Invented by Thangarajan; Parthiban, GENESYS CLOUD SERVICES, INC.

Contact centers are the backbone of customer service for many companies. But the people who work in these centers often get tired, especially after long hours and back-to-back customer calls. The patent application we’re discussing today introduces a smart way to track how tired agents are, using computers and automated tools. This not only helps the agents, but it also makes the entire contact center work better.

Background and Market Context

Contact centers are everywhere. If you have ever called for help with your phone bill or chatted online to fix a problem with your internet, you’ve talked to a contact center agent. These agents are expected to be fast, helpful, and friendly in every interaction. But working in a contact center can be very demanding. Agents often handle calls, emails, and chats all day, sometimes even through the night. It’s not surprising that many of them feel tired or even exhausted by the end of their shifts.

Fatigue among contact center agents is not just about feeling sleepy. When agents get tired, they become slower at typing, take longer to reply, and may sound less friendly. This can lead to unhappy customers, more mistakes, and even lost business. Supervisors try to watch out for tired agents, but it’s hard to know exactly when someone is feeling fatigued. Most of the time, supervisors only find out after the agent or the customer has already been affected.

The cost of agent fatigue is high. When agents are tired, they may not perform as well, leading to longer call times, more customer complaints, and higher turnover. Companies spend a lot of money hiring and training new agents, only to see them burn out and leave. This cycle hurts both profits and the reputation of the business.

Today’s contact centers use a mix of real people and automated systems. Some calls are handled by chatbots, but most still need a human touch, especially when customers have complex problems. The need to keep agents performing at their best is more important than ever. That’s why there’s a growing interest in tools that can detect fatigue early and help supervisors take action before problems happen.

This patent introduces a solution that uses computers and smart software to monitor agents while they work. It tracks how they type, move their mouse, speak, and other signs of how fast and well they are working. By comparing these signs to the agent’s usual “fresh” performance, the system can spot when someone is getting tired. If the system finds a problem, it can suggest steps like giving the agent a break or changing the kind of work they do. This is a big leap forward from the old way of just guessing or waiting for issues to show up.

Scientific Rationale and Prior Art

The science behind this system is both simple and smart. People show signs of fatigue in many small ways. They type more slowly, click around more aimlessly, or pause longer before answering. Their voices may become dull or slow. These changes can be measured with computers. By collecting this data over time, you can build a picture of what “normal” looks like for each agent. When the current behavior starts to drift away from this normal, it’s a sign that fatigue may be setting in.

Before this invention, supervisors had to rely on simple metrics like average call time or customer feedback to spot tired agents. Some companies tried to measure things like keyboard activity or time spent on breaks, but these methods were basic and could not truly understand how tired an agent was. There have also been efforts to use computer programs to track work pace, but most of these systems used fixed thresholds that were the same for everyone. This meant that natural differences between agents were ignored. A fast typist and a slow typist might both be fine, but their “normal” is not the same.

Machine learning, a branch of artificial intelligence, has changed how computers can find patterns in data. In recent years, models like neural networks and autoencoders have become popular for finding hidden changes in complex data. In an autoencoder, the model learns to compress and then rebuild the data, focusing on what is most important. If the data suddenly changes a lot from what the model expects, it signals that something is different – in this case, the agent might be tired.

Other industries, like trucking or aviation, have used similar ideas to track fatigue in drivers and pilots. They look at things like steering behavior or reaction time. But using these methods in a contact center is new. Each agent works differently, so a one-size-fits-all approach does not work. This invention takes the idea further by building a unique model for each agent, learning from their own best performance.

In summary, earlier methods for spotting fatigue were too simple, too slow, or did not adapt to each person. This patent uses the latest advances in machine learning to create a custom-fit solution, tracking many different signs of work pace and adjusting as agents change over time. It is this mix of careful data collection, smart modeling, and real-time feedback that sets this system apart from the old ways.

Invention Description and Key Innovations

Let’s break down how this invention works and what makes it special.

First, the system builds a “baseline” model for each agent. Think of this as a personal profile of how the agent works when they are not tired. To do this, the system watches the agent during parts of their shift when they are most likely to be fresh – usually early in the day. It collects data on things like:

– How fast and accurately they type in chat or emails
– How quickly they move their mouse to complete tasks
– How their voice sounds – for example, how loud or soft they speak, and how long they pause
– How fast they respond to each new customer message
– How many words they use in each reply
– How often, how long, and what kind of breaks they take

The system might also collect “context” data, like what time of day it is, what kind of shift the agent is working, or how long it has been since their last break. All this data is fed into a machine learning model, such as an autoencoder neural network. The model learns the patterns that mean “this is how this agent works when they are not tired.”

Once the baseline is set, the system keeps watching the agent as they work. It breaks each shift into smaller parts, maybe 15 or 30 minutes at a time. At the end of each part, it collects the same kind of data and compares it to the baseline. If the new data matches the baseline closely, the agent is probably still fresh. If the new data is very different, it means the agent might be getting tired.

The model outputs a “fatigue score,” showing how much the agent’s current performance has changed from their best. This score can be shown as a percentage or on a simple scale. If the score gets too high, the system can alert a supervisor or even take action automatically. For example, it might suggest giving the agent a break, routing fewer calls to them, or letting them handle only less urgent tasks for a while.

What’s clever about this system is that it doesn’t treat all agents the same. It learns what “good” looks like for each person, and keeps updating as the agent gains experience or their work style changes. The model can be retrained regularly, so it always knows what to expect from each agent.

If a supervisor gets an alert, the system can explain exactly what has changed – maybe the agent is typing more slowly, or taking longer breaks. The system can also recommend what to do next, and if needed, wait for approval before acting.

This approach has several advantages:

– It’s automatic and runs in the background, so agents and supervisors do not need to do anything extra.
– It adapts to each agent, so it’s fair and accurate.
– It gives clear, real-time feedback, so problems can be fixed before they get worse.
– It can suggest or even carry out helpful actions, like giving breaks or changing work assignments.

The invention can work for one agent or for groups. For example, it can track a whole team or agents working on a certain type of call. It can also learn from the effects of different actions, so over time, it gets better at recommending what really helps agents recover from fatigue.

The technology can be built into existing contact center systems, and works with both cloud-based and on-site setups. It uses standard computers and does not need any special equipment. Data is collected through software scripts that run on the agent’s workstation.

In technical terms, the patent covers both the method (how the system works step by step) and the system (the computer hardware and software needed to make it work). The claims are broad, covering different types of data, models, and actions, making this invention flexible and powerful for the changing needs of contact centers.

Conclusion

Agent fatigue in contact centers is a real problem, hurting both workers and businesses. This patent offers a new way to track fatigue using computers and machine learning. By creating a custom model for each agent, watching for early signs of tiredness, and taking action before performance drops, the system helps agents stay sharp and customers stay happy.

The mix of smart data collection, modern machine learning, and real-time feedback makes this invention stand out. It’s fair, accurate, and easy to use. As contact centers keep growing and customer expectations rise, tools like this will be more important than ever. By understanding and acting on agent fatigue, companies can boost performance, keep agents healthier, and make every customer interaction better.

Click here https://ppubs.uspto.gov/pubwebapp/ and search 20250217733.