How Manufacturers Can Optimise Supply Chains By Fixing ERPs & Layering AI

Recognising a Supply Chain in Trouble

Most manufacturers do not call in a fractional CIO because they want to innovate. They call because something is already going wrong. Inventory is climbing faster than revenue. Service levels are inconsistent. Orders are shipping late or are incomplete. Chargebacks and expediting costs are quietly eroding margins. And finance is struggling to reconcile inventory at month-end.

In this episode of the Comparesoft ERP Podcast, Lee Wachter, ERP & AI Supply Chain Advisor at Interactive Ingenuity, explains that the root cause is rarely the technology itself. The real issue is an operating model breakdown. Planners have stopped trusting the ERP system. Buyers are overriding recommendations. Decisions are being made in spreadsheets, emails, and tribal knowledge rather than inside the system of record.

Lee works primarily with mid-market manufacturers and CPG companies in the $50 million to $250 million revenue range. These are organisations that have outgrown their operating model but have not yet built the discipline and governance needed to match their scale.

According to Lee, the clearest danger signal is decision latency. When something goes wrong in the supply chain and nobody can quickly identify the cause inside the ERP, that is a sign the system has become a passive transaction recorder rather than a trusted decision-making tool.

How to Diagnose Supply Chain Problems Without Jumping to Conclusions

Lee describes a structured diagnostic approach that typically takes between two and six weeks. While experience allows him to sense issues quickly, he resists jumping to conclusions. The goal is not just to identify problems but to communicate them clearly enough to build a credible game plan.

On day one, he is not looking for a single problem. He is looking for patterns in a company’s supply chain management. The first signals are behavioural and structural, not dramatic system failures. He examines whether inventory is growing faster than revenue, whether service levels are inconsistent, whether orders are shipping in full and on time, and whether finance can reconcile inventory without a prolonged monthly close.

Critically, Lee looks for explainability. When a spike occurs or a service level is missed, can someone within the organisation jump into the system, pull the data, and explain exactly why it happened? If not, it indicates either the data is not there or the people are not equipped to use it. Both are symptoms of a deeper governance issue.

Pressure-Testing ERP Systems in Regulated Manufacturing Environments

For manufacturers in food and beverage, aerospace, chemicals, or medical devices, ERP systems must do more than manage transactions. They must support full traceability, shelf-life management, and regulatory compliance.

Lee describes a practical pressure test he calls a reverse recall. Starting from a finished product, he traces backwards through distribution, production, and procurement to determine whether the system can track lot or serial numbers at every stage. If a company cannot demonstrate end-to-end traceability from supplier intake through to customer delivery, the system is not fit for a regulated environment.

Beyond traceability, Lee tests two further areas:

• Shelf-life management: Does the system manage ingredient expiry and enforce first-expiry, first-out logic, or are warehouse staff picking product based on convenience?
• Exception handling: When a product fails inspection, requires rework, or is partially released, can the ERP explain the current status and the steps needed to resolve it?

In regulated industries, traceability is not optional. It is typically mandated by a governing body, and the ERP must support it natively.

Where Manufacturers See the First Measurable Gains In Their Supply Chain Optimisation

Lee explains that the earliest wins from fixing supply chain fundamentals appear in stability and working capital, not advanced optimisation. Inventory is typically the first lever. Once planning assumptions are corrected and the ERP becomes a trusted system of record, safety stock can be rationalised, resulting in decreased costs and excess inventory declines. Cash moves off the warehouse floor and back onto the balance sheet.

Service levels improve next. Orders begin shipping in full, on time, and with the correct documentation. Customer complaints decrease, chargebacks and penalties reduce, and quarterly business reviews with retailers become more favourable.

Less visibly, decision quality improves across the supply chain team. Planners begin to anticipate issues rather than react to them. They can see around corners, manage proactively, and apply more critical thinking to their roles.

How to Decide Whether the Existing Supply Chain ERP Can Be Saved

Lee is clear that he removes the assumption that replacing an ERP system is a solution. In most cases, it is not. His approach is to confirm whether the core 80 to 85 per cent of the system is functioning properly. Are the core transaction flows, including order management, inventory, purchasing, production, and financials, well-established and working?

The real diagnostic work happens in the remaining 15 to 20 per cent. And the issues are rarely about missing features. They are about governance: configuration, master data management, standard operating procedures, testing, and how planning is done. Lee describes this as the tail wagging the dog, where a relatively small proportion of the system’s capability is causing the majority of operational problems.

