Beating Back Industrial AI Bloat*

Industrial artificial intelligence (Industrial AI) is a big deal these days, at least according to the response to my prompt in Gemini on the subject. Kidding aside, I’ve been doing quite a bit of listening to and collaborating with others about the many, many perspectives on the subject. I’m lucky enough to interface every day with innovation leaders at energy, manufacturing, and other industrial companies, working through how to leverage the tools, where to invest, and, most importantly, examining the art of possible. Those interactions have informed some perspectives, based on discussing AI with some very smart, competent, forward-thinking people and teams. 

AI is inevitable to varying degrees, based on where and how it is being applied. It also has massive promise, certainly. On the flip side, in industrial settings, it is proving to be very costly to use at scale and when applied cross-collaboratively. Additionally, a consistent and persistent challenge has emerged as people attempt to apply it at scale -- the counterproductive bloat associated with industrial AI. Of course, it encompasses software bloat associated with stranded AI coding, unfettered access to prompting, model growth, and data and workload management, to name a few factors. In addition to adding cost, it also comes with massive resource intensity (e.g., energy, minerals, water, waste), at least currently. Within that lens, below are a few observations and learnings that can help companies move beyond industrial AI bloat and hype to a place where AI provides real and measurable business value. In no particular order.

• Just Because You Can Do Something Doesn’t Mean You Always Should

It’s tempting to apply AI to everything. What that has demonstrated it that just because you can use AI, doesn’t mean you should. It’s clearly now the shiniest of the shiny objects. As with any emergent capability, start with your desired outcomes and then work toward technology, and the benefits are more likely to be realized. There are countless examples of how forms of AI can be incredibly powerful when applied as a toolbox, not an outcome. 

• Focus on Innovation, Not Just Efficiency

Efficiency, optimization, and cost reduction are great (and yes you should tackle them), but they aren’t the market game changers that innovators strive for. If you focus too closely on margins as the outcome, you will get surpassed by someone who is laser-focused on competitive market signals. Also, in reducing one form of process bloat, you are likely just exchanging it for newer forms specific to industrial AI and data. 

• Strive for VOI, Not Just ROI

Because of the cost and complexity, those who focus on ROI versus VOI (value on investment) will hit a ceiling. And it will hurt. VOI is harder but exponentially more valuable. That’s how leaders and innovators approach it. Value is the objective of the  exercise in innovation, supported by these extraordinarily powerful AI tools.

• Be Honest When Defining Value

Value can be seen best through the lens of market signals and competitive excellence. Strategically, the objectives aren’t complex, through realizing them can be: clarity and freedom across the enterprise to make decisions (knowing they are the right ones); latency reduction in those decisions (removing artificial barriers to value); and the ability of the AI to then extend beyond those initial decisions to produce game changing results humans can’t deliver (value-based agency). I was privileged enough over the last decade to watch this process unfold as I worked with some real innovators in Mike Carroll, Ron Norris, Rajib Saha, and their teams as they built out an extremely advanced form of #casualAI. It wasn’t easy all the time, but their mantra was damn the egos, truth mattered more. 

• Understand that Betting the Farm Doesn’t Mean You’ll Win

What I’m typing now doesn’t apply across the board, as I have seen thoughtful, careful and very effective leadership specific to AI. Yes, leadership on the provider side is having a moment, but moments are always temporary, though their impact might not be. I realize software providers are taking a lot of risk, banking on the certainty it will all work out over what they expect to be a fairly short horizon. However, pushing all your chips to the middle of the table doesn’t entitle you to win. Ultimately, your investment (and speed of investment) isn’t the market’s problem. 

• Don’t Forget About the Humans

I’ve seen a lot of leaders who want to have it both ways. They tell workers they need to get on board in learning AI or they will be replaced (trust me, they know), and then double down, indicating that even if people get on board they are going to get replaced anyway. That’s not leadership, and that type of decision making cannot sustain itself. As we’ve learned in the era of disruption, every business model and company is replaceable based on the propensity for human ingenuity to flourish when challenged. Lead with thoughtfulness and wisdom.

Unfortunately, an outcome of that thinking has often led to an unforgiving mistake -- the rampant rush to replace human knowledge with the artificial variety too quickly. Careful what you wish for. I can’t predict the actual timelines, but while that decision path likely won’t be as debilitating in 10 years, for the next 5 that knowledge is crucial.

Moving Along

Don’t worry, I’m almost off my soapbox, and thanks for bearing with me. AI has produced a bit of a cottage industry around influencing. Not surprising given its rise in other social mediums. Relative to that influencing, two more quick points that are, well, a bit more random:

• We can never have enough leadership on readying the workforce for and organizing to scale outcomes that require industrial AI. Without people in the mix, especially up front, technology isn’t going to work. Traditional change management isn’t designed to support the continuous change that industrial AI creates, which is the new “steady state.” ARC will do its part in providing AI thought leadership, but we also welcome other voices to our community that push the conversation forward.
• To all you influencers, declaring AI will kill something doesn’t make it so. I come to this with humor when I say I could do with never seeing again the declarations that AI is killing PowerPoint. Those click-bait statements seem silly, particularly those that then go on to list a bunch of AI tools that can be applied to PowerPoint. Making PowerPoints faster to build and more effective for both less and more-experienced users isn’t killing it. It’s extending its life. Kind of petty, I know, but it has become a pet peeve 😊.

We can help

ARC is uniquely positioned at the center of industrial innovation, the intersection of the operations and enterprise, including people, processes, and technology. From educating industrial AI users on separating value from noise and developing actionable roadmaps to assessing the right solutions and introducing you to those who have succeeded, we can help. 

*Other than the image, no AI was used in the making of this blog. Seemed only right, given the topic.