud APS in its own right, so businesses face a genuine choice between adopting IBP and selecting an independent platform such as Kinaxis, o9, or Blue Yonder. The right answer depends on your wider SAP roadmap, your planning complexity, and the maturity of your planning process, and it should follow a clear-eyed assessment rather than a default to the incumbent.

Do I need an advanced planning system?

Not necessarily. If your supply chain is small and simple, with stable demand and reliable lead times, your ERP plus good spreadsheets may be adequate. You should consider an APS when the scale and complexity of your supply chain make manual planning genuinely unworkable, when multi-echelon optimisation and probabilistic forecasting would deliver real value, and, crucially, only once your planning process and data are sound enough to build on. An APS amplifies a good operating model and accelerates a bad one.

Does an APS replace my ERP?

No. An APS sits above the ERP, draws on its data as the single source of truth, and feeds planning decisions back into it for execution. The two systems do different jobs and work together. You keep the ERP as your system of record and add the APS as your system of decision.

Should I sort out planning during an ERP migration?

Yes, an ERP migration such as a move to S/4HANA is the natural moment to design your planning layer deliberately, but do not assume the new ERP will close the planning gap on its own. A newer ERP is still a system of record, not a system of decision. Treat the ERP and the planning capability as related but distinct decisions, fix the data and process foundations first, and then decide whether an APS is justified.

Related reading: Supply Chain Planning: A Guide for Australia · How Advanced Planning Systems Transform Supply Chain Planning · How to Improve Demand Forecasting Accuracy · S&OP That Actually Works in Australia

Supply Chain Planning: The Disciplines That Decide Cost and Service

Most Australian businesses are carrying too much of the wrong stock and not enough of the right stock, often at the same time. They are expediting freight to cover shortages in one category while writing down ageing inventory in another. They are running promotions that the supply chain never saw coming, and holding safety stock against a demand pattern that stopped being relevant two years ago. None of this is a warehousing problem or a procurement problem. It is a planning problem, and it is the single most under-managed part of the supply chain in this country.

Supply chain planning is the set of disciplines that decide what you make or buy, how much, when, and where you hold it. Get it right and you free up working capital, lift service levels, and take cost out without anyone having to work harder. Get it wrong and every other function in the business spends its time firefighting the consequences. The frustrating part is that planning rarely gets the attention it deserves, precisely because it sits upstream of the problems it causes. The stockout looks like a supplier issue. The excess looks like a demand issue. The expedite looks like a logistics issue. Underneath all three is a forecast that was never going to hold and an inventory policy that nobody had reviewed in eighteen months.

This guide walks through the core planning disciplines as they actually work: demand forecasting, inventory optimisation, lead time management, supply and replenishment planning, production planning, network design, and the sales and operations process that ties them together. It is written for the people who own the outcomes, the CFOs watching working capital, the COOs answering for service, and the supply chain and operations leaders who have to make the numbers reconcile. The tools have changed enormously over the past decade, and we will come to where advanced planning systems and platforms such as GAINS fit. But the disciplines underneath are timeless, and most of the value is still left on the table by businesses that have never properly mastered them.

What Supply Chain Planning Actually Is

Supply chain planning is the process of matching supply to demand across time, at the lowest total cost and the service level the business has committed to. That definition sounds simple, and the trap is in the words "across time" and "total cost." Planning is not a single decision. It is a connected sequence of decisions made over different horizons, from a five year network strategy down to tonight's replenishment order, and the quality of the whole chain depends on how well those horizons talk to each other.

It helps to think of planning as a set of distinct but interlocking disciplines. Demand planning estimates what customers will want. Inventory planning decides how much buffer to hold against the uncertainty in that estimate. Supply and replenishment planning turns the plan into purchase orders, production orders, and stock movements. Production planning sequences the factory. Network design decides where the inventory and capacity should physically sit. And sales and operations planning, often shortened to S&OP, is the management process that aligns all of it with the commercial and financial plan of the business.

Each of these can be done well in isolation and still produce a poor result overall, because the disciplines are dependent on one another. A brilliant forecast is wasted if the inventory policy ignores it. A sophisticated inventory model is undermined if lead times are unstable and nobody is managing them. A perfectly optimised network is irrelevant if the S&OP process never surfaces the demand signal that should have triggered a change. This is why the businesses that win at planning treat it as a system, not a collection of tasks owned by different people who meet once a month and disagree.

The other thing worth saying early is that planning is a probability game, not a precision game. You will never know exactly what demand will be. The goal is not a perfect forecast. The goal is to make good decisions in the face of uncertainty, which means understanding the uncertainty rather than pretending it away. Almost every planning failure we see traces back to a business that treated a single forecast number as a fact, built its entire supply chain on that number, and then blamed the number when reality arrived.

