Before AI: Practical Automation for Manufacturers
Originally published on June 10, 2026
“A lot of things really shouldn’t use AI. Try to find things that you can automate first. Have that be the first thing you can take an approach to.” — Daniel Shorstein, President of Digital Services, James Moore & Company
In this Moore on Manufacturing episode, Mike Sibley and Kevin Golden are joined by Daniel Shorstein, president of digital services at James Moore & Company, for a grounded conversation on how manufacturers can use automation and AI in ways that actually work. The discussion moves from data quality and governance to workflow improvement, strategic planning and knowing when dabbling is no longer enough.
Shorstein brings experience across multiple industries, including federal government, higher education and manufacturing, giving him a clear view of the patterns that hold organizations back and the steps that help them move forward.
Watch the Episode
Resources
Full Transcript
[00:01] Mike Sibley: Hi, I’m Mike Sibley, partner and leader of the manufacturing team here at James Moore & Company. I’m joined by Kevin Golden, partner and also member of the manufacturing team, and today we’re excited to welcome Daniel Shorstein, president of our digital services here at James Moore. Daniel works with organizations to help bridge a gap between technology strategy and business execution with focus on automation, AI, digital transformation, and operational efficiency.
[00:29] Mike Sibley: This is a topic that data in general is just everywhere, whether we’re talking about automation, AI, a combination of all of those things. Companies are filled with data and often don’t even know what to do with it. Today we’re going to talk about how manufacturers can use automation and artificial intelligence to improve efficiency, reduce repetitive work, strengthen decision-making, and really better position for growth and capacity. AI is everywhere. It’s one of the biggest topics in business, and businesses are all over the place in their implementation or their ability to execute. In a lot of cases, manufacturers are maybe just dabbling and trying to determine what’s practical, what’s possible and what’s not.
[01:21] Mike Sibley: I also think automation is a very important tool that is available, that can work together with AI, can work separately, but it’s not a one-size-fits-all. So, Daniel, welcome to the show. Glad to have you.
[01:42] Daniel Shorstein: Thank you. Happy to be here.
[01:44] Mike Sibley: As we get into a general state, maybe that’s a good place to start. Before we go into manufacturing specifically, you work across industries. Are you seeing the same struggles with data and AI out there in general?
[02:10] Daniel Shorstein: Yes, that’s correct. I don’t focus primarily on manufacturing, but I have some manufacturing clients and a lot of experience across a number of industries. Federal government, state and local government, and higher education are some of the large groups that I serve. I do see a lot of consistent challenges between the different industries I work with, including some of the manufacturing clients I speak to. A lot of it starts with data. You’ve got data that comes from different places, and depending on the client, it could be at different levels of reliability and consistency.
[03:05] Daniel Shorstein: That’s even just the biggest starting point that a lot of clients I talk to about AI or even automation mention. They say, “I can’t even automate anything. I can’t even use AI because my data stinks. I can’t rely on that.” That’s a common thread regardless of who you talk to, but especially in manufacturing. There’s a lot of data in manufacturing. But on the automation versus AI front, there’s a similar thread. There’s a lot of focus especially at the board level saying, “We need to do more AI. Let’s do everything with AI.” But in reality, a lot of things really shouldn’t use AI. I usually like to say try to find things that you can automate first. Have that be the first thing you can take an approach to with rules-based automation. If you can’t automate a process through traditional rules-based methods, then you can graduate to using AI potentially. But if you start by just trying to throw AI at everything, you might end up complicating something more than necessary, making it take longer, cost more, and have a bigger risk that the outputs will be incorrect.
[04:13] Mike Sibley: Yeah, and everybody’s throwing around AI like it’s the answer to absolutely everything. AI is still in a tricky phase right now in my mind because it can do a lot of good things, but there’s a lot you have to be careful about. Systems, security, governance. You have to put things in place to make sure that data remains confidential, that people are reviewing outputs, because you have to review it and make sure the data’s right. It’s not perfect. So I think AI, while it’s good, is getting a little out there in terms of people thinking it’s the absolute answer to everything.
