*How AI and ML are revolutionising the optical industry and creating an optimised customer experience*
By Anja Fordon, EMEA Staff Writer
Max Weber, Head of People & Workplace Operations at Mister Spex, recently spoke to Workday about the use of AI and ML, and key themes revolutionising the way companies like Mister Spex operate. Learn more right here.
The use of AI and ML is revolutionising the way companies operate. But what does this mean in actual terms? How does the integration of these advanced technologies change the operations of a company that exists in both the digital and physical space?
The potential applications of AI and ML are many and range across different areas – from optimising inventory management to personalising the customer experience. But how does it work? And what does it mean for you as the end consumer?
In a recent interview with Max Weber, Head of People & Workplace Operations at Mister Spex, we found out. If you’re ready to take a look behind the scenes and learn how AI and ML are revolutionising the world of optometry, then read on and get inspired. Because the future is here – and it looks dazzling.
How is Mister Spex using AI and ML in their operations?
Mister Spex, as an omnichannel business, implements AI and ML in its operations in many ways. We use these technologies in both our retail and online businesses.
One prominent area where we use AI is our online shop. AI plays a central role in identifying the right products for each individual customer. One example is our recommendation algorithm, which suggests products based on the customer's preferences.
Another practical application for AI is staff scheduling. We use a solver AI to optimise this process and ensure we always have enough staff available to meet our customers' needs.
We also use technologies such as natural language processing in various areas. For example, it’s used in communication with our staff and helps us evaluate the feedback we receive from our clients.
In general, we have various use cases for AI and ML integrated into our business model, ranging from smaller to larger. However, what we don't currently have is a comprehensive ML architecture. For us, the most important thing is to generate value for customer experience and in some cases for employee experience, rather than constantly asking fundamental questions related to our technical architecture.
AI and ML are valuable tools for us, provided they’re used in a way that respects the privacy and autonomy of employees.
To what extent has the recent acquisition of Tribe influenced the use of AI and ML at Mister Spex?
The acquisition of Tribe has had a significant impact on our customer experience and has significantly increased our knowledge in the field of AI. For example, if you visit our Mister Spex website, you can see that we use AI to interpret your face shape and eye distance live on screen to measure pupil distance.
The acquisition of Tribe has given us additional knowledge and technology that now allows us to offer 3D-printed eyewear in our first shops. Using digital measurement techniques, we can produce eyewear that is perfectly tailored to the shape of our customers' faces. This unique offering can give us a significant competitive advantage in the highly competitive eyewear retail market as we roll this product out further.
Tribe, which joined our family in 2021, was a small company with a specific use case aimed at giving us a competitive advantage. Therefore, the acquisition of Tribe has paid off in every way.
How will AI and ML impact HR for you and in general?
The application of AI and ML in HR is a challenge because you always ask, ‘How far do we want to go? To what degree do we want to use data to predict people's behaviour?’ When we apply AI and ML to product development or risk management, it has a less direct impact on individuals' lives. However, if we try to predict employee behaviour, such as when an employee might quit, it could lead to us making decisions that heavily interfere with people's lives.
That said, we have used AI and ML in some areas, particularly to improve work processes rather than to predict individual behaviours. For example, this year we launched Workday Scheduling, an AI-powered tool that automatically creates staff schedules. It takes into account both customer behaviour in our shops and the availability of our staff to create optimal schedules. This tool has significantly reduced the complexity of the process and saved us valuable time, which has ultimately translated into an improvement in our profitability of more than 10 percentage points.
We have also introduced Workday Peakon Employee Voice, which uses ML to identify overarching talent trends and find issues that engage employees. However, this is not about making predictions about individuals, but about the whole organisation. The goal is to improve the organisation as a whole rather than managing individual cases. AI and ML are therefore valuable tools for us, provided they’re used in a way that respects the privacy and autonomy of employees.
What challenges have you experienced when implementing AI and ML into your operations? Can you give specific examples where there were difficulties or things went differently than expected?
One of the biggest challenges in applying AI, for example in workforce planning, is to give it a clear goal to optimise. This goal is derived from various data, such as visitor numbers, expenditure, age distribution and so on. But defining a clear goal is often the biggest problem, as it depends on verifying qualitative correlations in a way that a machine can understand and target.
At the same time, AI also needs valid data for learning. One advantage of using AI products, such as those from Workday, is the ability to also benefit from learnings from other customers' data. However, one must always take into account the context and the specifics of your own company, as suggestions from an AI can’t always be directly applied to your situation.
A good example of this is that AI at one company may learn that a lack of sales advisors leads to lower sales figures. This information couldn’t be directly applied to another company, which may have a different distribution or relationship between customers and sales advisors. The challenge, then, is to bridge the gap between general information that AI learns from various data sources and specific information that is specifically relevant to your own company. Data-driven decisions should always take this distinction into account.
One advantage of using AI products, such as those from Workday, is the ability to also benefit from learnings from other customers' data.
How do employees feel when they learn their deployment planning is done by AI? Are there any internal hurdles or resistance?
We avoid full AI staff planning to avoid statistical overfitting and to use the valuable input of our store managers. Overfitting occurs when an AI is trained on 100 percent of the available data. In this case, the AI tends to replicate the world exactly as it has always been, leaving little room for future change. A practical example would be if an event, such as a Justin Bieber concert last year, led to an increased number of customers, a fully data-driven AI would expect to see the same amount of customers again on that date this year, even if no such event is planned.
To avoid this, we use AI to do most of the staff scheduling, leaving room for store managers to make adjustments and bring in their experience and intuition. This combination increases the acceptance of AI-supported planning among employees.
In addition, we have the ability to incorporate employee preferences into scheduling, for example, if a part-time employee can never work on Fridays. This feature, provided by Workday, also promotes the adoption of AI-assisted scheduling. Digital tools also allow us to make processes such as shift swaps easier and more transparent, which also contributes to the acceptance of AI-assisted scheduling.
How do you ensure that the data used for AI is valid, clean and sufficient? Do you have any kind of human supervision or control to ensure the quality of the data?
Yes, we definitely have human oversight, especially in terms of staff scheduling. Once a plan is generated by the AI, a manager checks if the results are valid. In addition, all staffing data consumed by the algorithm is constantly checked and must be approved. Every change is audited and reviewed by at least two people, which allows us to use the ‘four-eyes’ principle to ensure that our database does not contain any erroneous data.
As far as traffic data is concerned, it’s a challenge to map it 100 percent accurately – for example, if two people go through a light barrier at the same time – but that is a big data problem. An error rate of four to five percent is acceptable in this context because the rough direction is correct and we can rely on the vast majority of the data.
To what extent do you see technologies such as AI continuing to influence and change the optical industry in which Mister Spex operates? Do you see any other potential or areas where these technologies could have a significant impact? Also at Mister Spex itself?
There’s certainly a lot of potential for the application of AI in the optical industry. For example, AI can play an important role in content creation and customer experience, such as through the development of chatbots. Giving customers the instant option to solve the problems regardless of opening hours of the customers service will be a benefit for customers and Mister Spex.
As a company with a long online history, we have a treasure trove of data that we can use strategically. For example, using spectacle frame size recommendations based on face recognition or historic order behaviour we can reduce returns and increase customer satisfaction. We’re offering 3D-printed glasses based on AI featured measurements, which are a completely new type of eyewear, in more and more stores. Order and return predictions help to optimise work scheduling and the speed of delivery. Although I can't go into detail, I can assure you that we have a full list of topics on the agenda.