Demand Forecasting and Demand Sensing

Forecasting is the foundation, because every downstream decision inherits its assumptions. If the forecast is biased high, you will carry excess everywhere. If it is biased low, you will chase shortages everywhere. And if it is volatile and nobody trusts it, the business will quietly route around it, with sales building their own spreadsheet, operations building another, and finance using a third. One business, three versions of the future, none of them owned.

Traditional forecasting leans on statistical methods that extrapolate from history: moving averages, exponential smoothing, and seasonal decomposition. These are not obsolete, and for stable, high-volume items they are often perfectly adequate. The mistake is applying them to everything. A slow-moving spare part, a fashion item with a twelve week life, and a staple grocery line do not behave the same way, and they should not be forecast the same way. Good demand planning segments the portfolio and applies the right method to each segment, rather than running one model across the catalogue and accepting whatever falls out.

The bigger shift over the past decade has been the move from purely statistical forecasting to probabilistic and machine learning approaches, and then to demand sensing. Probabilistic forecasting produces a range of likely outcomes with their associated probabilities, rather than a single line, which is far more honest about the uncertainty and far more useful for setting inventory. Machine learning models can incorporate variables that classical methods struggle with: promotional calendars, pricing changes, weather, public holidays, even leading indicators from outside the business. Demand sensing takes this further by reading short-term signals such as recent orders and point-of-sale data to adjust the near-term forecast quickly, so the plan reflects what is happening this week rather than what a model assumed last quarter.

The practical lesson for Australian businesses is to be clear about what problem you are solving before reaching for the most advanced technique. Forecast accuracy is not free, and chasing the last few percentage points of accuracy on an item that is cheap and fast to replenish is wasted effort. The value of a better forecast is highest where inventory is expensive, lead times are long, or the cost of a stockout is severe. Concentrate the sophistication there. A few rules of thumb hold up well: forecast at the level you can actually plan at, measure forecast bias as seriously as forecast error because persistent bias is what quietly inflates or starves inventory, and never let the forecast become a negotiation. The forecast is a best estimate of demand. It is not the sales target, the budget, or the stretch goal, and the moment those things get blended into it, the supply chain starts making decisions on a number that was designed to motivate people rather than describe reality.

Inventory Optimisation and Why Multi-Echelon Changes the Game

Inventory is where planning decisions show up on the balance sheet, and it is where the cost of poor planning is most visible. The Institute for Supply Management puts the cost of holding inventory at roughly 20 to 30 per cent of its value every year once you account for the capital tied up, storage, insurance, handling, shrinkage, and obsolescence. The largest single component of that is usually the cost of capital, the return the business forgoes by having money locked in stock rather than working elsewhere. With the Reserve Bank cash rate sitting at 4.35 per cent through the middle of 2026 after three increases earlier in the year, and the Bank signalling it is in no hurry to cut, the money tied up in excess inventory has rarely been more expensive in recent memory. Every dollar of stock you do not need is a dollar you are paying to hold, at a higher rate than you were two years ago.

Inventory optimisation is the discipline of holding the least stock that still delivers the service you have promised. The classical tools are safety stock calculations and reorder points, which set a buffer based on demand variability, lead time, and a target service level. Done properly, this is already a step up from the rule-of-thumb approach most businesses default to, which is to hold a fixed number of weeks of cover across the board. A flat weeks-of-cover policy guarantees that you are over-invested in your stable, predictable lines and under-invested in your volatile, hard-to-forecast ones, which is exactly backwards. Variability, not volume, should drive the buffer.

The most significant advance in inventory planning is multi-echelon inventory optimisation, usually shortened to MEIO. Most businesses still set safety stock one location at a time, as if each warehouse or store existed on its own. In reality, inventory sits in a network: suppliers feed national distribution centres, which feed regional centres, which feed stores or branches. When you optimise each echelon independently, you systematically hold too much, because every node is buffering against the same demand uncertainty without any credit for the buffers held elsewhere. Multi-echelon optimisation looks at the whole network at once and decides where in the chain it is cheapest and most effective to hold the buffer, often pooling stock upstream where it can serve multiple downstream locations rather than pre-positioning it everywhere.

The result, for businesses that make the move properly, is the same or better service from materially less inventory, because the network is no longer paying for the same protection several times over. This is precisely the kind of problem that spreadsheets cannot solve, because the mathematics of optimising service and cost across a multi-tier network with thousands of items is beyond what a manual model can handle. It is also one of the clearest cases where an advanced planning platform earns its keep, and a capability that systems such as GAINS are specifically built around.

A final point on inventory that is easy to lose. The objective is not minimum inventory, it is optimum inventory. It is entirely possible to cut stock too far, and a business that congratulates itself on a lower inventory number while its service quietly erodes and its expedite costs climb has not optimised anything, it has just moved the cost somewhere less visible. The discipline is to set the buffer deliberately, against a defined service target, with full sight of the trade-off, rather than swinging between too much and too little depending on which complaint was loudest last quarter.