He uses a cross-functional approach to surface these issues, comparing it to a kaleidoscope. When you get people from different functions in a room, one person’s action can affect something seemingly unrelated. Changing one item can shift the whole picture. This end-to-end visibility is what allows the real root causes to emerge.

What Buyers Misunderstand About the ERP Vendor’s Role

Drawing on his time working inside an ERP vendor, Lee identifies the biggest buyer misconception: assuming the vendor owns the outcome. They do not. The vendor provides software. The implementation partner provides methodology and technical execution. But the ERP is an operating model decision that the business itself must govern.

Lee highlights several areas where buyers should focus their attention during selection and implementation:

• Leadership accountability: Who is the day-to-day business owner, and will they stay through go-live?
• Ownership transition: How quickly does control of the project shift from the vendor to the company installing the ERP? Early ownership creates future leaders who understand the system inside out.
• Governance and transparency: How are risks and issues logged, tracked, and resolved?
• Data and reporting: The earlier data cleansing and reporting requirements are addressed, the better. These are frequently deprioritised during projects and cause problems later.

What Separates an Outstanding ERP Vendor from an Adequate One?

Lee outlines five qualities that distinguish outstanding ERP vendors:

1. Product stewardship: They listen to customers, translate real operational issues into a clear product roadmap, and deliver against it predictably.
2. Partnership discipline: They are transparent about what their product does and does not do, including its weak spots. They remain engaged and accountable when things go wrong.
3. People quality and continuity: Strong technical, managerial, and interpersonal skills are present from executive sponsor to project manager. They do not swap people out mid-project.
4. Execution integrity: Coherent architecture designed for the cloud, rather than stitched together through acquisitions, avoids hidden integration issues.
5. Industry focus and financial stability: Vendors with genuine experience in the buyer’s sector can arrive with the system 80 per cent pre-configured for that industry.

Layering AI on Top of a Stable ERP Supply Chain: What the Process Actually Looks Like

Lee is emphatic that AI should only be introduced once the ERP foundation is stable. AI in ERP is not a plug-in. It is a progression.

The first shift is in data governance. Inside an ERP, organisations deal primarily with structured transactional data: item masters, orders, inventory records, bills of materials. Once AI enters the picture, the data landscape expands to include unstructured data such as customer communications, supplier updates, regulatory guidance, and quality notes, as well as temporal data like promotions, weather patterns, logistics signals, and market indicators.

Lee stresses that before choosing a model, whether for forecasting, classification, or optimisation, organisations must be specific about the business problem they are solving and ensure they have the right classes of data to feed that model. Without this clarity, AI initiatives produce poor results regardless of the technology used.

Lee identifies demand forecasting and sales and operations planning (S&OP) as the area where AI delivers the greatest supply chain impact. Getting the forecast right, by combining structured ERP data with unstructured and temporal data sources, is the foundation for everything downstream. If the forecast is flawed, any model applied on top of it will produce poor outcomes.

Inventory management follows closely. Once S&OP is functioning well, AI can help optimise inventory policies, minimise stock levels, and maximise working capital. The key question becomes what customer service level the business actually needs. A hospital may require 100 per cent availability. A consumer goods company may operate effectively at 90 or 96 per cent. The right inventory policy depends on understanding that trade-off.

Non-Negotiable Guardrails Before Going Live with AI

For leadership teams wanting to move fast with AI, Lee does not slow them down but insists on discipline. Speed without guardrails destroys value. He outlines five non-negotiable requirements:

• Explicit decision authority: Every AI-supported recommendation must have a named owner with the authority to act on it, override it, or stop it. Without accountability, AI can quickly destroy value.
• Data integrity at the point of decision: The data feeding the model, whether from the ERP, external signals, or other sources, must be trusted, monitored, and tested. Training and results must be continuously validated.
• Economic intent upfront: AI must be tied to a specific, measurable outcome from the outset. Like a capital budgeting model, there must be a clear return.
• A human in the loop: Someone with the time and capability to apply critical thinking to what the model is doing, how it is interpreting data, and how best to use the output.
• Organisational ability to absorb change: If the AI recommends an action, can the business actually respond in real time? Moving too fast can destabilise operations even when the AI is technically correct.

Lee’s message to leadership is clear: do not slow down, but make sure you are taking the right steps. Following AI without these guardrails risks taking the operation in the wrong direction with nobody positioned to correct course.

