Main content

IFS Connect 2023 - why talk about generative AI when robotics and connected manufacturing are ready for prime time?

Jon Reed Profile picture for user jreed June 16, 2023
IFS avoids technology festivals in favor of industry shop talk. But at IFS Connect, I found customers knee deep in AI preparations. It's all about getting your plant floor machines talking. And, for CDF Corporation, robotics are already in action.

Alex Ivkovic - CDF - CIO - at IFS
(Alex Ivkovic of CDF Corp at IFS Connect )

You can't attend an enterprise show without hearing extensive pronouncements about generative AI from the keynote stage. Well, except for IFS Connect North America.

Not that IFS didn't talk about emerging technology - their "asset health" demonstration captured attendees' attention, and mine as well.

Why? Because the goal is to solve industry problems, not win a science fair. If tech makes that happen, bring on the tech - but generative AI use cases aren't ready for enterprise prime time yet. Besides, that's not where you start. Your AI is never going to be better than your company's data platform - that goes for any type of applied AI, generative AI included. 

IFS customers - enterprise AI needs plant floor data

That means your manufacturing equipment better be able to communicate with the rest of your IT systems. If you're going to do "AI" in the asset-intensive industries IFS serves, you simply can't overlook the most expensive equipment on your plant floor - and many of those machines have years on the job, with no business case to replace.

IFS customer KLN Family Brands is a revealing example. As KLN Family Brands' Drew Johnson told me, they are definitely looking ahead to partnering with IFS on AI-based initiatives. But Johnson's team knows, you have to start with getting sensors on your equipment, and your data platform in place. As he told me:

I mentioned our MES and SCADA project that we're working on. That'll give us a lot of data points into our equipment, as far as our extruders and different manufacturing equipment. One thing we liked that we saw here is that we can feed that data in, and AI will say, 'Hey, something doesn't look right here,' once it starts to recognize what's good, and what's bad for that equipment. That will help us be a lot more proactive.

As far as keeping our equipment running, AI can provide that to our maintenance teams, and create work orders to get things fixed and replace before we have downtime.

The "we're doing generative AI too!" vendor hype festival has a way of obscuring advances in more mature areas of enterprise AI. IFS, for example, has deep AI pursuits under its belt, including a mature Planning and Scheduling Optimization (PSO) engine that is a sneaky big story, because it can be utilized by service organizations in many industries - even by IFS competitors. Robotics is also overlooked these days, but conquering physical movement is one of AI's biggest challenges/opportunities.

IFS robotics in action - the CDF Corporation story

Talk about generative AI all you want, but plant managers everywhere might prefer to talk robots. Finding the right mix between effective robotics and ever-scarce human labor is a shop floor obsession.

Cheer Park robotics with IFS
(Cheer Park's IFS robotics project in action)

It was no different talking to Alex Ivkovic, Chief Information Officer at CDF Corporation. He is excited about generative AI possibilities, but he had plenty to say about the robotics partnership with IFS. 

The result? Robots in action, on the plant floor in CDF Corporation's Cheer Pack North American facility in West Bridgewater, Massachusetts (Cheer Pack is a partnership between several companies including CDF Corp).

You can see the heavy payloads this robot can handle above. As Ivkovic says in an IFS video on the project:

We expect to save over $1.5 million per year. In addition, every employee affected will be retasked into higher skilled positions, greatly helping us with our labor shortage.

 How did CDF Corporation get to operational robotics? As IFS explained:

Cheerpack North America wanted to create a competitive advantage and derive efficiency in their production. They needed a partner to help them create capability and value in Process Automation. For this, they turned to IFS, and the IFS labs team went to work.

For Ivkovic's team, this all started at an IFS event:

We've long been discussing a robotic material handling project. After seeing the IFS Labs demo at the World Conference, it seemed to make a lot of sense to have IFS applications orchestrate our robots [Author's note: CDF Corporation has been running on IFS applications for fifteen years].

That paid off:

The robots we put in the plant that are run by IFS save a tremendous amount of time.

My only objection to these types of robots-in-action videos? They make robotics look easy. Even in a fairly-controlled setting like a plant floor, robots have limitations in adapting to unpredictable variables - something that Chris Middleton has documented extensively on diginomica. The consequences can lead to safety issues, or failed pilot projects without meaningful ROI. But as Ivkovic told me, this is not the case for these robots - though a miscue from time to time is expected:

We do get [miscues] occasionally. But they're autonomous - they can adapt to a changing plant map. We're down right now, because they ripped out walls and they're expanding the plant, and there's no way they could survive that. But in the day-to-day, they are traffic-aware. They'll cross each other, and know that they're seeing each other.