Lead Time, the Variable Everyone Ignores

Ask most planning teams what drives their inventory and they will point to demand. Demand matters, but the variable that quietly does as much damage, and gets a fraction of the attention, is lead time. Specifically, lead time variability. Safety stock exists to protect against two things: uncertainty in demand and uncertainty in supply. A supplier whose lead time swings between three and nine weeks forces you to hold buffer for the nine week case, even if the average is five, because you cannot afford to run out during the long ones. Stabilise that lead time and the required buffer falls, with no change to demand at all.

The problem is that most businesses treat lead time as a fixed parameter, a single number sitting in a field in the ERP, set once and rarely revisited. It is almost never a single number. It is a distribution, and often a wide and shifting one, affected by supplier reliability, port and freight conditions, customs and biosecurity processing, and the supplier's own production schedule. Planning against the average lead time when the real distribution is wide is a guaranteed way to be short during the long tail, which is exactly when shortages hurt most.

This is where one of the more useful recent advances comes in: predicting lead times rather than assuming them. The same machine learning techniques that improved demand forecasting can be turned on the inbound side, using historical performance, supplier behaviour, and external signals to forecast how long replenishment will actually take, and how much that time is likely to vary. It is a discipline very few Australian businesses have built, which makes it both an opportunity and a genuine source of competitive advantage, and it happens to be one of the areas where the more capable planning platforms have invested heavily. Treating lead time as something you forecast and manage, rather than a constant you inherit, is one of the highest-return changes a planning function can make.

The management lesson is straightforward. Measure your lead times the way you measure demand: track the actuals, understand the variability, and feed that variability into your inventory policy rather than a static assumption. And manage lead time as a lever, not a given. Supplier reliability is negotiable. Order frequency, mode of transport, and how far ahead you commit are all choices that change the lead time distribution, and therefore the inventory you are forced to carry. The buffer in your warehouse is often a direct, measurable price you are paying for instability further up the chain, and that price is usually larger than anyone has bothered to calculate.

Supply and Replenishment Planning

If demand and inventory planning decide what good looks like, supply and replenishment planning is where the plan becomes action. This is the discipline that converts the forecast and the inventory targets into actual purchase orders, production orders, and stock transfers, in the right quantities and at the right times, while respecting the real constraints of the business. It is the least glamorous part of planning and arguably the part where execution quality matters most, because a sound strategy ruined by sloppy replenishment delivers the same empty shelf as having no strategy at all.

The core of replenishment is deciding when to order and how much. Reorder point logic triggers an order when stock falls to a calculated level. Periodic review orders on a fixed cycle up to a target level. Each suits different situations, and mature planning blends them across the portfolio rather than forcing everything into one model. Layered on top are the real-world constraints that textbooks gloss over: minimum order quantities, supplier batch sizes, container and pallet rounding, shelf-life limits, and supplier capacity ceilings. A replenishment plan that ignores these produces orders that cannot actually be placed, which is how planners lose faith in the system and quietly go back to ordering by feel.

Two failure modes are worth naming because they are so common. The first is the bullwhip effect, where small swings in customer demand amplify into large swings in orders as they move up the chain, because each tier reacts to the orders below it rather than to true end demand, adding its own buffer and its own batching as it goes. The classic cause is each link planning off the orders it receives instead of the underlying demand signal, and the classic fix is visibility, getting true demand information further up the chain so that upstream tiers are not amplifying noise. The second failure mode is allocation under constraint. When supply is short, someone has to decide who gets what, and if that decision is made ad hoc by whoever shouts loudest, the business ends up serving its least valuable demand and starving its most valuable. Good supply planning has an allocation logic decided in advance, based on customer priority, margin, and strategic importance, so the scarce stock goes where it does the most good.

Replenishment is also the discipline that benefits most directly from automation, and the one where automation is most safely applied. The decisions are high volume, rules-based, and repetitive, which is exactly the profile a planning system handles better than a person. The aim is not to remove planners. It is to automate the thousands of routine reorder decisions so the planners can spend their time on the exceptions, the new products, and the constrained situations where judgement actually adds value, rather than rekeying orders that a system could have generated and that a person, working at speed across a large catalogue, will inevitably get wrong some of the time.

Production and Fabrication Planning

For manufacturers and fabricators, planning extends into the factory, and production planning is its own discipline with its own hard constraints. Where a distributor's main lever is when and how much to buy, a manufacturer also has to decide what to make, in what sequence, on which line or machine, with which materials, and in what quantity, all while juggling finite capacity, changeover times, and material availability. The complexity is higher, and the cost of getting it wrong is more immediate, because idle capacity and the wrong production sequence translate into cost and missed orders within days.