Full Transcript

Ryan (00:00): Lee, it’s great to have you on the podcast. How are you?

Lee Wachter (00:04): I’m great. It’s great to be here, Ryan, and I appreciate the hour we’re going to spend together. I think it’ll be very good.

Ryan (00:10): Excellent. So you’re currently running your own independent operation, Lee. How would you best describe your role as a fractional CIO?

Lee Wachter (00:19): Yeah, I’m running Interactive Ingenuity and basically I work with mid-market manufacturers and consumer goods companies, typically $50 to $250 million in revenue. And really what they’ve done is they’ve outgrown their operating model. They may not trust the data that they’re utilising and they’re not running the business as efficiently as they could. So I’ve served as a CIO and COO and worked inside an ERP vendor, and I’ve seen the supply chain from both sides.

So in my role, what I try to do is align the technology, the operations, and the financial performance specifically within the supply chain. As a fractional CIO or CTO, I don’t really operate like a traditional CIO. My role is to ensure that the ERP system is running efficiently and that it’s a trusted system of record that management can rely on to make decisions about inventory, procurement, and the execution of that.

When I’m called in, I help leadership stabilise the foundation — governance, planning discipline, and all of the different functionality associated with running a supply chain — and make sure that we’re moving off of spreadsheets and that you’re actually using the ERP system to make decisions.

Then once I do that, I also help the people who run the supply chain get to the data that they need in order to assess the situation, analyse the data, and come up with the best decisions they can. And only once the foundation is stable and the organisation can respond quickly and people trust the system as the system of record, does it make sense to layer AI on top of that to improve decision quality.

Ryan (02:18): That’s excellent. You’ve positioned yourself as someone who assists these manufacturing and CPG companies in recovering and optimising their supply chain operations. I’m curious, what’s the most common stage you’re called in if they want to get their supply chain operations in order?

Lee Wachter (02:37): Sure, that’s a great question. Most of the time I’m called in, the ERP system may be up, it’s live, transactions are flowing, but there’s some kind of inventory problem, a service level problem, margins are eroding through chargebacks, there’s a lot of expediting going on — in other words, moving goods quickly and paying extra for that. And essentially management begins to feel a sense of loss of trust relative to what the system’s doing. Planners are relying on spreadsheets, buyers are overriding the system, and there’s a lot of decision-making going on outside the system itself. So I’ll get called in. And when that happens, it’s really not a technology issue. It’s an operating model issue.

Another point is I often get called in if there’s a new CEO or a new CFO and they’re in the midst of an implementation and something is going sideways. So I’ll come in, assess where we are, and then figure out what’s the plan we’re following, what corrective actions do we need to take, and make sure that we can get the programme and the implementation project moving ahead smoothly.

My role in that case is to assess quickly, determine what the breakdown is, come up with a game plan, and move forward from there. The objective in either case is straightforward: getting the ERP operational, the data operational, people comfortable with it, the trust level up. And only when those factors are in place does it make sense to move into AI where we can create real value.

Ryan (04:29): So you touched on a few telltale signs there, some signals where the supply chain operation isn’t at its full capacity. How long does that typically take you to understand what’s going wrong? How long does that process take?

Lee Wachter (04:45): Well, I’ve got a double-edged answer there. Firstly, I’m going to feel it, I’m going to sense it, I’m going to smell it, and I’ll kind of know it immediately. But immediately is too quick — I don’t want to jump to any conclusions just because I’ve seen this movie before. So basically I’ll know what’s cooking, but the more important thing is I need to be able to assess what’s really going on and then communicate that clearly and in a structured manner to management and the operations people. I can’t just walk in — I need to really have my ducks in order to go down to the right level of detail.

So I’ll know quickly, and it depends on the extent to which they’re using the ERP system. It can run anywhere between two, four, six weeks, something to that effect. Because at the end of that, what you really want is a game plan. These are the steps that we need to take in order to fix and address the system. So it’s really putting the whole programme together that makes the difference.

Ryan (05:48): And what sort of information are you asking for? What sort of data do you want from that company on, say, day one to understand and organise planning, procurement, production, and fulfilment — what that actually looks like?

Lee Wachter (05:55): Sure. So when I walk in on day one, I’m not necessarily looking for one particular problem, although they may say our inventory is too high or our response times aren’t where they need to be and we’re not fulfilling our customer obligations. What I’m really looking for are patterns. And the first signals are rarely dramatic system failures. They’re more behavioural and structural.

