Artificial intelligence is already changing how manufacturers manage quality and compliance. The difference between success and frustration often comes down to how AI is applied. When implemented with intention, guardrails, and a clear return on investment, AI can remove friction from quality systems instead of adding complexity.
At Ledge, Inc., AI is used as a practical tool to support people, strengthen systems, and improve consistency across operations. The focus is not experimentation for its own sake, but real workflows that manufacturers can deploy and trust.
Where AI Is Already Delivering Value
Many manufacturers are under pressure to do more with fewer resources. Purchasing, quality, engineering, and sales teams are often stretched thin, and critical reviews can get rushed or skipped entirely.
One of the earliest and most impactful uses of AI has been contract and requirement review. By comparing incoming customer contracts against quality manuals, supplier requirements, and customer-specific obligations, AI can quickly highlight gaps that might otherwise be missed. This helps teams understand risk earlier in the process without consuming hours of valuable time on bids that may never be awarded.
The result is faster decision-making and better visibility into compliance requirements before work begins.
Turning Repetitive Tasks Into Repeatable Workflow
A key principle behind Ledge’s AI approach is repeatability. Rather than one-off prompts, workflows are designed to be reused consistently with new inputs.
This approach has led to measurable gains in areas such as safety data sheet generation. For organizations producing complex SDS documentation, AI tools allow users to input chemical data, CAS numbers, and percentages to generate a draft document. Engineers and chemists then review and finalize the result.
Starting a task nearly complete is far more efficient than starting from a blank page. This shift alone can save hours per document while improving consistency.
Supporting Inspection and Verification Where Expertise Is Limited
Another area where AI has proven valuable is inspection support. In welding and fabrication environments, experienced inspectors can quickly identify defects, but not every organization has that expertise in house.
AI-based weld analysis tools provide directional insight by analyzing images of suspect welds. These tools are not a replacement for skilled inspectors, but they help teams decide whether further review or corrective action is needed.
The same principle applies to material certification review. Material certifications are often assumed to be correct, yet errors are more common than many expect. AI can analyze incoming material certs, even poor-quality PDFs, and compare them against specifications to flag potential issues immediately.
This allows receiving teams to identify risks without relying solely on engineering resources or tribal knowledge.
Adapting Faster as Teams and Requirements Change
As organizations grow and change, keeping people aligned on revisions and specifications becomes more difficult. Training tools instead of retraining entire teams can be more effective.
When specifications or formulations change, AI workflows can be updated once and applied consistently going forward. This reduces the risk of outdated knowledge and helps maintain compliance even as personnel or processes evolve.
AI as a Support Tool, Not a Replacement
Ledge’s perspective on AI is clear. AI is meant to support skilled workers, not replace them. Its greatest value lies in capturing knowledge, reducing low-value tasks, and helping teams focus on work that requires judgment and experience.
AI can help:
Capture institutional knowledge from experienced workers
Support faster onboarding and training
Reduce dependency on informal processes
Free up time for higher-value problem solving
The long-term challenge is not eliminating junior roles, but helping people progress faster by removing unnecessary friction from their work.
Security, Control, and the Right Tool for the Job
Security remains one of the biggest concerns around AI adoption. Ledge addresses this by deploying AI tools that are intentionally designed not to learn from client data. This closed-system approach reduces intellectual property risk and gives organizations more confidence in how their information is handled.
Another critical factor is model selection. Different tasks require different strengths. By providing access to multiple models, teams can choose the right tool for each job rather than forcing every workflow into a single solution.
Building a Platform, Not a One-Size Solution
AI at Ledge is treated as a platform rather than a packaged product. Leadership provides access and guardrails, while teams identify where help is needed and build solutions around real pain points.
A common analogy is a 3D printer. The value is not in the machine itself, but in what users choose to create with it. With the right policies and support, teams can develop solutions that directly address their operational challenges.
Practical AI for Quality and Compliance
AI in manufacturing does not need to be overwhelming. When applied thoughtfully, it can strengthen quality systems, improve compliance, and help teams operate more efficiently without sacrificing control.
By focusing on security, repeatability, and real-world workflows, Ledge is helping manufacturers use AI in ways that deliver measurable value today while supporting long-term growth and resilience.