The foundation is material requirements planning, or MRP, which works backwards from the production schedule to determine what components and raw materials are needed and when, so that materials arrive in time to build but not so early that they clog the floor and tie up cash. MRP has been around for decades and is built into every ERP, but on its own it assumes infinite capacity, which is why so many MRP-driven schedules are quietly infeasible. Finite-capacity scheduling adds the missing constraint, sequencing work against the real capacity of machines and people, accounting for the changeover time lost when you switch from one product to another, and producing a schedule the factory can actually execute rather than one that looks tidy on paper and falls apart by Wednesday.

The deeper strategic question in production planning is make-to-stock versus make-to-order, and where to place the decoupling point, the line in the process before which you build to forecast and after which you build to actual orders. Push it too far downstream and you carry finished goods inventory you may never sell. Pull it too far upstream and your lead times to the customer blow out because you are starting from scratch on every order. The right answer depends on demand variability, customer lead time tolerance, and the cost of holding stock at each stage, and getting it right is one of the highest-leverage decisions a manufacturing business makes. For fabrication and engineered-to-order environments specifically, the planning problem becomes one of scheduling shared resources across jobs with very different routings, which is well beyond what spreadsheets or basic MRP can handle and is increasingly the domain of dedicated production planning and optimisation tools.

For Australian manufacturers and fabricators, the prize is significant, because production planning sits at the intersection of cost, capacity utilisation, and customer service. A factory running a better schedule makes more with the same assets, holds less work in progress, and quotes more reliable lead times to its customers, all of which fall through to margin. It is also an area where the planning discipline and the planning technology have to fit the realities of the specific operation, which is why a generic implementation rarely lands and why the value comes from configuring the approach to how the plant actually runs.

Supply Chain Network Design

Network design is the most strategic of the planning disciplines and the one most often treated as a one-off project rather than an ongoing capability. It asks the structural questions: how many distribution centres should we have and where, which customers and stores should each serve, where should we hold inventory across the network, and how should product flow from source to shelf. These decisions shape cost and service for years, because they determine the fixed footprint within which all the day-to-day planning has to operate. You can run a brilliant replenishment plan inside a poorly designed network and still lose, because the structure itself is working against you.

Historically, network design happened once a decade, usually triggered by a lease expiry, a merger, or a crisis, and was run as a discrete strategy engagement that produced a recommendation and then sat on a shelf. That cadence no longer fits the pace at which demand patterns, costs, and customer expectations now move. Fuel and freight costs shift, e-commerce changes where demand physically lands, property and labour markets tighten, and a network that was optimal three years ago can quietly become a liability. The better approach treats network design as a living capability, with the model maintained and rerun as conditions change, so the business can answer "should we change the footprint" with analysis rather than instinct whenever the question arises.

The technique at the heart of network design is optimisation, building a mathematical model of the network with its costs, capacities, and service requirements, and solving for the structure that minimises total cost at the required service level. Done well, this surfaces trade-offs that intuition misses, such as the fact that adding a distribution centre can sometimes reduce total cost by cutting transport and improving service even though it adds fixed overhead, or that consolidating to fewer, larger sites can lower cost while lengthening delivery times in a way some customer segments will tolerate and others will not. The value is in quantifying those trade-offs so the decision is made with eyes open. Scenario modelling matters as much as the base optimisation: a robust network is one that performs well across a range of plausible futures, not one that is perfectly tuned to a single forecast that will not eventuate.

The point that ties network design back to the rest of planning is that structure and flow are not separable. Where you decide to hold inventory in the network is both a design decision and an inventory decision, and the multi-echelon optimisation discussed earlier is really the operational expression of a network design choice. Businesses that keep these conversations in separate rooms, with strategy designing the network and planning running the inventory, leave value on the table at the seam between them. The two disciplines answer different time horizons of the same question, and they should share the same model and the same assumptions.

Sales and Operations Planning, the Integrating Layer

Everything above describes individual planning disciplines. Sales and operations planning is the management process that pulls them into a single, agreed plan and aligns that plan with the commercial and financial direction of the business. Without it, you have a set of technically competent functions optimising in different directions: sales chasing revenue, operations chasing efficiency, finance chasing the budget, and supply chain caught in the middle trying to serve all three with one set of stock. S&OP, done properly, forces those tensions into the open once a month and resolves them with one plan that everyone has signed up to.

A healthy S&OP cycle moves through a recognisable rhythm. It starts with an unconstrained view of demand, the honest forecast, separated from the sales target. It then tests that demand against supply: can we make it, buy it, store it, and move it, given our capacity and constraints. Where demand and supply do not reconcile, the gaps and the choices are escalated to a leadership review where decisions get made, including the commercial ones such as whether to invest in capacity, accept a service trade-off, or shape demand through pricing and promotion. The output is a single approved plan, in units and in dollars, that the whole business runs to until the next cycle. The discipline is in the sequence and in the seniority of the people in the room. An S&OP process that does not connect to the financials, or that the executive does not attend, is a production meeting wearing a strategic title.