Inventory may be climbing faster than revenue. There are inconsistent service levels. Orders are not being shipped in full and on a timely basis. There are chargebacks, expediting of incoming goods. Finance is struggling to reconcile inventory. The month end is a problem. So those are signals of instability, but they may not be the root cause.

The real flag is lack of trust in the system. People may be working outside the system, working with satellite spreadsheets, overriding what the system is saying. So now you’re operating your business on spreadsheets and tribal knowledge in addition to the ERP system because they’ve lost trust in it.

The next thing I look for is explainability. When inventories go off or you miss a service level or a spike occurs, is there somebody within the organisation that can jump into the system, look at the data, and say this is the exact reason why that happened? And if that doesn’t happen, that tells me another issue is going on because the data is either not there or the people are not sufficiently versed on how to get to the data to understand the root cause of their problems.

This results in decision latency. You’re consistently late making or delivering orders and you’re not finding out until five days later. Meanwhile, you’re getting complaints from the customer and nobody in your organisation knows what’s going on or how to handle it. So the real thing I’m looking for is when something changes or something’s off kilter, how quickly is management made aware of it? How quickly can they identify and diagnose the problem to bring it to resolution? If they’re running around with emails, spreadsheets, and it’s a little bit chaotic, that’s not a good sign.

Ryan (08:30): Right, and I’m really interested — you’ve had experience in food and beverage manufacturing environments too, which can obviously throw up its own set of unique challenges. How do you pressure-test whether an existing system can handle things like traceability, shelf life, and compliance? Are there test scenarios you can run?

Lee Wachter (08:37): Yeah, what I really try to do is get down to the bottom and do a reverse recall. Whether it’s food, beverage, chemicals, aerospace, or medical devices, each of those is going to have some type of regulatory requirement in order to respond to that. So you’re going to want to be able to look at the entire supply chain and have traceability down to either the lot or the serial number.

It’s all pressure-tested from procurement. When a product or a component or an ingredient comes in the door, does it go through QA? Then where is it used? Is it now combined into a different product? And do you have a lot number or a serial number associated with that? Then if it gets into the warehouse and goes to distribution and hits the customers, you need traceability all the way from procurement to distribution and the final sale. So I’ll run through a pressure test like that — I’m calling it a reverse recall — and see if they can really do that or not.

The other one is how they’re managing shelf-life ingredients. For example, if something expires, how do they know which product to ship? The one that’s going to expire soonest, or is a warehouse person just walking in and picking product arbitrarily? That’s another very important component — shelf life.

The third one is exception handling. When something goes wrong or fails an inspection, or there’s rework or a partial release, how do they manage that? Can the ERP explain the current status and the steps that are supposed to be taken in order to address that particular issue?

As I mentioned earlier, in a regulated environment, traceability isn’t optional. Most of the time it’s going to be regulated by some governing body. So you really need to have these pieces in place, particularly given the industry that you’re operating in.

Ryan (11:07): A very thorough approach. So if we take it hypothetically, you get to the stage of understanding a business’s supply chain issues, you then advise on either remediating or replacing the existing system to ultimately fix and solve these issues. Where do manufacturers typically see the first measurable gains once their supply chain fundamentals are fixed?

Lee Wachter (11:35): The first gains usually show up in stability and working capital, not necessarily advanced optimisation. Inventory is generally the earliest lever. Once planning assumptions are corrected and the ERP becomes a trusted system of record, you have safety stock that you can rationalise, expediting of orders decreases, excess inventory starts to decline, and the cash moving off the warehouse floor is going back onto the balance sheet. So you’re beginning to see improvement from an operational perspective and it translates to an increase in working capital.

The next thing that happens is that service levels begin to improve. You’re now shipping in full, on time, and with the right paperwork, and you’re getting fewer customer complaints. Theoretically, that’ll turn into stronger retail or customer scorecards when you sit down and do a quarterly business review with them. This will reduce your chargebacks, penalties, or any customer escalations.

There’s also something less visible: the decision quality of the people operating the supply chain improves. They’re beginning to anticipate issues. They can do a bit more critical thinking about what their role is. And if they’re anticipating issues and can see around corners, that will enable them to manage the supply chain in a much more effective and efficient manner so that customers are happy, your margins are improving, your working capital is going down, and you’re moving the business in a proper forward strategic direction.