Applying AI to connected manufacturing

Given this all sparked from an IFS event, it's no surprise: Ivkovic is a regular at IFS shows. So what did he think about how IFS addressed AI? Ivkovic said he is drawn to the link IFS made between AI and connected manufacturing. He referenced Six Sigma:

You have your upper control and your lower control limit, but there's also patterns in between, and they talk about it - but no one has the manpower to actually look for these patterns. Let's face reality, right? 

So a lot of the data goes into the database, and we'd look later when there's a problem. But I'm really excited about the aspect of AI where it could find these patterns and notify people, and predict problems. And maybe looking at history: when we saw this pattern before, this had to be replaced. I think there's a lot of possibility to tie things together.

IFS customers may not know what they'll do with AI next, but they know this: they have to get their machines talking.

We've been doing some stuff, even with our older machines. We'll put a really cheap PLC on a machine, with just a bunch of sensors. It's not actually running the machine, we're just using it as a sensor platform, so our SCADA system can talk to the machines, and get this data. The other thing we're looking at: there's a product called Crosser, which will allow disparate systems - or in our case, we're very heavily invested in SCADA - to talk directly to IFS, and to transfer the data back and forth.

These days, automating plants is not about getting rid of talent. It's about finding and keeping it:

It took me six months to find a programmer for my SCADA system - and I was looking hard.

IFS Cloud - two customer reactions

If IFS could pick the headline for this piece, it might be IFS Cloud. IFS Cloud is the go-to-release, the core of IFS' "evergreen" strategy. It certainly got plenty of attention from the keynote stage (As per IFS, 400 customers are running on IFS Cloud now). Indications are that the evergreen approach is working. All customers that initially moved onto the first IFS Cloud release (21R1) have now upgraded again.

"IFS Cloud" can be a confusing product title, because customers can actually deploy IFS cloud on-premise, or in their own hosted environment, as well as on Azure as an IFS service. Though I was not able to do an in-depth interview with an IFS Cloud customer at IFS Connect, you can count CDF Corporation and KLN Family Brands as two more companies planning to make the IFS Cloud move.

The details of IFS Cloud's node-based architecture, which enables multiple deployment scenarios, are beyond the scope of this piece - but IFS firmly believes that they have a modern answer for their customers, many of which are in heavily regulated industries, and might need their operational software on-premise. Yet these same customers don't want to miss out on the benefits of modern software, from embedded artificial intelligence to ease of upgrades.

Moving from older versions of IFS onto IFS Cloud is not trivial. There is a new UI to get employees up to speed on. Prior code customizations must be evaluated, with an eye towards moving closer to standard, or shifting to custom fields. Technical upgrades have limited value - and we all know it. Instead, IFS encourages customers to focus on the business results from the IFS Cloud move; IFS' Digital Business Value Assessment (DBVA) solution, which I've written about before, provides an underrated framework for mapping out where the value will be found, and how the path to further ROI will be achieved, beyond go-live.

IFS pitches IFS Cloud as a "composable" platform. One aspect of that: customers can buy components separately, including IFS's fast-growing Service Management solution , as well as IFS EAM (Enterprise Asset Management). As for Ivkovic, he's got one advantage for his IFS Cloud move: no customizations. Yes, you read that right - after 15 years of running on IFS, CDF Corporation has entirely avoided customizations. Well, with one exception:

My one modification is actually the robots. And that's written by IFS Labs. So it's not like that give me a hard time upgrading that.

Avoiding modifications has made selling upgrades to CDF Corporation management easier. Therefore,  Ivkovic says he didn't have any trouble getting management to green light IFS Cloud (they plan to go live in November).

The wrap - generative AI scenarios for plant managers

As for generative AI, Ivkovic has all kinds of ideas for how it could be applied. Here's one interesting scenario. As expected, it goes back to making the most of machine-generated data:

Picture my tech at his machine. He pushes his button and says, 'We've got an electrical problem.' Now the AI is working. It's going to send the technician; the proper technicians go. But in the meantime, the AI is working in the background, and it says, 'Hmm, the last three times that type of machine had an electrical problem, we saw this pattern.'

The next time it sees that pattern, it puts a ticket in before the guy pushes the button. It says line 23, predictive maintenance. We're seeing this pattern, right. The last three times we saw that pattern, it was this problem, and not only just an electrical problem - maybe it knows the exact problem.

So what did we just save timewise, right?  Let's say it's a major repair. Now it sends it to the ERP the scheduling engine you're talking about. And it says there's an 80% probability it's going to be down for two months, or two weeks. And it sees what that does to the shop floor orders. Maybe it lets customer service know that there's a probability that their order might be late. I mean, you could tie it into all this disparate stuff.

Some of the tech to pull this scenario off is already there. The rest is not far behind (Ivkovic sees generative AI helping with communications in these scenarios, e.g. from shop floor engineers who often don't excel at the writing side of things). It will be interesting to see how IFS customers fare. If they move like these two customers did to get their data house in order, I like their chances.

A grey colored placeholder image