[05:13] Daniel Shorstein: That’s an interesting dynamic I hear across the board. Sometimes I’ll hear folks who want to throw it at everything because they think it’s the end-all, be-all. But then sometimes people are at the complete other end of the spectrum, saying, “I don’t want to use it for anything because I can’t trust it at all.” The honest answer is it really should be somewhere in the middle. If you find a process that AI could be a good solution for, you just have to dig in a little bit more. You think about it not just from an individual AI-to-automate-a-thing perspective, but from a workflow perspective. Like lean manufacturing, you can apply the same concept to your processes in the business office. You look at the end-to-end process and figure out what step takes the most time, has the most risk, and could easily be automated either through a traditional method, or if it takes some kind of decision making or inferring, that’s where traditional automation can’t work and AI comes in. But governance is really important. You want good policies around how and where you can and should use AI, where you shouldn’t, and what types of risk assessment you want to take before implementing it, as well as what monitoring you want in place. That way you can make sure AI is giving you accurate, reliable outputs and that the combination of AI and human can be put into the workflow safely.
[06:58] Kevin Golden: One central theme I’m hearing through this is trust. If cars started driving us tomorrow and I don’t have to do anything, I know I’m not going to be a first adopter because it’s a trust factor. Some people are hesitant to adopt, and then you’ve got people who go the exact opposite and dive way in. Would you say, Daniel, that this mostly comes down to trust and what you can do to get comfortable with it and build that trust?
[07:54] Daniel Shorstein: 100% agree. I would even take it one step further. If you’ve heard the term “trust but verify,” that works perfectly here. I like to compare AI to a human doing something. You typically wouldn’t just hire somebody, give them a very important task, and then let them run with it without any oversight. You still have to hire people and give them responsibilities, and the way you build trust in that person is similar to how you might build trust in an AI tool. You give it a role, some expectations, and you start to trust it more as it does a good job, but you still want to verify it as it goes. The more of a relationship you build with the process and its capabilities, the more you might trust it, but you still want to verify it. Even somebody who’s been leading a company for many years still has a board of directors they report to, because you still need to verify that incentives are in the right place. From the human side it looks a little different with AI, but that verification still needs to be there from a system standpoint.
[09:35] Mike Sibley: Let’s dive a little deeper now. Where do you see manufacturing companies right now in their adoption of automation and AI?
[10:02] Daniel Shorstein: Maybe I’ll talk specifically about the financial management space and then more broadly, because that’s the world I live in. It is a little different depending on the area and the size and maturity of a company. Some of the manufacturing clients I talk with are maybe a little smaller, don’t quite have a full AI program or leadership with AI training. They’re probably a little earlier in what I call the maturity curve of applying automation and AI. They’re often having individual users use things like ChatGPT or Claude to do individual tasks, but not looking at the broader workflow perspective. That’s fine. You have to start somewhere. At that stage I usually suggest they start with governance and assign responsibility at a leadership level for someone who is responsible for growing AI capabilities, but doing it responsibly and thoughtfully. Then starting with light experimentation, working on the data side, and once they have the data and governance in place and understand where the use cases are, they can go from there.
[11:06] Daniel Shorstein: The biggest disconnect I see in manufacturers, even at the mid-market and larger level, is that the financial management folks often don’t have the same level of AI training and skills. The language about using it and the creative thought process. Not that accountants aren’t creative, but often we’re more focused on our process and what we understand. It’s challenging when the IT group leading the AI initiative doesn’t know enough about the accounting process, the data, and the challenges, and tries to force AI into the accounting world when they don’t really understand. There’s a lot more opportunity, but the language isn’t really there between the two groups. It really takes somebody who has enough skills and experience in both the accounting and reporting space, as well as the right understanding on the AI side, to make good progress building out workflow-based use cases.
[12:24] Mike Sibley: Going back to something we were both talking about: governance. I think it’s really important. There’s a framework you and I have talked about, Daniel. You’ve got to have security around this. Cyber and all that kind of security with policies to protect your data, your customer data, your employee data. The second thing is a governance policy around this. What’s acceptable and what’s not. And then from there, the next important thing is training. How do you use it? Everyone can sort of get started by creating an email or doing a search, and those are great starting points. But the larger point is getting into workflows and how to insert AI and automation into workflows. That’s when you’re really starting to get the power of AI, and that’s operations all the way through finance. But I wouldn’t go down that road until you look at the security and governance pieces first. Do you agree with that?