Video Transcript
Good afternoon and welcome to another video that we’re putting on today. I’m Adam Marsh from Ledge Inc. And I’ve got one of my teammates here, Chad Yankst. Today we’re going to do an introduction or a discussion around how we’re using AI and sort of where we’re going with that. Chad has a couple questions he thought he’d kind of send my way. We’re going to bounce back and forth, talk about them, but we’re diving into what are we doing? What are our customers using it for and how are we making sure that it works for
Them? Yeah, thanks Adam. So what’s one way that AI is already helping roles like purchasing, sales, engineering, and today at our clients?
So we’ve really seen opportunities within our clients to help with things like contract review. And so as customers are getting new contracts in, we’re able to drop in things like their quality manual and the supplier requirement or the customer requirements and compare them really quickly. So imagine on the front end of a project, being able to ask the customer to have a review completely done that looks at all those connections, looks for things you may be missing. Think about record retention requirements or quality requirements that you didn’t know you had. Some of these documents can be really long and hard to get through, and so your estimators are wasting a lot of time going through ones on jobs they might not even get or they’re not doing that review at all. And so think about what kind of risks do you see if they don’t do that review, right?
Right. Yeah. Yeah. Especially if your win rate’s fairly low, you don’t want to spend a ton of time upfront reviewing contracts and putting all the resources in for expensive employees doing non-value added
Work. Exactly. So we’ve built a couple tools for folks doing that and then we can really specify what that tool’s looking for and customize it based on the client. But the idea is how do we use generative AI and turn it on in a way that is repeatable? So we’ve built workflows that are repeatable. We know what we’re looking for. We continue to use that same prompt on a new input and it’s been really effective.
Great. So what are some real examples of workflows that we’ve built that have helped make a measurable impact for our
Clients? So great question. We have a couple clients that want one who has to generate some really interesting safety data sheet reports and it can take a couple hours to do this based on what they’ve designed. So what we’ve been able to do is actually allow them to input chemicals that they’re using, input the CAS number and a percentage and let AI build that safety data sheet, identify things it might not be sure about, and then let the engineer or the chemist go ahead and review that and complete it. So think about starting a project 90% of the way done is a heck of a lot easier than starting it with that blank sheet. And so we’re able to allow people to dive in and do those kind of things. We just built and tested a weld analysis tool. And is it as good as your best welder or your best weld inspector looking at a weld?
No. But does it send you in the right direction? That’s what we’re looking for, right?
Yeah.
And I check some of my own welds and I fail miserable.
Yeah. I was going to ask how you did on
That. Not
So good. Another benefit to a weld inspector tool is if you’re a fab shop, you have people who know what a bad weld looks like. If you’re purchasing from a fab shop, you may not. You don’t have a weld inspector, you don’t have those welders with that knowledge. And so this gives you the ability to take a picture of a suspect weld, what’s wrong with it?
Yeah.
Do I need to send this back to my fabricator or not?
Yeah. And so it’s a really interesting point. I would say the other area we’re doing that where they don’t have experts is on the chemical and physical properties. And so when material comes in, are you doing a review of that material cert to say it meets the spec, it meets what we need it to meet. And so we can quickly grab those PDFs, which by the way, are the worst PDFs in the world. I mean, it’s like a copy of a copy of a copy- From 1965. And so we’re able to take that and analyze it. AI is really good with object recognition and compare it to the specification and quickly do that. So imagine your receiving person who may not be qualified to look at things like chemical properties or physical properties and compare them. And so they’re looking at it, not great, or we’re spending our engineer’s time to look at a material cert when it comes in, which could be a waste.
Instead, run it through an AI check, get a heads up if it’s not good, and then take it from there. We can do that immediately when it comes in.
So I was at a fab shop yesterday and I was talking to them about material cert inspection review and they’re like, “Why would we review material certs?” We told them what we wanted. We paid for material that meets an ASTM standard. Why would we have to review the cert?
I built a tool to do this and I found a certain that we had from a long time ago and threw it in. And the first cert I checked was wrong. The first cert. You want to tell me what the odds of that are? They come in wrong a lot and just like you would inspect a part when it comes in, we should be inspecting that material, but we just haven’t had a great way to consistently do that. This answers that. The other place I think it’s really key is if your team is maybe changing a lot. We have a customer that maybe plans some chemistry, but might change that once in a while. And so they could quickly update the tool rather than updating their people to say, “Hey, remember, we’re on revision B of this now, so you have to look for this makeup.” Instead, the tool is now updated and when they check it, it’s there.
So it’s a lot easier to train the tool than it is to train people.
Yeah. Yeah, that’s true. So how do you see AI tools like ours changing the role of skilled workers over the next few years?