This is the area where Australian businesses most consistently underperform, and the reasons are rarely technical. The common failure is an S&OP process that exists on paper but has degenerated into a backward-looking status update, where teams report what happened rather than deciding what to do next, and where the difficult cross-functional trade-offs are politely avoided rather than resolved. Another is an S&OP that never bridges to money, so the plan is expressed only in cases and pallets and the finance team runs a parallel forecast that never reconciles, which means the business is effectively planning twice and trusting neither. A third is the absence of a genuine decision-making forum, so issues are surfaced but never closed, and the same problems reappear cycle after cycle.

The fix is rarely more software and almost always better process and clearer ownership. S&OP is where the planning disciplines stop being a technical exercise and become a leadership one, because the trade-offs it surfaces, service against cost, revenue against capital, this customer against that one, are commercial decisions that only the leadership team can make. The most advanced demand and supply models in the world will not help a business that has no forum to act on what they reveal. When organisations ask us to fix their planning, the diagnosis very often lands here, on a process that surfaces the right questions to the wrong people, too late, in a currency the business does not actually run on.

From Spreadsheets to a Planning System: Where Advanced Planning Systems and GAINS Fit

Almost every business starts planning in spreadsheets, and many never leave. Spreadsheets are flexible, familiar, and free, and for a small, simple supply chain they are genuinely fine. The trouble starts as the business grows. Spreadsheet planning does not scale, it breaks when the person who built the model leaves, it cannot handle the mathematics of multi-echelon inventory or network optimisation, it offers no single source of truth when several people maintain their own versions, and it consumes enormous planner time on manual rekeying that adds no value and introduces error. The symptoms are easy to recognise: planners spending their days maintaining the spreadsheet rather than planning, decisions made on stale data, and nobody quite able to explain why the numbers in two reports disagree.

An advanced planning system, or APS, is purpose-built software that handles these disciplines at scale: demand forecasting, inventory optimisation, supply and replenishment planning, production planning, and network design, on a single platform with a single version of the data. It sits above the ERP, which is a system of record built to transact, not to optimise, and is one reason ERP planning modules so often disappoint: they were designed to execute the plan, not to work out what the plan should be. The modern generation of these platforms has moved well beyond rules-based calculation. They combine optimisation, heuristics, and machine learning to generate the plan, simulate the trade-offs, and increasingly automate the routine decisions so planners can focus on the exceptions.

This is the context in which a platform such as GAINS fits, and it is one Trace knows well from implementation experience. GAINS frames its offering around decision engineering and orchestration, combining AI and machine learning, heuristics, and optimisation across supply chain design, planning, forecasting, sales and operations planning, and replenishment on a single platform. It is strong in precisely the areas where most businesses struggle and spreadsheets cannot follow: multi-echelon inventory optimisation, predicting lead times rather than assuming them, demand sensing, and network design. The longevity matters too, as a planning suite with several decades behind it has been through enough real implementations to know that the technology is the easy part. Its own delivery approach reflects that, leaning on a structured, proven implementation methodology rather than treating the rollout as a software install, which is the right instinct, because the way these systems are implemented is what decides whether they deliver.

The honest caution is that an APS is not a silver bullet, and the businesses that get the least from these platforms are usually the ones that bought the technology hoping it would fix a problem that was really about process, data, or ownership. A planning system amplifies the quality of the planning operating model it sits on top of. Implement it on clean data, a sound segmentation, a working S&OP process, and clear accountability, and it is transformative. Implement it on the same broken process that produced the spreadsheet chaos, and you get faster, more expensive chaos. The decision to invest in an APS should follow a clear-eyed view of where the planning capability genuinely is, not a hope that the software will supply the maturity the organisation has not built. For most businesses the sensible sequence is to fix the process and the data discipline first, or in parallel, so the platform has something solid to optimise.

Why Planning Initiatives Fail

After enough of these engagements, the failure patterns become predictable, and they are worth naming because avoiding them is most of the battle. The first is tool-first thinking, buying a planning system before fixing the planning process, on the assumption that the software will impose the discipline the organisation lacks. It will not. It will automate whatever process you point it at, good or bad, and a poor process running faster is not progress.

The second is neglecting data. Planning systems run on master data, item attributes, lead times, costs, supplier information, and bills of material, and most businesses badly underestimate how poor their data is until a system tries to use it. Wrong lead times, missing costs, and stale attributes produce wrong plans no matter how good the optimisation engine is, and the credibility of an entire implementation can be lost in the first month if planners see the system generating obviously wrong answers from obviously wrong inputs. Data readiness is not a preliminary box to tick, it is a workstream in its own right.