[13:47] Daniel Shorstein: I would definitely agree. It depends on the maturity of an organization and the skill level of different people at the organization. At James Moore, we have a group doing some experimenting with some of the more advanced tools, but we’re not deploying them at full scale yet because we are taking the time to make sure we’re comfortable from a governance standpoint. That’s the same thing we recommend. If there are users with the right level of capabilities and creative thinking to come up with use cases, it’s actually great to allow and encourage them, give them a safe space to play with them, whether it’s with fake data or in a sandbox environment, while the organization overall figures out how to really deploy that at scale, but still safely. If you don’t have employees with the right level of skills to do that kind of experimentation, you can grow those capabilities alongside the organization growing in its governance and safety.
[15:02] Kevin Golden: Even the education part, because just like any new technology, until you get in there and start using it and thinking about it. I like that idea of a creative space, Daniel, because we talk about on the show all the time how operations and finance connect. That could be the same thing you’d be using later for automation or AI on top of that. Even someone thinking, “Okay, what problems do we have?” Most operational folks can help figure that out. But thinking about how this technology or automation could layer on to make that better. Now you’ve got the safe space and the governance around it. That actually works. Here’s what it’s actually resolving or making quicker and easier. And so yeah, once someone starts thinking that way, they list off five other things in their work where AI or automation could improve things for the company as a whole.
[16:10] Daniel Shorstein: 100% agree. The biggest challenge I see everywhere is unlocking that creative thinking part of the process. I don’t know how to solve for that yet. It’s hard. There are certain people who are just naturally curious and will spend free time thinking of different ways to do things better and faster. But I haven’t seen it as much more broadly. The challenge is how do you teach someone to think creatively? There’s no textbook on it, especially when it comes to business processes. It’s almost as much of an art as a science. But to me, one of the best ways to teach people is just by showing them or working with them. If you have one person on a team doing solo creative thinking and automation, instead have them work with one or two other people and bring them along. Talk out loud. Why am I picking this thing to automate? Well, it’s because I don’t like doing it or there are a lot of repetitive steps. I’m just going to take 15 minutes to try this out with Claude or ChatGPT and see if it works. Enabling people to fail, and teaching them that failing is okay and good. In this case, you could fail very quickly. When I say fail, I don’t mean you’re actually failing at something. You might try something with AI and nine times out of ten, it might not work. But it’s that one or two times out of ten where you’re like, “This is amazing. I didn’t even think it could do this.” That unlocks everything for you. And then you learn to find the failures even more quickly, so the iteration process happens even faster. But people are sometimes naturally afraid to fail or to try something and have it not work.
[18:36] Mike Sibley: Let’s switch gears a little and talk about practicality. You just spurred some thoughts, Danny. One of them is that experimentation often doesn’t happen from eight to five. It’s that quiet time on weekends when you’re playing around on your own personal R&D time. I make that investment to try to do that. And the other thing is I think in terms of systems and processes and flowcharts. When you start drawing a picture of a process, you can start seeing the areas where automation or AI might play in. Now I’m going to put a shameless plug in here. You may not know how to do it, but that’s why we’ve got people like Daniel who can come in with imagination about how to do these automations. Some things are simple, but some things are really complex. That’s where you start thinking about what process is really manual and takes a lot of time. How do we automate those things? I even had a scenario with a month-end Excel spreadsheet that was tedious, took a lot of work, and always had a lot of errors. When it got to me from the client, I used AI to create an automation within the Excel itself that also self-audited it. So every month the automations are working. I’m not using AI on a monthly basis. It’s the Excel spreadsheet because we built in the logic and the formulas to make it work. That improved the process dramatically. Now, Daniel, what are some practical ways you’re seeing automation and AI in the manufacturing world?
