In this episode of Great Question: A Manufacturing Podcast, Smart Industry's Scott Achelpohl spoke with ARC Advisory Group research lead Inderpreet Shoker about data and advice on how to use and implement technologies successfully in smart factories. 

The conversation was part of Shoker's recent three-part Smart Industry series on connected worker platforms. In Shoker's first article, she talked about the increasing complexity of manufacturing, including the division of knowledge across different systems.

Part 1: Understanding connected worker solutions: Expectations vs. reality 

Part 2: The AI layer in connected worker solutions: Promise and practicality 

Part 3: From pilot to plantwide value: How to scale connected worker solutions

She also pointed out that connected worker solutions offer ways to digitize this knowledge and link the assets and systems employees need to do their jobs. 

In her second article, Shoker delved into the AI layer of connected worker solutions, and how AI can be used to extend connected worker capabilities.  

In this episode of Great Question, Shoker expanded on these themes and spoke about how manufacturers can successfully implement these emerging technologies in connected worker settings.

See also: Smart Industry's annual State of Initiative Survey is open!   

Below is a partial excerpt from the podcast: 

Scott Achelpohl: What prep would you recommend an organization do before asking their employees to interface? There's an IT word for you with these solutions. And here's another word for you. We talk a lot about data. What kind of shape does their data need to be? 

Inderpreet Shoker: Scott, there is one thing that I always talk about whenever it comes to getting ready for any kind of technology adoption and not just, you know, connected worker technologies. I think a lot of the times end users are looking at a technology and then take that as a first step. I think the first step should be to start off with a problem. Look where are the issues in your plant or in your facility, [ask] what are these problems, and then dissect these problems and see what are the technologies that are out there that can help you solve these problems? I think instead of thinking of technology first, the first thing I think we need should do is to think of the problem first. It's also important to work with the frontline workers, with the plant workers, with the operators, with technicians, supervisors, subject matter experts, because these are the people who know where do the problems exist. What are the processes that slow down the whole plant? What are the variations from shift to shift? What are the areas where we rely too much on one or two people for their expertise? These are the different things that we should think before we think of technology. 

Podcast: Scaling AI won't fix a shaky ERP foundation  

Another thing I often hear from end users is that they think of the problem, they bring in the technology, and then they never paid attention to change management or training. These are some of the areas we, a lot of times, forget about. So start focusing on these beforehand as well. Don't think if you have a successful pilot when you are scaling and after that, we'll think of training and change and think of these things beforehand.  

Scott, you mentioned about data. That's also another question we often get from end users that, okay, I want to start with this technology investment, but my data is not ready. A lot of time what we hear from forward-looking end users who have been successful with technology implementation is that there is never a good time to start with the technology thinking that my data is ready now. There'll probably never be a time where your data is 100% ready. And the good thing about these solution is that you can start off even when your data is not perfect. But yes, definitely do pay attention to some of the basics. So make sure that your procedures are current, make sure that your data is clean, you are consistent with asset names, asset tags, wherever the information needs to be linked, it's linked to the right sources. Clean up the duplication.  

See also: Physical AI: Where opportunity really sits—now, today—for manufacturers  

SA: How can connected worker solutions utilize AI to kind of what you also talked about in your story to go beyond chatbots for the worker standing at a plant workstation, for example. Just tell us what more is there than AI dashboards? 

IS: A lot can be done with AI. I think chatbots are becoming very, very common. So that's actually a good thing that end users are having experience with chatbots and also realizing how useful these can be. But again, having just an interface to do your searches is one thing, but to get meaningful result, I think it depends upon how strong your data foundation is, how well you have contextualized your data.  

The goal should be that we want to give our workers the right information at the right time, answers that they can trust. Definitely there's an opportunity for chatbots to become more than just chatbots, to be intelligent assistants to the workers out if they're in the field in the plant. A lot of end users are focusing in that area and as well as a lot of solution providers are trying to improve the quality of the chatbots. So that's an area where there's ongoing improvement.  

Another area I think where AI is proving to be very helpful is to help capture worker expertise. Some tools, you know, and connected with worker AI have been used for a long time now. You know, we see a lot of capabilities around computer vision, voice tool that captures if a worker is saving any data using voice tools, which is very helpful in the field because a lot of time workers are using both of their hands. To fill forms, to input any information, voice tools are very helpful. Computer vision is very helpful to record equipment conditions, what is going on, to record procedures, processes as well. But what now AI has done, primarily generative AI, is converting all this information that is in your audio form or your video form to actually convert that into knowledge.  

See also: Survey: AI adoption continues apace, despite uncertainties and failure risk   

[...] Another good thing about AI is that now technology providers are embedding it into their existing solutions. It's not like end users have to learn something completely new. It's already there within the existing solution that you already have in your plant. Scott, I think it's clear from end users as well that they want something just beyond AI chat bots.