The third is chasing forecast accuracy as an end in itself, pouring effort into a marginally better forecast while ignoring the inventory policy, the lead time variability, and the S&OP process that would deliver far more value for far less effort. Accuracy matters where it pays. Elsewhere it is a vanity metric. The fourth is treating planning as a back-office function, staffing it juniorly, giving it no executive voice, and then wondering why the cross-functional trade-offs never get resolved. Planning decisions are commercial decisions, and a planning function without seniority and without a seat at the leadership table cannot make them stick. The fifth, related, is the absence of clear ownership, where demand, supply, and inventory each have several part-owners and therefore no real owner, and accountability dissolves into a monthly meeting where everyone agrees there is a problem and nobody is responsible for fixing it.

The common thread is that planning fails for organisational reasons far more often than technical ones. The mathematics of forecasting and optimisation is well understood and increasingly handled by capable software. What is hard, and what separates the businesses that get real value from the ones that do not, is the operating model around it: the process, the data discipline, the seniority, and the willingness of leadership to make the trade-offs that planning surfaces rather than deferring them. A business that gets the operating model right will do well even with modest tools. A business that gets it wrong will struggle with the best tools money can buy.

How Trace Consultants Can Help

Trace works with Australian businesses across retail, FMCG and manufacturing, health, government, and infrastructure to build planning capability that actually holds, combining the discipline of the methods with the practical reality of how each organisation runs.

Planning and operations capability that lasts. We design and improve the full planning stack, demand forecasting, inventory optimisation including multi-echelon approaches, supply and replenishment planning, and production planning, and we build it into your operating model so it survives after we leave. Our focus is the capability, not a slide deck, which means leaving your people able to run and improve the process themselves. You can read more about our approach on our Planning and Operations page.

Network and inventory strategy. Through our strategy and network design work we model your distribution footprint, flow paths, and where inventory should sit across the network, quantifying the cost and service trade-offs so the structural decisions are made on analysis rather than instinct. Explore our Strategy and Network Design capability for detail.

Advanced planning systems, selected and implemented properly. We help businesses decide whether they need an APS, choose the right platform for their situation, and implement it on a foundation of clean data and a sound process. We bring genuine implementation experience, including with platforms such as GAINS, and we are clear-eyed about what these systems can and cannot do. More on this sits on our Technology page.

Sector depth where it counts. Planning is not generic, and the realities of a retail business differ from those of an FMCG or manufacturing operation. We bring practitioners who have done the work in your sector, so the approach fits how your business actually operates rather than how a textbook assumes it does.

Explore our Planning and Operations services →Speak to an expert at Trace →

Where to Begin

If your planning needs work, resist the urge to start by buying software. Start by understanding where the capability genuinely sits today. A short, honest diagnostic across the disciplines, how you forecast, how you set inventory, how you manage lead times, how you run S&OP, usually reveals quickly where the value is, and it is rarely evenly spread. For most businesses, two or three disciplines are quietly costing the majority of the lost value, and concentrating effort there beats a broad transformation that tries to fix everything at once and finishes nothing.

From there, the sequence that tends to work is to fix the process and the data before, or alongside, any technology, because a planning system implemented on a sound operating model is transformative and the same system implemented on a broken one is an expensive disappointment. Get the demand forecast trusted and separated from the sales target. Set inventory policy deliberately against a service target rather than by habit. Treat lead time as something you measure and manage. And stand up an S&OP process with the right people in the room making real decisions. Do those things and you will capture much of the available value before you have spent a dollar on a platform, and you will be in a far stronger position to choose and implement one if you decide you need it.

The Bottom Line

Supply chain planning is the highest-leverage and most under-managed part of the supply chain in most Australian businesses. The disciplines, demand forecasting, inventory optimisation, lead time management, supply and replenishment planning, production planning, network design, and the S&OP process that ties them together, are well understood, and the tools to support them have never been better. Yet the value keeps going uncaptured, because planning sits upstream of the problems it causes and because doing it well is as much an organisational challenge as a technical one. The businesses that treat planning as a connected system, invest in the operating model as seriously as the technology, and give the function the seniority its decisions warrant, take out cost, free up capital, and lift service in ways that compound year after year. The ones that keep firefighting the symptoms will keep carrying too much of the wrong stock, and paying more than ever to hold it.

If planning is costing your business more than it should, Trace can help you find where the value is and build the capability to capture it.

Explore our Planning and Operations services →Speak to an expert at Trace →

‍

Agentic AI in the Supply Chain: Hype Versus What Actually Works

Agentic AI is the loudest story in supply chain right now. Every software vendor has agents in its roadmap, every conference from CES to Hannover Messe is demonstrating autonomous orchestration, and every strategy deck promises a self-managing supply chain that senses disruption, finds alternatives, and acts, all without human intervention. Some of this is real and arriving faster than many expected. A good deal of it is aspirational, a demonstration of what might be possible rather than what is running in production today. The defining skill for a supply chain leader in 2026 is telling the two apart.