[20:46] Daniel Shorstein: That’s a good point. You hear AI and you might think of all these exotic things like flying robots, but what I see in the real world, especially in manufacturing, is exactly what you described, Mike. It’s a little bit boring, a little bit nerdy, but really impactful from a business standpoint. The process I really like to take, which has been working well for me and the folks I’ve worked with, is taking an existing process and starting with just a conversation, using AI to help document that process and turn it into a diagram that shows each step. One of my favorite things to call out is that a lot of people think AI is only going to be useful when it’s baked into a process. But in a lot of cases, AI’s most powerful role can be helping you analyze the process and figure out how to keep a working process working, and only change the little pieces of it, either through fixing formulas or updating a template. You might even end with a process that has no AI in it at all, like your updated Excel process. You used AI to make it better and validate it.
[22:24] Daniel Shorstein: Where I’d like to see things go is use AI to help document the current process and then identify what in there can be automated using traditional automation, and what might benefit from AI. Then start working with people to determine whether it makes sense to automate each piece. The risk versus reward. You might not want to automate all pieces, but you could automate a few and add AI to a little bit of it. That to me is one of the most common and safest ways to start implementing that combination of AI and automation into a process without completely disrupting and overhauling it.
[22:55] Mike Sibley: And I think we’ve said it before. Some of it’s back to basics. Whether that’s your data integrity or even what you just said, you might end up with a process that didn’t even have AI in it, but AI helped you get there into a more efficient process. So it’s not always some really complicated or expensive solution. It’s just using the tools in a better way to come to a better solution than what you had previously. So beyond efficiency, Daniel, what else would you say AI can help best improve within a manufacturing organization?
[24:21] Daniel Shorstein: I’ll give my thoughts but I also want to hear from Mike. To me, besides efficiency, there are probably two or three things I see as significantly useful. One is around strategic planning. Looking at the current state of where you’re at. If you’re the leader of a business or an investor, you have the ability to drive decision-making. AI is really good at helping with this piece. You can give it your financial statements, any analyses that have been done, all of the issues you’ve been seeing. You can even just give it the big questions you’re asking, like “Why does my margin keep shrinking? I don’t even know where to start.” A lot of that stuff that was difficult or might take big projects to answer, you can now unlock at least the strategic aspects of it with AI helping you. It’s a lot better if you have the best model available. The paid ones with higher capabilities will have a better understanding of how to solve these things. And then providing it with just the right amount of data and the right context. If you think about talking to AI like you might talk to a professional who’s been in the same industry for 30 years, think about giving them that same level of context. It can be really helpful from a strategy standpoint.
[25:36] Daniel Shorstein: The other thing is just having it help you get insights into your data. If you have a set of dashboards or metrics you’re looking at and you’re trying to interpret them, you can use AI to unlock certain things about that data, whether at the high level or the detailed level or a combination. A tool that Mike and I both love using is Claude or ChatGPT inside Excel. If you have a combination of dashboards at the high level and then the underlying data behind it, that’s a great place to start asking questions like, “Why does this set of sales start going up here and what’s the driver? What are some things I might be able to do to improve my margins with this product line? Come up with some scenario planning for me.” It can start spinning all of that up. You just want to make sure it’s accurate, but at least it gives you a starting point that would otherwise take you hours or weeks to put together.
[26:51] Mike Sibley: Those are great points and I’ll build off of that a little bit. One is the idea of using technology not only on the efficiency side, but to be able to handle more. What we hear from a lot of manufacturers now is that capacity is at a premium. People are busy. It’s hard to spend time on strategic planning and things of that nature. Kevin and I often talk about how to build value in a business. You need to be able to work on the business and not always in it. AI is a great brainstorming tool, but it won’t replace the natural experience that you bring. You still have to make those decisions. At the end of the day, Danny, as you and I were just talking about before the show, AI can sometimes tell you what you want to hear, so you have to be really careful about that.
[28:14] Mike Sibley: The other side of this is effectiveness. A really good example of that is a sales team that sets up a profile to help them practice a pitch before they ever go to the customer and work through that pitch. It can make you more effective as a training tool because you can talk to ChatGPT and it will talk back to you, whether you want to hear it or not. So there’s the leverage, there’s effectiveness, and all these other things Daniel went into. It’s a matter of learning to understand how to do these things, and I think that’s where manufacturers are still in the dabble mode. Our goal, Kevin’s, mine, and Daniel’s, is to bring that along and use the power of it.
