Rethinking decision-making processes with AI

Human decision-making is influenced by any number of factors — emotions, intuition, subjective experiences, environmental conditions, data, and so on. But when an AI system makes a decision, it’s purely based on data and pre-programmed logic, all optimized for efficiency, accuracy and predetermined goals.

When AI and humans work hand in hand, a true partnership in decision-making can emerge. In our global GenAI report 51% of manufacturers said they will use Generative AI-powered decision management systems to enhance back-/mid-office processes (Q.20) and 43% are already seeing improved business intelligence (i.e., better decision-making) as outcomes of AI deployment (Q14). As the AI innovation space continues to evolve rapidly, we can imagine that it won’t be long before we move from AI-informed human decisions and human-informed AI decisions to AI-only (or autonomous) decisions.

But a lot must happen before we reach that scenario. During a recent panel session at the Manufacturing Leadership Council’s Future of Manufacturing Project conference, we explored the current and future states of AI-driven decision making. With a few exceptions, the lion’s share of manufacturing decisions are currently being made by humans alone. Achieving greater human-AI collaboration — let alone autonomous AI decision-making — will take years and a lot of effort. Here are some of the observations and opportunities on this journey to the AI-driven decision-making future we discussed.

Context is the Achilles heel of AI — for now

Humans make better decisions when we have context. AI systems have the potential to provide the context for smarter, faster, more autonomous decision-making in the front office, on the shop floor or in the R&D lab, but we’re not there yet. For AI to comprehend context better, we will have to do a better job at describing and articulating context to business problems in a way that AI and computers can interpret more readily, completely and accurately.

Take, for example, training and maintenance manuals in manufacturing. In most cases, this technical information is well-structured for human consumption. It's got embedded text, images and videos and it’s optimized for the way humans process information. Unfortunately, it’s not always well organized for the way computers and AI take in information. For AI to get better at contextual decisions, we need to rethink the way we build and capture this information or even retrofit some of the existing tooling to make it more computer-consumable. As technology advancements like image recognition tools accelerate, so too will manufacturers’ ability to deliver the context these AI solutions and their users need.

A partner in decision-making

Manufacturing has been using automation and AI in the form of forecasting, classification and optimization tools for a long time to solve the why’s and how’s. Generative AI is a new tool in manufacturers’ toolbox, offering a step-change in decision-making at all levels of an organization.

As AI gets better and better, particularly generative AI, it can become a true partner in decision-making. But leaders need to make some effort in this process, learning to use AI as a thought partner and articulating what they want that partner relationship to be. Once this foundation is in place, generative AI can help you expand your horizons, asking you questions and nudging you to think bigger. Using AI in this way enables you to approach your strategy or any kind of creative thought process in a much more holistic way and exceed the boundaries of your own lived experience.

The possibilities you’ll uncover in your thought partnership with AI have the potential to enhance operational, strategic, organizational and tactical decision-making in new and unexpected ways.

Trust is key for accepting AI decisions

Trust in generative AI has been and will continue to be a challenge because of its potential to spread misinformation, perpetuate bias and lack transparency. However, we’re seeing rapid advances in technology that are helping with trust and a growing sophistication of agent-based AI technology that will better allow for AI tools to make decisions more collaboratively — and eventually independently.

For example, we now have technology patterns for using large language models (LLMs) to judge the accuracy and quality of the output by other large language models, a.k.a “LLM as a judge.” These solutions essentially act as ombudsmen, ensuring checks and balances in the system and thus the context, accuracy and quality of the outcome. This will also help speed the adoption of AI because of the multi-threaded oversight of the AI solutions.

But building trust cannot be established by technological advancements alone. Manufacturers need to recognize and address the human emotions and behaviors that AI in the workplace is creating — from fear to enthusiasm to confusion — at every level of your organization. A quick personal example: When I took my first autonomous taxi in Phoenix a few months ago, I excitedly shared a video of the experience with my husband, a seasoned technology professional like me. But his response, “Aren’t you scared?” wasn’t at all what I expected. It was, however, a good lesson for me, and I think for all leaders that we need to anticipate and tackle the gamut of worker reactions with consistent communication, robust training and, above all, empathy. In other words, lean into AI literacy and lean in hard across all constituents and stakeholders: your workers, your peers, your executive suite, your board and even your partners.

The panel closed with what sounds like a challenging question: “What is the one thing manufacturers need to focus on when it comes to AI?” But the answer is simple: Don’t wait to get started with AI if you haven’t already. The technology is here to stay, and it will transform the manufacturing industry faster than you may think possible. Waiting to see how things shake out will only make it harder for you to catch up.

Watch the discussion replay.