This matters because the cost of getting it wrong runs both ways. Dismiss agentic AI as hype and you cede ground to competitors who are using it to react faster and operate leaner. Believe the hype uncritically and you spend heavily on autonomous capability your data, systems, and processes cannot actually support, and you end up with expensive agents that produce confident, fast, wrong decisions. The pragmatic path runs between those errors, and finding it requires understanding what agentic AI really is, where it works now, where it is still a promise, and what separates the organisations that get value from it from those that do not.

This article is a practitioner's view for supply chain, procurement, and operations leaders who want to cut through the noise and make sensible decisions about agentic AI.

What agentic AI actually is

The term gets used loosely, so it is worth being precise. The progression of AI in supply chains has gone through stages. Traditional analytics describes what happened and, at its best, predicts what will happen. Generative AI, the wave that arrived most recently, answers questions and drafts content, summarising data, writing reports, responding in natural language. Agentic AI is a further step: it does not just inform, it acts. An agentic system is given a goal and the permission to pursue it, and it orchestrates and executes workflows across multiple systems autonomously to achieve that goal, taking actions with limited or no human intervention.

The difference is execution. A generative AI tool might tell a planner that a supplier is at risk and a reorder is advisable. An agentic system monitors the signals continuously, identifies the risk itself, finds an alternative supplier, adjusts the procurement order, and initiates the action across the connected systems, escalating to a human only where its rules require. That shift from advice to autonomous action is the whole point of agentic AI, and it is also the source of both its promise and its risk.

Where it works now

Stripping away the hype, there are areas where agentic AI is delivering real value in 2026, and they share a common trait: they involve high-frequency, data-rich decisions where speed matters and the cost of a wrong move is contained.

In planning and inventory, agents monitor stock levels, sales signals, and demand forecasts continuously across locations, trigger reorders, and redistribute inventory between facilities faster than traditional planning cycles can react. Where demand shifts in real time, this responsiveness is an advantage over periodic planning runs.

In logistics, agentic systems replan routes automatically when disruptions occur, reallocating stops across a fleet without dispatcher involvement, balancing competing objectives like speed, cost, emissions, and service commitments simultaneously, and learning from delivery outcomes to improve future decisions. The result is higher first-attempt delivery rates, better vehicle utilisation, and less manual firefighting.

In procurement, sourcing and supplier-risk agents continuously scan supplier financial health, geopolitical exposure, and ESG indicators, flagging instability before it disrupts supply, and supporting spend analysis and contract review. Given how much of procurement is now risk management and supplier intelligence rather than pure cost control, this is a natural fit.

In disruption response, orchestration agents detect a problem, find alternatives, reroute, and execute contingency plans across interconnected systems, compressing a response that used to take days of human coordination.

And in decision support and data access, agents that translate natural-language questions into queries against supply chain data are removing one of the quieter frictions in the function: the gap between a manager's question and the answer buried in a system nobody has time to interrogate. Digital twins, which let teams simulate and test supply chain changes before committing to them physically, are extending this into scenario testing.

These are real and worth taking seriously. But notice what they have in common: most of the value today comes from agents augmenting and accelerating human decision-making, or autonomously handling well-bounded, lower-stakes, reversible actions. That is the honest current state, and it is rather different from the marketing.

Where it is still hype

The vision being sold, a fully autonomous, lights-out supply chain that runs itself end to end, is largely still aspirational, and it is important to say so plainly.

In practice, the credible deployments are human-plus-agent hybrids. Agents operate as permissioned participants alongside human teams, handling the routine and the time-critical, while humans retain control of the consequential decisions, the ones with large financial, safety, regulatory, or relationship stakes. Autonomy is being applied selectively, not universally, and the most autonomous, end-to-end orchestration remains earlier in maturity than the demonstrations suggest. There is a real gap between a polished vendor demo running on clean sample data and the same capability running reliably on a real organisation's messy systems and exceptions.

This is not a reason to dismiss agentic AI. It is a reason to be precise about what you are buying and what you will actually get in the first year versus the third. Autonomy is a spectrum, not a switch, and most organisations will move along it gradually, expanding what they let agents do as confidence and capability grow.

The foundations that decide success

Here is the part the hype skips, and the part that matters most. Whether agentic AI delivers value or destroys it in a given organisation comes down to foundations that have nothing to do with the cleverness of the agent.

1. Systems integration

An agent that cannot read from and write to your core systems in real time, your ERP, whether SAP, Oracle, Microsoft Dynamics, or another, and the planning, procurement, and logistics platforms around it, cannot actually execute. It can only generate insight, which makes it a more expensive dashboard, not an autonomous actor. Real-time, bidirectional integration is the critical technical dependency, and it is where many agentic ambitions stall.

2. Data

Agents act on data, and they act fast. Fragmented, inconsistent, or unreliable data does not just produce a bad report, as it would with traditional analytics; it produces wrong actions executed automatically at machine speed before a human notices. Spend data scattered across business units in different currencies and taxonomies produces misleading analyses. Inventory data that is not trustworthy produces bad autonomous reorders. The old principle becomes sharper: garbage in, garbage executed.

