How We Use AI To Better Customer Experience
As part of the build up for the upcoming AI & Big Data Expo World Series, we have prepared a virtual session on how Artificial Intelligence (AI) and Machine Learning (ML) is used to better customer experience.
Extract of some of the points shared in the session.
How does Reflex Media use AI?
Ruben Buell, CEO - Reflex Media: At Reflex Media, we use AI for a number of things. First, we are a dating platform that runs to millions and millions of users every day. At Seeking.com; our flagship product, we have over 40,000 photos uploaded every day. We have 2 to 3 million messages crossing paths on our platform on any given day alone. Using AI, we’re able to make sure that the photos that come in, are photos that we would like to see on our platform. We make sure that photos that are sent across and messages, are photos that members want to see and don’t shock people when they open them up.
We use AI to look at the text messages to make sure that users are not being bad actors. We use this technology to look for fraud, find people who might be out and not people who are actually using the platform to find love and dates.
We also have the company use AI for other items, like financial forecasting. Using it in finance so that we can see where our platform is going, where our revenues are going and other items of that nature.
What are some examples of AI in use on Reflex Media?
Sufiyan Rahmat, VP of Software Engineering: A very good and simple example of how AI is being used on Reflex Media is with image recognition. A user’s photo upload is first labeled by AI and automatically moderated based on the given labels. This in turn helps to speed up approval time and requires less human moderators to step in.
How do you think AI can help predict customer experience needs?
Daryl Webster, VP of Customer Experience: AI can help us predict a better customer experience by learning the behaviors of that customer and then making an experience on the dating site to them, so they will be able to have more robust interactions with customers in the future.
How does AI improve customer experience?
Stephanie Pelayo, Associate Director of Customer Service: AI can be used to shorten approval wait times. We can use AI on the back end to help with our moderation process and that can really get our users into our ecosystem a lot quicker.
How has AI helped your workflow to become more streamlined and/or effective?
DW: AI has helped our workflow become more streamlined. We still look at each profile individually with a human moderator, but the AI has helped us make that decision faster by giving us the details in advance and then make a determination that quickly.
SP: AI has really helped on a two pronged approach. On one end it helps us automate those moderation processes to improve users getting into our ecosystem. And on the other end it has helped us flag potential bad actors and improve the dating experience.
How is Human in the Loop used with AI with Reflex Media?
RB: Human in the loop is used in parallel with Machine Learning. With human in the loop, we’ve allowed the machine to make a decision but they don’t have to make a decision 100% of the time. If the machine can make the decision to either approve or deny 60% or even 70% of the time, the remaining content can be escalated along with a recommendation from the machine. The machine may recommend that the content be approved, but it’s not sure enough to actually approve it itself, so we have a content moderator; a person, actually look at that piece of content and make a determination.
Do you also use Machine Learning, and how does it relate to your AI?
RB: Machine learning and AI is a great piece of technology that’s come about in the last several years. One of the things that is great about machine learning is that we are actually able to train models in a method that’s different than just searching a database.
A great example is you can take any pieces of data, and in essence, label those pieces of data. Then the machine can actually use those labels to learn and make decisions about what the content is, that they’re seeing. A simple example may be to take messages that are from users on your platform looking for dates; for instance, and to label those as proper messages. You can take messages where people are actually trying to get you in a crypto scam or something else and label those as negative messages.
In essence, just that little bit of data alone is enough for the machine to learn and to start recognizing what legitimate messages look like and what actual fraudulent or negative messages look like and can tell us with extreme accuracy whether a person is using our platform properly or is actually misbehaving on the platform.
What steps do you take to evaluate the effectiveness of Reflex Media’s ML model?
Andrew Choflet, Sr. Product Manager: When it comes to evaluating the success of our AI, we first want to make sure that we’re training our intelligence on how to think, how to identify problems, how to look for the right pieces of information that are actually gonna move us towards the goal that we set.
And then we can come in with our data science team to evaluate this process and actually reflect on the results and say, “you know what this might need to get tweaked, that might need to get tweaked”, and then we just really hone in on how we can really deliver value to our users in a way that just saves everybody time.
Do you also use Deep Learning as part of your Machine Learning?
RB: Deep learning is a method of machine learning which really goes after very large data sets. With deep learning, we look at things like photos, looking at the photos and recognizing patterns. Whether we’re searching for things like people that might be trying to upload nudity or hate symbols or text that we don’t allow. We can use Deep learning to recognize that.
Deep Learning is also utilized in sophisticated text patterns. So when we’re looking at messages or other things for specific items or pieces of content, we can use deep learning to go a little deeper.
What is an example of Deep Learning in use at Reflex Media?
SR: Using Deep Learning, we trained a Machine Learning model to ingest, process, and analyze unstructured data to distinguish good versus bad content. In turn, this AI will help us automatically moderate user submitted content without the need for human interaction.
In the future, 80% of new features or improvements should be derived from user behavior. To achieve that, Deep Learning models can be used to build out the user behavior and determine the next steps for the users.
Do you envision replacing human staff with this?
RB: Artificial intelligence is a tool, it’s like any tool that’s come about whether it’s the internet or the microprocessor or the telephone. All these tools we have used to make life better. So here at Reflex Media, we utilize these tools to make the experience on our platforms more enjoyable for our users.
We use it to also help our customer service staff so that what they’re looking at is items that have been escalated from the tools. But all of it is a tool none of it is meant to replace any of the individuals. It is to create better experiences.
DW: We know to use AI as a helper to the human moderation. It’s not the end all be all with AI. We also use that with our human moderators to make decisions more quickly. But the human has a way to look at context more than AI does. So we use AI to help benefit the human moderation.
SP: Our human agents will always be useful and valuable and we don’t foresee replacing them with AI. AI will still be used on the back end to help with our automation and our efficiency. But those human agents will still be used for that interaction with the customer.
SR: Human staff will continue to play an integral role. However, with AI, productivity as well as moderation quality has improved over time.
RB: We do not expect this type of technology to replace staff. What this allows us to do is to create a better environment for our users. To be able to take our customer service staff and other individuals and have them actually be responsive more to the customer one on one, than to be sifting through photos or text messages looking for bad actors all day. The machine can handle a lot of that while, we handle our interaction and better the lives and experiences of the people on Seeking.com.
Where does your AI go from here?
AC: So as we’re stepping into this new domain where technology is starting to learn from how we’re using it and we’re able to train intelligence to solve more complicated problems. Yes, there will be some areas that those tasks that we just do over and over again. We need to have more people showing and thinking about problems in ways that technology doesn’t have the ability to solve for. In ways that lead us into the new direction of where we can go, where we can support people. And the more that we can do that here Reflex Media, the more we can empower our people to solve problems for the future that we might not even see today.