That’s a really good question. Right now I see it impacting sort of the white collar workers more than the skill workers, but I think we have huge opportunities to take learnings from those skilled workers, right? I think we have opportunities to use AI to maybe capture some knowledge from those skilled workers, right? What are they doing, especially as a lot of them are heading towards retirement. So we can train our new workers, identify places … No, there are things we don’t want to do, right? And so if we can use AI to supplement areas that aren’t ideal for humans, then let’s do that. I don’t see it replacing heavy skill. And I would say the best example I have of that is we’ve talked about this a lot where there’s a concern with developers and AI is getting pretty good at writing code, but does AI write great code?
No. So senior developers, they’re in great shape, right? So they’re going to keep making stuff. Junior developers, I don’t know, because I can go to AI to be my junior developer right now. Now, the problem there is how do you become a senior developer?
Yeah.
But I see this across a lot of industries. And so that tells us we probably need to get really good at training because we can’t have people at junior level. We got to get them up to speed and fast.
Yep. So evolve, there’s a million AI tools out there. It’s overwhelming when you look at what you can use. Why this tool? Why our solution?
So we’ve been asked a lot over the past year to help people implement AI. And we’ve been looking for tool sets that kind of answer a lot of the questions. The biggest question we get before we go anywhere is, is this secure? How do we deal with this? And so we’ve been looking for a solution that intentionally does not learn, right? And so that way you can feel safer with your IP. You can start to use it in ways that you might not feel safe using some of the other large language models that have the opportunity for leakage or have the opportunity for sort of that break-in. So we deploy a tool that doesn’t learn intentionally. It’s a closed system, so it’s pulled off of those models. And we also offer multiple models. And that’s really one of the big tricks here is multiple models give you the ability to choose the right model for the task that you’re asking for.
Great example. I’ve been working on evaluating some calibration certificates and it’s about 200 certificates in one file. I can tell you ChatGPT-4 cannot do it. I see that it’s a little lazy. It gets the first three or four, and then it says, “You do the rest.” If I use Gemini, get some all. And so what we start to do is start finding the right tool for the right operation. And so with our tool, you can start testing and saying, “Hey, we’re doing this kind of operation. Let’s use this tool.” And so we give you sort of a broad range for that, and then we give you access across the business. That’s the real key to AI is this is not going to be leadership driving your solution and saying, “Hey, here’s the tool, use it. ” Instead, it’s a leadership giving you access to a tool to build your own solutions and identify those solutions.
So the employees at the company are going to be saying, “I need help with this. ” And now we can help them there. We’re not sort of pushing it at them. We’re saying, “What do you need help with and let’s build it? ” So this is not so much a one size solution. This is a platform that we’re building on. I think you and I have talked about this. I’ve turned it into … AI is like a … I’m selling you a 3D printer. You get to decide what you want to print.
Okay. That’s a pretty good analogy.
So you’re going to design your part, and you can print the stuff that’s online, you can use the stuff that’s already out there, or you can start building your own stuff. So I think that’s where the good place to be at, but do it in a safe way, do it with guardrails and with policy.
Yeah. Yeah. I know. I was showing someone the other day some of the things that you can do with it and they said, “Well, those don’t really apply to us. And what could we do? ” And I said, “Well, what are your pain points? What are some things that are bogging down your skilled employees, your employees that are maxed out, your quality folks, your purchasing, what are those things that we can … If you can take some of that workload off of them, they can get more done and use their brain for something that’s more worthwhile.”
Well, and I think we’re seeing it work across industries. So when we develop tools, we’re not making them for maybe … We’re helping to make it for a customer, but we’re also making it to plant a seed to just get your folks thinking. When I go and start showing them like that, I start to see the light bulb go off in the room of people like, “Oh, could it do this? Could it do this? ” And that just kind of goes, but you got to put the tool in their hands and start showing them what it can do and then they’ll understand what those solutions are.
Yeah.
Okay, great.
Well, thank you very much, Adam.
Well, thank you, Chad. So today, we covered sort of what we’re doing with AI where we see it going, how we’re helping manufacturers get there. I can tell you it’s not just manufacturers. We’re helping across multiple industries, accounting firms, all those kind of things, service companies. And so we’re seeing great opportunity there. There’s some resistance, some concern, but we’ve got some answers to those concerns. We’ve got great options on security, on toolkits, and on how to roll it out to your team. So thank you very much and look forward to our next video.