3. Process clarity

You cannot automate a process you have not defined. Agentic AI amplifies a well-designed process and accelerates a poorly-designed one, and an agent let loose on a broken process simply breaks things faster. This is the same sequencing rule that applies to every supply chain technology, and it is why we consistently argue for getting the process right before layering the tool on top, whether the tool is an advanced planning system or an autonomous agent.

4. Governance

Agents that take autonomous action need defined permissions, clear boundaries on what they can and cannot decide, escalation rules for when to involve a human, and native audit trails, the latter being non-negotiable in regulated industries where every decision must be traceable. Deciding what an agent is allowed to execute on its own versus what requires human sign-off is a governance design exercise that has to happen before deployment, not after the first costly mistake.

5. The human role

This shifts rather than disappears. The work moves from performing the task to designing, supervising, and governing the agents that perform it, handling the exceptions agents cannot, and exercising judgment on the consequential calls. Organisations that imagine agentic AI as a headcount-removal exercise misunderstand it. The people who used to execute become the people who orchestrate and oversee, and that transition has to be managed deliberately.

Get these five foundations right and agentic AI can deliver. Skip them and no amount of agent sophistication will save the investment.

How to approach it pragmatically

A sensible adoption path follows from all of this. Start with the problem, not the technology. Identify where autonomous, high-frequency execution would create value in your supply chain, rather than starting from a desire to deploy agents and hunting for somewhere to put them.

Fix the data and process foundations first, at least for the area you are targeting. This is unglamorous and it is the work that determines the outcome.

Start narrow and well-bounded. Begin with lower-stakes, reversible, high-confidence decisions where an error is cheap and recoverable, prove the value, and build organisational trust before extending autonomy to higher-stakes territory. Keep humans firmly in the loop for consequential decisions, and widen the agent's remit only as confidence is earned.

Build governance from the outset, not as an afterthought once something has gone wrong. And sequence the whole effort in the right order: people, process, and data first, then the agents that act on them. The organisations that treat agentic AI as the last and easiest step on a solid foundation will outperform those that treat it as a shortcut around the foundation. Our broader perspective on supply chain technology and how planning systems actually deliver applies directly here.

The Australian context

For Australian and New Zealand businesses there is a particular trap and a particular opportunity. The trap is that many organisations in this region are still maturing their planning systems, data foundations, and process discipline, which means the temptation to leapfrog straight to autonomous agents, skipping the foundations, is both strong and dangerous. Agentic AI deployed on an immature data and process base will disappoint expensively.

The opportunity is real too. Persistent labour constraints, the geographic complexity and distance that characterise ANZ supply chains, and the constant pressure on cost and service all make intelligent automation attractive. The organisations that build the foundations and then adopt agentic AI deliberately will get a real edge in responsiveness and efficiency. The ones that chase the hype without the substrate will spend money and learn an expensive lesson. Pragmatism, not enthusiasm or scepticism, is the right posture.

How Trace Consultants can help

At Trace Consultants, we take a deliberately pragmatic, technology-agnostic view of agentic AI. We have no platform to sell and no agent to push, which means our advice is about what will actually create value in your supply chain, not what is fashionable.

We assess where agentic AI fits. We identify the decisions and workflows in your supply chain where autonomous execution would create real value, and, just as importantly, where it would not, so investment goes to the opportunities that will pay.

We build the foundations that make it work. We fix the data quality, process design, and planning and operations discipline that agentic AI depends on, so the technology amplifies a strong process rather than accelerating a weak one.

We design the governance and the human-in-the-loop model. We help define what agents can execute autonomously versus what needs human judgment, the guardrails and audit trails, and the redesigned roles for the people who will orchestrate and oversee the agents.

We connect it across the supply chain. From procurement supplier-risk and sourcing through to logistics and distribution, we help you adopt agentic capability where it fits, sequenced sensibly and integrated with the systems and processes you already run.

Explore our supply chain technology capability →

Speak to an expert at Trace →

Where to begin

If agentic AI is on your agenda, resist starting with the technology. Start by asking where in your supply chain fast, autonomous, high-frequency execution would help, and be honest about whether your data, systems, and processes in that area are ready to support it. For most organisations, the first real work is foundational: cleaning and consolidating data, defining the process, and confirming the systems integration that lets an agent actually act rather than merely advise.

Then start small, on bounded and reversible decisions, keep humans in control of the consequential ones, build the governance early, and expand autonomy only as the results earn it. Treat agentic AI as the capstone on a solid foundation rather than a substitute for one.

The autonomous supply chain that runs itself is not here yet, and the breathless version being marketed is still some distance off. But agentic AI that meaningfully accelerates planning, logistics, and procurement, under human governance and on solid foundations, is real and worth pursuing now. The winners will not be the organisations that adopt it fastest or talk about it loudest. They will be the ones that build the foundations first and then let the technology do what it is good at.