[29:01] Mike Sibley: If we switch gears for a second, Daniel, one of the things you and I talked about a few weeks ago, going back into the framework once you have your governance and security and you’re upskilling people: you’ve got data in ERP systems, your equipment has data, your CRM has data. We’ve got all of this data but it’s not necessarily usable where it sits because we’ve got to get it out and place it somewhere. I call that data centralization, and I think that’s part of where automation comes from and kind of what you help companies do, putting it in a place that automation or AI can do something with it. Feel free to correct me where I’m wrong, but that’s kind of my next line of thinking.
[29:59] Daniel Shorstein: That’s almost exactly where I’m going with the thought process. If you think about data similar to how you might think about raw materials from a manufacturing process, you might have a lot of raw materials that could be really useful and build a lot of really powerful products that give you a lot of value. You also might have a lot of scrap materials mixed in. It’s important to understand the quality of your data and the type of data, and what it can be used for. Then going back to first principles: what are the things that you need to know that you don’t know today? What information are you trying to get or what questions are you trying to answer that you can’t? Use that as the driver rather than just saying, “Let me collect all my data and create a bunch of charts because that would be cool.” To what end? As a business leader in manufacturing, you have a lot of questions that are important to answer in order to make business decisions and grow. The data is going to be critical to doing that. But you have to know what questions you have first. Then you can use that to drive the collection of the data and start putting it into a warehouse, structuring it in a way that you can test and reconcile it and start using it for reporting on a daily, weekly, and monthly basis. That way you have reliable answers to those questions.
[31:47] Mike Sibley: In terms of the quality of the data and what data you want, because all those different sources don’t always make it easy to find the data set you want or to get it out on an automated or regular routine schedule. There’s some work that goes into that. Am I describing that right?
[32:14] Daniel Shorstein: Oh, 100%. Even when you think you’re looking at the right data, we do this all the time. You look at a dashboard that says sales and then you look at another report that says sales and you’re like, “Why don’t these numbers tie?” Oh, well, that one’s net of discounts and that one excludes e-commerce because we have a different space that shows up on that dashboard. All of a sudden you start seeing the complexities around what seemed like a simple number, but the devil’s in the details. Whoever wrote the underlying query made the decision, maybe without talking to somebody about which fields should be included and which nuances should be added. That’s where it’s important to have a data strategy and some architecture and documentation, so that whoever is trying to build these things for you, or if you’re trying to build it yourself, you have some consistency and documentation around the definitions of what you’re even trying to look at. That way you’re not using numbers that are slightly close to correct but not exactly what you need to make decisions.
[33:17] Mike Sibley: Right. In the manufacturing world, that’s called the inventory issue. Every single one of them we’ve dealt with to some extent. Either a lack of clarity, or like you said, “Oh, well, that one over there includes this or that, or that doesn’t include.” You start having exceptions as opposed to a consistent truth that’s repeated throughout financial systems, throughout inventory management systems, throughout your organization. That goes back to your point, Daniel, and then allows you to make proactive decisions about where you’re heading and what you’re doing and why, as opposed to reactive decisions based on maybe half-correct information.
[34:28] Mike Sibley: The idea in my mind around everything we were just talking about is getting the right data out of the system at the right intervals to create those summaries, whether you want to call it a dashboard or whatever, in a way that as a leader I can look at this data and make decisions minute by minute, daily, weekly, monthly. Because the old way is you get these KPIs and you might update them once a month or once a quarter. You get financial statements, hopefully 15 to 20 days after month end. But by then, how much time has passed? How much product has gone through? How many problems? The idea here is taking all of that and putting it in front of your face at a moment’s notice with a quick refresh if it’s set up right. Data and data automation is a premium and a value add for a company.