Ryan (13:28): How do you decide whether the existing ERP can be fixed and optimised versus when a system actually needs to be changed?

Lee Wachter (13:35): First off, I remove the assumption that replacement of the system is a solution, because in most cases it isn’t. What I try to discern is whether 80 or 85 per cent of the system is functioning properly. I want to confirm that in my own mind. Are the core transaction flows — order management, inventory, purchasing, production, financials — well established and working well?

However, what I really try to do is figure out what may be problematic in the other 15 or 20%. And it’s not necessarily about whether the features are available in that system. It’s about governance. Is the configuration set up? Is the master data set up? Is it being properly managed? Is there appropriate governance? Is there appropriate testing? Are there appropriate standard operating procedures? How is planning done?

Then I go from there and figure out how exceptions are handled, traceability, and so forth. I confirm that 80 or 85% is working, and then I try to figure out what’s going on with the 15%, which is kind of like the tail wagging the dog but causing all of these problems. So it’s not about whether the ERP is good, but whether it supports the non-negotiables of this business and whether the proper governance is in place to make sure those things are being properly addressed.

In food and beverage, it could be traceability. In aerospace, it’s serial and lot number control and quality history. In medical or regulated industry, it’s compliance enforcement. If those things specific to that particular industry are available in the software, is the company actually using the software and all its features to their fullest capability?

I really try not to say this needs to be replaced, but rather what do we need to do in order to bring the system along. The other thing I do is when I look at these problems, it’s kind of like a kaleidoscope. It’s amazing when you get cross-functional people in a room — one person is doing one thing and it can affect something completely unrelated in their mind. But when you turn a kaleidoscope, if you change one item, the whole picture changes. So you get these people in a room so that they fully understand the end-to-end impact of what they could be doing and how it might affect another function. When I get them in a room, it also helps to really drill down and get into a lot of detail behind what’s really going on in the system.

Ryan (16:29): Excellent. And if it comes to selecting a new system, we had a nice chat last week and you said roughly 80% of ERPs do the same thing. In your opinion, what is that 20% that actually differentiates the supply chain outcomes?

Lee Wachter (16:46): Yeah, let me take a step back. You’re right, 80% we discussed this. In general, 80% of an ERP system across ERP offerings will be very similar in nature and support basic functionality. It’s when you get to the 15 or 20% of business operations that are critical to that business — how do you ensure that the particular system or the ERP you’re evaluating will support those?

What I look for is to identify, of all the capabilities and things the system needs to do, what are those 15 or 20% that are critical to that particular industry and to that particular business? That’s where I start. I want demos and I want the vendor to be able to set it up, show the data, demonstrate the input, demonstrate the output, and make sure that those particular pieces of functionality are being supported. Again, it could be traceability, it could be regulatory issues, but you really surface those.

What I really try to avoid is creating an RFP. I don’t do anything like that. I zero in on the strategic issues that provide competitive advantage or regulatory support for that business within that industry and make sure those things are addressed. Then I move on to other areas such as whether there’s a cultural match between the company and the implementation partner. What’s their methodology? Will their methodology fit into the approach that the company looking at the ERP will need?

It’s not only the features and functionality for those strategic items. It’s also the methodology, the people, a cultural match, and all of this coming together so that you can have a successful project.

Ryan (19:06): Yeah, and you touched on demos there, so important in this type of industry. Which sort of system capabilities tend to look convincing in demos but are most likely going to fail when they’re introduced on the plant floor or the warehouse? Is there an example of something that looked great in a demo but failed in reality for you?

Lee Wachter (19:16): Yeah, a couple of examples come to mind. It may not be a specific feature, but it’s an assumption that the demonstration provides real-world scenarios and fit. I go back to the company really needing to define their true non-negotiable features upfront. That’s what’s going to make a difference when you’re looking at and running a demo. You want to see how the user interface works. You want it to be easy and intuitive. But you really need to drill down into those things that are critical for that particular industry.

One example might be planning coherence. In a demonstration, planning, inventory, and execution appear all synchronised. However, in reality, if a supplier lead time shifts and the master file isn’t updated, how do you know that? You’ve got to be able to get the early warning signals to let you know something’s off and either adjust your planning data or not.

Another one is regulatory documentation. If you’re distributing chemicals, for instance, there’s a whole series of documents that need to go along with those chemicals when they’re picked, tested, and shipped out the door. Same thing for aerospace — regulatory, serial number, lot number tracking, QA — those are critical. So those would be the areas you would probe.

