Are you looking to optimize your plant’s processes with AI? Use this short and sweet guide to cut through the AI hype and find the solution you need.

By Allison Buenemann, Product Marketing Manager at Imubit

Are you a heavy process industry leader getting upward pressure to explore new AI solutions? Are you concerned about your team’s capacity to support evaluating something so new and different? Are you getting 5+ emails a day from vendors talking about how their AI is going to transform the way you work? Are you confused?

If you answered yes to any of these questions, then this guide was created for you. For those who are hearing McKinsey say that AI leaders outperform industry peers by a factor of 3.4x. For those who recognize that AI can add value, but don’t know where to start. For those who aren’t sure you’ve got the right people in place to make an AI project successful. This guide is for you.

Industry leaders speaking publicly about the impact they’re expecting AI to have in both business and sustainability excellence.

The pursuit of process optimization perfection

Technology is one of the lowest cost investments that can be made to get improved performance and margins out of a plant with existing people, processes, and equipment. So how do you cut through the noise to harness new technology? How do you tell whether you’re looking at real, maintainable, value sustaining, true AI software… or if you’re looking at vaporware? These three questions are a great place to start.

1. Is it real?

Ok maybe the better question here isn’t is it real but is it new? This is usually the hardest of the questions to answer, so knock it out up front. Was the company founded on AI or have they started emphasizing it more recently? If the latter is true, has the company make-up changed? Have there been AI experts brought on to drive innovation?

Looking at company make-up and patents can be good ways to gauge the relative “newness” you’re seeing in their offering. It’s easy to market something as AI but hard to defend a solution to questions or demo requests if it’s really just a rebrand of statistics. Ask about the technology, and if met with “it’s proprietary” ask for the name of the technologist.

Lastly, make sure the technology is AI-first. A lot of “hybrid” AI solutions are emerging, especially in the space of industrial process optimization, where first principles have reigned king. AI models built on top of first principles models succumb to the pitfalls of model mismatch and non-convergence when reality deviates from theory.

2. Is it accessible?

Who are going to be the end users of your new AI solution? What skill-set do they have? Will they be able to use it, and more importantly, will they be able to understand it?

Up until recently, application of AI technologies have been limited to the programming savvy. The no-code / low-code revolution in the software industry has taken programming expertise out of the AI equation, enabling domain experts in a point-and-click environment.

If you check the box on can my target consumers use the AI solution, the next and most critical question, is do they understand and trust the solution? Explainability has become a hot topic in AI because the math itself is not fully understandable by all audiences. Explainability, however, can be used to build trust by understanding what outputs they can expect from a model under a given set of scenarios. Interfaces for performing what-if analysis or investigating model relationships can enable this.

3. Is it closed loop?

Will your solution analyze and advise or will it analyze and act? So much of the delta in captured value from different AI solutions can come down to whether or not they depend on a human to capture the value. This doesn’t mean that humans aren’t an integral part of the AI solution. The AI model doesn’t exist without the expertise of the people currently working to operate, analyze, and improve the production process. But when your human-informed AI model is telling you to move 22 variables on a five minute basis in order to drive to an optimal economic or energy target, you’re just not going to get that out of even the best operator you’ve ever had.

So many of the AI technologies on the market offer great multivariate insights, but offering them in an open-loop, advisory way it’s impossible to capture all of the potential value. Your complex AI model, of your complex real plant, is going to result in a complex optimization solution – requiring manipulation of dozens of process variables to get the best outcome. And This type of optimization is going to be most effective in closed loop.

Now you’re off on your journey to find the best real, accessible, and closed loop AI optimization solution for your industrial process. And if you need help finding the north star, we’ve got you covered.

Is True AI right for your plant?

WATCH our on-demand webinar: ‘Hybrid Models vs. True AI,’ to explore the transformative power of True AI and whether you’re making the most of your plant’s data. WATCH NOW.