[35:52] Daniel Shorstein: 100% agree, and that really should be the foundational table stakes for anyone who wants to grow the business. If you don’t have access to reliable data that you can trust, that you can use to understand how you’re doing, where you’re not doing well, and what the impact is on decisions that you’re making, that’s really tough. If it takes you a month or two months to see if a decision you made to offshore one of the parts you’re manufacturing was a mistake, and you don’t get the data around that until it’s too late to pull back, that can have a pretty significant impact on your bottom line. After you decide which data you need, that really is the next step in the process, and it usually ends up being a lot harder than most people think to get it right, especially when it comes to validating that it’s the right data and the right formulas and the right fields. What I usually suggest is start simple. Start with one or two pieces of information that you can test heavily over a period of time and get very comfortable and confident that it is the right information. Once you build those fast pipelines of gathering and presenting data, you want to be sure it’s good and reliable. Then you can move on to the next one. If you build it too fast, you might end up in a situation where you’re getting information fast, but if it’s the wrong information, it’s going to cause problems. Like, “I think my margin’s 40% but it ends up only being 20% because I forgot to pull that one field in.”
[37:29] Mike Sibley: Yeah, that’s exactly it. Kevin and I have both seen lately with some prospective clients who made decisions based on bad information that they didn’t realize was bad. That’s a killer for a business. It can cost cash, profits, everything under the sun. Your cautionary approach is the right approach. Make sure you’re getting it right so that you can trust the data.
[38:14] Mike Sibley: So Daniel, we’re kind of coming to the end, but one thing I wanted to ask is: at what point do you think dabbling is no longer going to work? When does this have to become a real part of a manufacturer’s organization before they get left in the dust by their competitors by not adopting some level of automation or AI?
[38:54] Daniel Shorstein: Like most good consultants, I’ll say the answer is it depends. A lot of it comes down to where you’re at in an organization. Are there processes that end up causing so much time and headache that you’re paying a person to do much more of a certain process than you should, and it’s actually impacting your productivity or your bottom line? Are you not getting access to the data you need, and it’s so bad that it’s actually causing you to not be able to grow your business the way you want to? Think about it from the standpoint of whether it’s a true business pain. Are you not able to run your business, grow your business, or strategize the way you want to, and it’s actually that bad? Or is it just that you feel you should be doing more because somebody else is? It’s important to always understand what the capabilities are and to always be trying to move in the direction of adding some AI capabilities. But let the pains, let the issues, really drive the need and drive that next step. Just know what those next steps are when the time is right.
[40:04] Mike Sibley: That’s great advice, and the only thing I would add is go back to the beginning. If you don’t know where to start, you can always come to us. We can help. But the first step is understanding security and governance, then walking into the training, then walking through those pains that Daniel just talked about, and walking through those processes and just getting some comfort level around that. Daniel, thank you very much for being on today. Any last bit of advice before we jump off?
[40:33] Daniel Shorstein: Last thing I’ll say is just unlock that curiosity. Be creative, have fun, listen to some of those podcasts. If you don’t know how good AI is and you aren’t thinking about all the amazing things it can do, you might not have actually seen the latest capabilities of the latest models, because it’s still blowing my mind and I hope everybody else at least gets to experience that.
[41:06] Mike Sibley: Daniel, again, thank you. Appreciate you for sharing all your insights. It’s a fun conversation. We may have to have quite a few more of these because AI changes by the minute, it seems. And for everybody listening, thank you. As always, reach out with comments, questions, and things we can help you with. Have a great day.
[41:28] Outro: To learn more about James Moore and Company’s manufacturing services, go to jmco.com. Don’t forget to subscribe to our Moore on Manufacturing series to receive updates when new videos and podcasts are released. If you’d like to be a guest or if there’s a topic you’d like to see covered on a future episode, contact us on our website. You can also follow us on social media for more news as manufacturing continues to evolve.
If your manufacturing business is ready to move beyond dabbling and build a smarter approach to automation and AI, explore more episodes and resources from the Moore on Manufacturing series.
All content provided in this article is for informational purposes only. Matters discussed in this article are subject to change. For up-to-date information on this subject please contact a James Moore professional. James Moore will not be held responsible for any claim, loss, damage or inconvenience caused as a result of any information within these pages or any information accessed through this site.