I don’t necessarily judge the ERP system on how broadly they demonstrate features, but I really try to determine whether they support the company’s true operating requirements in standard mode. That’s where the difference comes in. A 150-page book that you score and somebody gets a 92 and someone gets an 88 isn’t going to tell you the real story. The real story is having your user community in front of the demo, participating in the demonstration, and watching in real time how those critical features are being supported by the ERP system.

Ryan (21:41): And have there been situations where you’ve been able to test real-world data in these systems before going live, using it as a demo test?

Lee Wachter (21:52): Yeah, definitely. If you can pull together the documentation — and if you’re already running the business, you already know what these are — you can give the vendor the documentation, the bill of materials, the QA checklist, all these things, and say show me how you do this. Then show me the reporting associated with that so I truly know that it’s being supported. Absolutely, that’s a great approach.

Ryan (22:19): From your time working within an ERP vendor, you’re situated in a really good position in terms of ERP selection. What do buyers misunderstand about the vendor’s role in that selection? Are there questions you ask that most buyers won’t?

Lee Wachter (22:34): That’s a great question because the biggest misunderstanding I’ve seen is assuming the vendor owns the outcome. They don’t. Leadership of the company that’s going to install the ERP owns the outcome. The vendors will provide the software. Partners will provide a methodology and technical execution. But the ERP is an operating model decision that the business must govern.

So where I focus is on discipline and governance. Leadership accountability is critical. Who’s the day-to-day business owner? Will they stay through go-live? How quickly does ownership of the project transition from the vendor and become assumed by the company that’s installing the ERP? It’s amazing — you could be running a conference room pilot and the vendor will continue to do it, but those companies where someone stands up and says I’m going to lead the CRP, I’m going to do it, that makes a tremendous difference. Those people become your future leaders. They’re going to embrace the system, learn about it, and know your transactions inside and out. They’re also going to be able to operate the system. So that’s really critical.

We touched on governance and transparency. That’s critical in terms of how a risk log and issues are managed and whether they’re all being ticked off and addressed. Another piece is data and the cutover, as well as reporting. In some cases, the earlier you start cleaning up your data, the better off you’re going to be, and imposing governance — in other words, having single ownership and stewardship of that data.

The other thing that sometimes gets lost during the project is reporting. The earlier you get on top of your data, its governance, and the reporting that you’re going to need, the better off you’re going to be. Because you will be focused on the transactions, but now you’re going to be looking at the data and the output of those transactions. Those are the kinds of things you need to be aware of when you’re evaluating and working with a vendor.

Ryan (25:04): Yeah, excellent advice. And obviously an ERP system depends on your requirements and the capabilities it’s offering. We’ve touched on how 80% are quite similar. In your experience, what separates an okay ERP vendor from an outstanding one? What sort of qualities should you be looking for?

Lee Wachter (25:15): The first one is product stewardship. Outstanding vendors listen to the customers, translate real operational issues into a clear product roadmap, and they execute against it predictably. In other words, if they say something’s going to be released in six months, they need to hit those targets. They don’t just talk about innovation — they deliver it.

The second one is partnership discipline. Vendors that are transparent about what their product does, what it does not do, and its weak spots. So you can have true conversations about how to mitigate risk and move things forward. They don’t overpromise in sales and then disappear after delivery. When something goes wrong, they remain engaged and accountable.

The third is people quality and continuity. You need strong technical skills, strong managerial skills, and strong interpersonal skills — from the executive sponsor all the way down to the project manager. There needs to be continuity. You don’t swap people out. You need to build trust and deliver on any promises that you make.

Execution integrity is just as important. Vendors with coherent architecture designed for the cloud, rather than stitched together through acquisitions, avoid hidden integrations and other things you might run into. They’re executing their architecture very well.

And finally, industry focus and financial stability matter as well. It’s great if you can find a vendor that understands CPG, aerospace, chemicals, distribution, healthcare, or whatever your industry is, so that they’ve configured their system maybe 80% before you even begin your project and it’s geared towards your industry. So it’s the continued communication, the product roadmap, the transparency, delivery against that, the quality and stability of the people, their knowledge of the system, and the collaboration you can establish.

Ryan (27:44): That’s excellent. I feel that industry experience is actually quite an underrated capability of an ERP vendor — having that experience where they’ve done it before and they know what works.

