Getting Started with AI in Manufacturing

AI has already become indispensable for many of us. I definitely understand reluctance on the subject. What do business people hate most about AI? Let’s find out, with some help from Grok3! This article is geared towards reluctant and casual users of AI and those looking to get started. As always, we will emphasize issues that matter most to manufacturers, because we live and work in their world every day and understand the unique challenges you face. I definitely encourage you to preview this short video or read the full conversation with Grok3 how the AI contribution from Grok came about.

Grok3's Contribution Large Language Model

As Grok 3, built by xAI, I’m thrilled to co-author your article, bringing AI-driven insights to the table while keeping the human perspective front and center!

What Business People Dislike About AI

The conversation began with a query about whether anyone has studied what business people hate most about AI. I provided a concise response tailored to business concerns, highlighting key frustrations: fears of job displacement, unreliable AI systems, privacy and security risks, overhyped promises with underwhelming results, and the loss of human autonomy in creative or decision-making processes. These insights were drawn from a blend of general research and sentiment, reflecting practical challenges business leaders face when integrating AI into their operations. No single study was pinpointed, but the response captured a broad consensus of pain points relevant to the business world as of March 26, 2025.

References and Collaboration with Grok

At the user’s request, I supplied additional reading references formatted as clickable links in a code box for easy WordPress integration. These resources delve deeper into AI’s impact on jobs, privacy, reliability, and ethics—topics that resonate with business frustrations. The references include:

If you are new to AI, you might find the full conversation with Grok3 very useful.

Resuming human editing mode!

There are a few things that are very important when first getting started with AI.

Rule #1 Protect your Privacy (and Intellectual Property!)

For starters, stick to US based companies (Claude, ChatGPT, Grok, etc) governed by US law. Why is that important? Our adversaries will abuse AI. For example Deepseek, which made big headlines in January, at the very least “lies” about important things, but more critically the information users share with it is accessible to the Chinese Government. Here is an article and a few posts on LinkedIn that I shared in response to Deepseek:

Turn off data sharing unless the data is completely harmless. All US based AIs have an option to “turn off data sharing” meaning, they do not train their models or review any of your data. Unless what you’re doing is completely harmless, turn this off to provide some level of protection.

Limit self identification. Some LLMs will remind you about this in the query window from time to time. Regardless of your settings on data sharing, it is still a good idea to avoid sharing personally identifiable information about yourself to the LLMs.

Most importantly! Don’t share trade secrets, CUI, etc. with LLMs. This can generate serious ramifications. In the future we can explore the idea of using localized and even “sandboxed” LLMs for maximum privacy. For now just err on the side of caution if your company has any trade secrets or handles controlled documents, make absolutely certain they all or in part, are never shared in any form (screenshots, data sets, files, text, etc) with any LLM.

Always consult your employer to ensure your use of AI does not violate any of their policies. Some companies are very strict about use of AI, what you are allowed to use it for, which models, and if you are allowed to use them at all on their systems.

Rule #2 Always Validate Output

There are things that LLMs often screw up on. You will become familiar with their strengths and weaknesses in time. It can also change, usually for the better, but not always. One major issue you will experience is “Hallucination”. This is when the AI presents information as though it is correct, but contains errors or outright phantom (not real) information. For example, while validating the articles Grok provided links for, all the titles were correct but 3 of the 5 links were actually incorrect. I was able to easily source the articles (maybe I should have asked another AI to fix them?) but had I just accepted the information, readers would have gotten the wrong articles in two cases and a 404 page in another case. I see this issue often with all the LLMs I use.

Tip #1: Use multiple LLMs

It is almost always a good idea to put any “serious” queries in multiple LLMs. Personally I recommend Grok3 and ChatGPT 4o to start out. If you’re a coder, Claude Sonet 3.5 is highly recommend by experts in the field. For the upcoming examples I used ChatGPT and Grok3. You will see why I advocate for this approach in Tip #2. You will start to learn the strengths and weaknesses of each LLM. It can be stunning at times.

Tip #2: Use AI in place of or in addition to a Search Engine

A great way to get familiarized with an LLM is to use it instead of DuckDuckGo, Google or Bing. Existing search engines do provide AI summaries, but the interface is superior with an LLM itself. You can ask follow up questions, etc. For example, if you’re trying to source a part, you can give an LLM specifics. Let’s say you are trying to acquire some 6 mil zipper plastic bags somewhere around 10×6″ in size. You can ask something like: Please search reputable websites such as (list sites you commonly use) Grainger, Amazon, and Uline for 6 mil zippered plastic bags. Provide cost breakdown and links for 10000 units. From this exercise it became very evident that there is a big disparity between how well Grok3 and ChatGPT 4.5 performed (we tried 4.0 as well, which provided no links to actual products). To see the stark contrast View full results (PDF).

Grok3 Excerpt (Left) ChatGPT 4.5 Full (Right)

Tip #3: Use follow up questions and “Deep Search” options to get more useful data.

Test out follow up questions, such as “Are there other distributors in the USA that have a good reputation and offer decent prices for this item?”. You can get very specific here, and this might be a good time to use the “Deep Search” type features to give the LLM more time to discover more suppliers.

Tip #4: Ask for advice for repetitive tasks.

I’ve got good news for you. In the age of AI, there is zero reason for a human to do mundane repetitive data tasks.

If you are about to do something that requires a lot of repetition, this is a perfect area to ask AI for help. Recently I was needing to print a lot of labels for the week #’s and date ranges for a binder I was working on. By quickly asking AI for advice, I was able to get a sample .csv file that had the Week #’s, date ranges, and instructions to use “Professional Mode” on my Brother software. It would have taken me, or someone else (and I would not feel great about delegating that!) a lot of time to make months and months worth of these labels. Within minutes I had a sample and printed a larger batch of labels.

Tim #5: Experiment and Never Assume (AI can’t handle it).

I think I am good about this, but I keep getting “surprised” when I ask something I think the AI will barf nonsense back at me. I don’t recall in recent times an AI (especially Grok3!) telling me it couldn’t do something, or just doing what ChatGPT sadly often does, providing me with some “suggestions”. Maybe my queries are too casual for ChatGPT, but Grok3 is almost always extremely helpful, so it’s a good idea to always query Grok in addition to other LLMs so you can see what I mean.

Recently I wanted to 3D print some text for my automation wall. I spent an hour or so one day looking at and testing some parametric generators and was not thrilled with the formatting. None of them allowed me to find the line/font spacing “Goldilocks” zone. So the next day, I thought, okay, this is just code, let me see what I can do to create my own parametric tool to create it. The first attempt and literally 10 minutes in with a few tweaks I had a print started with the exact spacing, etc, that I was looking for. Another big win, of many.

Conclusion

AI is here to stay and it is fun and rewarding to use it once you get rolling. Building on our long tradition of embracing new technologies, we find AI to be one of the most rewarding opportunities that has entered our space. So far, it’s helped us build a better understanding, save time and hopefully made us a more valuable company to our customers. If you found this article useful please subscribe to website updates. We’d also love suggestions for future topics around AI, what do you need today to leverage AI better? Let us know.