So you're ready to join the AI revolution. But are your workers?
Alexander Khaytin
Alexander Khaytin
When integrating AI into your business, you might have a lot of expectations for the algorithms. But there's one very real issue of AI implementation we've encountered, and this one is entirely on humans. Today we are sharing some useful insights into the trust issues between humans and technology and how to address them in your company.
Many businesses are thinking about bringing digitalization to their factories and production processes. And that's a great idea! AI can bring a lot of value to an industrial company.

There is, however, something that can stand in the way of successful AI Implementation.
In machines we trust?
For one of our clients, Mechanica AI developed a predictive machine learning solution to optimize their processes. Using the past data, the algorithm makes a prediction and suggests a recommendation, that on-site experts can then act upon.

But some surprising results came through. Soon we realized, that we are not getting the expected outcomes. But what went wrong?

Turns out, many on-site technicians simply did not trust our model's predictions.

¼ of the teams completely ignored the recommendations provided. Moreover, not a single team followed 100% of the recommendations.
The last step to this full-scale integration was challenging, however, it helped us gain critical insights on how to embed the AI into the complex environment of the big steelmaking plant.
How did that happen?
The first issue is the lack of human trust in new technologies.

The shop floor specialists are not interested in AI projects. They are expert metallurgists that have been doing their work for years, and they don't see the benefit of Machine learning model implementation. And why would they? They have the expertise and they've successfully done their job all this time by themselves. Why is improvement needed?

The second issue is the old habits, built throughout a worker's career.
Man is largely a creature of habit, and many of his activities are more or less automatic reflexes from the stimuli of his environment.
G. Stanley Hall
We humans do not like to change our routines. We want things to stay the way they are, and the same can be said about your employees.
So how can you address that?
An implementation plan is key. No matter how good your AI accuracy is, you won't get any results if it's not used. It is important to understand that there are going to be problems. The implementation of new technologies always comes with challenges. But, we've got good news too!
Forewarned is forearmed
There are ways to address such issues. We in Mechanica AI always make an effort to discuss these potential mishaps beforehand with the customer so that a reasonable action plan can be created.

Firstly, we make time to sit down and talk with some shop floor specialists to explain our processes and what is required of them. We demonstrate what the algorithms do and how they can improve the operators' work. It is a good idea to have some discussions with middle management as well. The goal is to increase awareness and remove the fear of uncertainty.

When implementing our algorithms in the full-scale use, we now introduce a familiarization period. During that time, the AI-based recommendations are provided, but none of the teams are required to use them yet. That creates a sense of comfort, and taps into the human curiosity, lowering the distrust barrier.

Another solution that we found effective is introducing daily reports for management – not on the algorithms, but on the humans. Monitoring the algorithm is just not enough, and in the early stages of integration, it's way more important to monitor how it's used. These have to be absolutely as basic as possible – nobody has time to analyze detailed progress sheets. The methodology should be clear and transparent.
Example report
Having that data allows for a simple way for the managers to address the issue. The teams with the lowest implementation percentages can be contacted and requested to follow the procedures agreed upon.
In our experience, these methods led to a 20% to 80% increase in recommendation uses.
Innovations become the new normal
It's not the first industrial revolution and neither it's the last. People might be unsure or worried, and that is absolutely natural. They might think it will lower their quality of work instead of improving, or concerned that new tech will steal their job. But with time it will become evident, that it is not the case. In fact, the algorithm is learning from their expertise, it is a hard-working student, looking for a way to help.
Simple steps for successful implementation
Choose a contractor with experience

Many companies are offering AI implementation nowadays. You hear about endless "successful pilot projects" and their determination to help you. Unfortunately, pilot projects can't show many of the difficulties of the process and proper full-scale integration experience is crucial.

Find innovation-spirited people

There might be many people who don't care for innovation and AI. But there are also people who are. Find those with natural curiosity and enthusiasm for the subject in your operation teams, because these are the people that will help you get the most efficient implementation.

Bring an economist along to your digitalization team

It is a great idea to have an economist by your side. AI implementation is a complex process and sometimes it's tricky to evaluate the value it brings to your business. In our experience, different employees might come up with different ways to evaluate the success of the project, which can get rather confusing. A perspective of an economist can help to get some clarity on your progress.

Find a moment to track results daily

As mentioned before, it is very helpful to introduce daily management reports. Reviewing such reports on a regular basis might seem like an extra hassle, but it does make a major difference for the implementation process. The reports should be simple, including only essential information, so it shouldn't take much time. These 5 minutes of your time will help to identify the problem areas and take actions in a timely manner.

Want to learn more about the kind of experience Mechanica AI has in the industry?

More posts you may like