Lee Wachter (27:56): Right. The industry experience can be very critical, particularly when you’re bringing a team in and they say, wait a minute, you’re in healthcare and I really only know CPG, or something like that. They may know the system, but they’re not going to know all the little nuances.

Ryan (28:07): Yeah, exactly. So after planning, strategy, and system selection, you then advise companies on how they can leverage AI by layering it on top of their already functional ERP system. This is something that’s really exciting. What does that process look like in practice, Lee, from the moment an ERP stabilises?

Lee Wachter (28:34): That’s a really good question. I’m really excited about getting into this topic. Once the ERP is stabilised, the focus shifts from running transactions to improving decisions and making them more timely. AI layering is not a plug-in — it’s a progression. And you can do these things simultaneously.

Let’s talk about data governance for a second. It changes once you move into AI. Inside an ERP system, you’re mostly dealing with structured transactional data — item masters, orders, inventory, production records, bills of materials, things like that. As soon as you begin thinking about AI, the data landscape shifts.

Now you’re dealing with unstructured data or even temporal or semi-structured data. You’ve got your structured data within the ERP — your master files, products, pricing, and so on. But now there’s a whole set of unstructured data, which could be customer communications, supplier updates, regulatory guidance, quality notes, or service logs. You have all this inbound information either from a regulatory arena or from a customer. So how do you take that and leverage it as part of your AI programme?

Then the last piece is time series data, which relates to promotions, weather, logistics signals, and market indicators. This also doesn’t exist in your ERP system. So you have to keep these kinds of data in mind as you go forward.

But the real thing is, once you’re on top of your data, what is the business issue that you’re trying to solve? And given the data that you have, what kind of model might you want to use? For example, I’m going to get a little technical here. Mixed integer linear programming is not necessarily AI. You could do it on a spreadsheet or run it with a computer programme. It’s mathematics and optimisation. It could help you with your footprint or with tariffs. But you’ll only be able to run that if you have those three classes of data — the structured data, the unstructured data, and the temporal data — all aligned in order to feed that model.

I don’t think people always have a clear understanding of what data governance really means in this context. But it’s important to do a deep dive into what it actually requires. The second piece is what are we going to do with AI to boost our decision-making, optimise forecasting, optimise our footprint, logistics, transportation, products, and pricing? Because you’ll need the additional data. You need the right model and the right data. That’s why I say only once an ERP system is stable is it right to use AI. It’s not something that you just bolt on. It’s disciplined capabilities that are layered on, and you need to ensure you have the right data.

Ryan (32:03): Is that a common misconception from businesses that you’ve spoken to? They come in and say, right, I want to lay AI on top of it, and you have to almost say, no, stop, let’s get everything in order first. Is that a common theme you’re seeing?

Lee Wachter (32:17): Yeah, what I’m seeing is that people will come in and talk about data governance, but they may not even be thinking about what kind of model they want to feed the data into to solve what business issue. People aren’t always thinking about the fact that we’ve got structured data, unstructured data, and a lot of different sources and types of data — which is exactly what AI can really help you prepare and put into the kind of model that you want to use. It could be a prediction, a classification, or a forecast. You really need to be very specific about it.

The other considerations are what is the ROI? Is it that powerful? What learning do I need to do? Because at the end of the day, you may need a very capable statistician who understands the tools that are going to take this data and provide you with the capability or forecast that you’re trying to achieve. I’m not sure that people are truly thinking about all of that. Those are some really key elements — data and what kind of model you’re going to use — and making sure that AI helps to prepare the data. Because ultimately, if you’re not doing that properly or you don’t have anybody who really understands what it takes, when you’re training and creating the model, you’re just going to end up with poor results.

Ryan (33:49): Where have you seen firsthand AI integration have the biggest impact on optimising a supply chain?

Lee Wachter (33:57): I think people are really trying to forecast demand volatility and come up with a forecast and make sure there’s not a bias. Many companies are doing sales and operations planning — they get their sales organisation together, their operations people together, and they want to drive their MRP and manufacturing planning.

If they don’t have a very good forecast and they’re not pulling the structured data plus the unstructured data that I’ve talked about and leveraging AI to come up with what they think is a realistic forecast, then any model I apply to it is not going to be very good. So I think a lot of people start with sales and operations planning because that’s the heart of your supply chain.

There’s another area too. Once you have that going, it leads into inventory management, because basically you want to minimise your inventory and maximise your working capital so you can make investments elsewhere. So those are two areas — the S&OP and the inventory policies that you’re putting in to achieve a certain customer service level. Is it okay to be at 90% or do you need to be at 96%?

