Manufacturing change management wisdom is a give-and-take between technology and people. However, high-impact AI solutions must adapt to a business’s needs first, not the other way around. Manufacturers, for their part, understand that the business of manufacturing is complex. Plants are highly contextualized, variable, dynamic, and people-centered. Common sense dictates that data-driven optimization technologies for the smart factory somehow mesh with these qualities. On this last trait (human-centricity), collective human intelligence on the factory floor cannot be discounted. After all, it is the plant team’s responsibility to preside over the engine room of industrial creation. Operators, managers, and plant engineers are, therefore, guided daily by their individual impact on a complex physical process. This is even more the case as manufacturing processes are further automated — technological advances in manufacturing render the upskilling and reskilling of people more critical than ever.
Successful change management initiatives for an AI-driven factory environment acknowledge these production realities. They leverage insights into the manufacturing process itself and engage with the operation’s human custodians.
THE AIAAS APPROACH—HUMANS-IN-THE-LOOP FOR AI CAPABILITY
Collaborating with a technology vendor around an AI use case can jumpstart a manufacturer’s digital maturity engine. Employing a hands-on (human-in-the-loop) approach via AI-as-a-Service (AIaaS) bridges the divide for industrials—between AI as a concept and AI as an explainable tool put to practical use.
Mikhail Royeppen is DataProphet’s Solutions Architect (a Chemical Engineer by training). He comes in during the business development phase of the AI implementation journey. Mikhail emphasizes that establishing an understanding of a factory’s day-to-day operational workforce dynamics and priorities is crucial since success will entail working with specific plant teams on a unique challenge:
“Enabling our solutions depends on plant operators enacting the changes our data science solution is suggesting. Because of the prescriptive nature of our work at DataProphet, we’re very much humans in the loop. So, we need to take humans into account. This means generating reliable prescriptions in an understandable format, allowing operators to do their jobs even better.”
Mikhail has found the following indispensable to establish from the outset:
- understanding the current operating paradigm of the plant itself.
- determining how teams currently go about optimizing their operations.
HUMAN INTUITION + AI = CHANGE MANAGEMENT WISDOM
We can draw a useful implication from Mikhail’s experiential insight into impactful Fourth Industrial Revolution (4IR) change initiatives. Digital transformation of production is built on consensus.
To manifest this consensus, individuals need to cultivate confidence in algorithmic prescriptions. Research bears this assertion out. To earn trust in next-generational computational thinking, the validity of human intuition must be acknowledged and integrated. In factory environments, subject matter expertise and operational experience matter. For manufacturers, the plant’s existing wisdom must not be discarded on the path to digital maturity. On the contrary, technology vendors are well-advised to see it as an essential part of the solution process.
“If we consider preserving pre-existing expert knowledge within our system, we’re looking at historical plant states. These states show where the plant has run the best. Plant teams know why those plant states occurred. More often than not, they also know when and how they occurred. Finally, they know, to a degree, at least with a gut feeling, why things change when the plant isn’t operating as well as it should be”, Mikhail elaborates.
Even so, as critical as factory personnel are to 4IR manufacturing, all efforts will fall flat without the right base technology in place.
THE HUMAN-MACHINE-PROCESS NEXUS
The basic premise for AI-guided technology in factories is clear — augmentation of the human decision-making process to uplevel operations (to be more agile, efficient, and productive). Put succinctly, technological change in plants happens at the human-machine-process nexus. This is the give-and-take of the smart factory journey. To ensure human buy-in, the AI-guided decisions factory personnel are asked to work with should yield a crystal clear positive impact.
Without safe, transparent, and measurable improvements to production KPIs—all faith in data-driven decision-making is lost. It is no accident that a whole sub-field of explainable artificial intelligence (XAI) has emerged to strengthen people’s underlying grasp of what machine learning models are actually doing.
3-D GLASSES FOR SMART FACTORY STEP-CHANGE
In evaluating a data-driven production technology, a holistic mindset works best. Seeing an AI application in three dimensions leads to more reliable judgments around its inherent value. As outlined above, the best data-driven solutions are context-adaptive and problem-oriented. They start with pain points and work towards agreed business outcomes. Additionally, as the European Commission advocates, they are also safe, technologically robust, and allow for human agency and oversight.
So-called “plug-and-play” solutions designed to optimize a complex physical process often assume the problem must somehow fit the technology. This approach comes with limitations, which are exposed in light of three interdependent variables:
- the targeted industrial process
- the adopted technology
- the people interfacing with it.
DataProphet’s AIaaS approach takes this 3D view, illuminating how deep learning applied as prescriptive analytics adaptively modifies a production line’s current process in all its complexity—thus moving the plant to a more optimal state.
MANAGING COMPLEXITY — HUMAN INTUITION VERSUS DATA AS A SOURCE OF TRUTH
For Mikhail, this is where the change management dimension of digital adoption comes in:
“Any solution that we deploy into a given environment needs to augment the current working process instead of changing it completely. For this reason, when we’re doing the initial AI Readiness Assessment (and even before that, during the sales process), it’s really important to initiate and sustain the right dialogue with plant teams from the outset. For example, we will engage with operators, plant engineers, and even IT systems owners to find out how they’re doing what they currently do. We can then glean from them how (were our product to be deployed) they would be able to use it to make their lives better on the plant—to reach KPIs and targets without the intervention having a detrimental impact on the way they currently work.”
On this last point, Mikhail again underscores the importance of preserving the plant’s existing expert knowledge. To secure the initial buy-in for AI deployments—from the plant manager and engineers to the plant operator—AI vendors need to appreciate that plant personnel fundamentally understand how the plant operates.
However, subject matter experts need to work with data scientists (preferably with manufacturing experience) to interpret the available plant data at the next level. Again, because the plant is such a dynamic, ‘living’, thing, plant teams have insight into the subtle nuances of the plant environment during any given shift. Factoring this in, data scientists can use machine learning models to parse the sheer mass and complexity of the right data emanating from machine sensors and other plant data sources.
As Mikhail observes:
“Where DataProphet comes in is, through collaboration with plant personnel, integrating and modeling the data that accurately reflects the historical operation. We then generate prescriptions from this data to get the plant back to its most optimal and stable state—deliberately, safely, and repeatedly.”
With this in mind, Mikhail sees part of DataProphet’s job—through the results it achieves—as earning the trust of plant teams in their data as the source of the truth rather than mere intuition.
This sense of ownership is critical to change management along the digital adoption curve. Manufacturers may be impressed by the operational impact of AI. However, sustaining belief in the base technology going forward goes deeper. Ideally, people cultivate a sense of their own impact as vital to the digital transformation of manufacturing.