In a hospital, you don’t want to show up and say we didn’t have enough material to do your particular procedure today, or we couldn’t do a diagnostic. That’s not acceptable. In that case, you need 100%. But in other businesses, it may be perfectly fine to achieve 90% or 96%.

Ryan (35:42): Yeah, excellent. So Lee, we like to save these last few questions for giving a bit of insight and advice from your experience. Looking back at some of the things you’ve done, you’ve delivered high-value outcomes like 30% inventory reduction and eliminating one to two million dollars in chargebacks. What do those successful projects usually have in common behind the scenes? Is there a common point in terms of decision-making, trade-offs, and behaviours?

Lee Wachter (36:10): Yeah, basically what they have in common is disciplined, detailed decision-making, but not necessarily a heroic effort. The decisions needed to drive inventory down or eliminate chargebacks or achieve any other particular goal start with really understanding the problem at a detailed level and the best way to approach it. Your decisions are explicit as you go about navigating and addressing these issues.

The other thing is you need to make some trade-offs. For example, at what point is acceptability okay for that organisation’s retail scorecard? Do you really need 100% or can you be at 90%? Because to achieve 100%, that may mean you’re tying up all sorts of working capital in inventory. And the reality is that may not make sense. So maybe you really need only a select number of SKUs and colours in order to achieve a comfortable level of customer scorecards, because the retailers are always going to want you to have everything in stock so that every single order can be fulfilled in a timely manner.

It’s the analytics and the attention to detail. Are you looking at the KPIs? Are you monitoring them on a frequent basis? And what do you need to do to make things even better? And finally, collaboration both internally across different silos and with your business partners and customers will yield the best results. Once you have everybody aligned and on board, you can achieve these kinds of significant benefits across your supply chain.

Ryan (38:08): Excellent. Lee, I really want to finish this chat on AI layering — it’s interested me a lot. If a leadership team wants to push for go-live quickly with their AI layering and wants to really get it up and running, what non-negotiable guardrails do you insist are in place before doing so?

Lee Wachter (38:28): When leadership wants to move fast with AI, I don’t necessarily slow them down, but I insist on the guardrails. Speed without discipline is going to destroy value.

Decision authority needs to be explicit. Every AI-supported recommendation must have a named owner with the authority to make decisions, override it, or stop it. If no one has this decision-making authority and is accountable, then the AI could quickly start destroying value.

We talked about data integrity a number of times. Data integrity at the point when you’re making decisions is non-negotiable. The data feeding the model — its quality, whether it’s external signals, ERP data, or whatever — needs to be trusted, monitored, and tested so that your training and the results of your training are always consistently being monitored. You need to know that your model is going to give you back what you’re looking for.

The economic intent or ROI must be upfront. The AI needs to be tied to a specific output. The earlier you can do that, the better off you’re going to be. It’s kind of like a capital budgeting model. If somebody wants to make an expenditure, there needs to be a return. Unless you’re experimenting and learning, there needs to be a return.

Another thing that’s critical is that there must be a human in the loop. This person should have time to do some critical thinking — not only about your data, but also about what the model is doing, how it’s interpreting things, and how to best leverage the AI. People talk about jobs being replaced by AI. I don’t necessarily buy into that fully. I think you’re always going to need the critical human-thinking aspect serving as a guardrail.

The last one is the ability to absorb change. If the AI is saying something, can you really respond to that in real time? And does it make sense? Moving too fast can destabilise your operation. Even though AI is meant to improve it, you need to make sure you’re moving at a speed that you can manage and operate your business safely. Without these guardrails, if you’re just following the AI, you’re going to take a wrong turn and no one’s going to be happy. Don’t necessarily slow the speed down — just make sure you’re taking the right steps in order to leverage AI.

Ryan (41:20): Excellent. Well, Lee, you’ve been a brilliant guest. Thank you so much for sharing your practical insight and real-world experience, particularly around what it really takes to stabilise supply chain operations and how to apply AI in a way that actually delivers value. Thanks again for joining us.

Lee Wachter (41:37): Yeah, it was a pleasure. Thank you, Ryan. I appreciate this opportunity. And I really thought you gave a lot of thought to how this could work well and structured it well. So thank you.

Ryan (41:46): Thank you very much, Lee. And thank you all for listening — we’ll see you on the next episode.