Building Your Own AI Chatbot for Customer Support With Ken Benet

Welcome to the AI Customer Service Podcast, where we explore how artificial intelligence is transforming customer service and support through conversations with industry leaders and innovators.

Today, we have Ken Benet as our guest speaker. Ken is the Manager of Customer Experience and AI-Enhanced Support at We The People NYC. Before this, Ken studied UI/UX at Pratt Institute with a focus on AI.

Ken, thank you so much for taking the time to join us on the podcast. How are you doing today?

I’m well, man. Thank you so much for having me. It’s an honor to be here.

My path into AI and customer experience did not come about in the most traditional way. I actually started as a musician. I earned advanced degrees in music and even founded a music and art school on Long Island, New York.

That is where I learned how to break things down in a way that people could actually connect with.

Eventually, I moved into legal document preparation, and I now wear many hats at We The People in New York City, including customer experience, marketing, and social media.

Working with clients who are struggling to understand legal procedures taught me something very important. The real problem is not people’s ability to understand and comprehend. The real problem is often the systems. They are built without considering how humans actually process information, especially during stressful times.

That realization, along with my love of design, pushed me into UX/UI design at Pratt Institute, where I focused on empathy, human-centered design, and creating transparent user flows.

AI became the natural extension of that work.

I use AI as a creative partner, not just as a tool. It helps me prototype quickly, rewrite dense legal explanations into plain English, evaluate user journeys, and stay consistent across service categories.

It is really a collaboration between human intuition and AI speed.

At We The People, this partnership lets me completely redesign how customers get their information. I built an AI assistant, a custom GPT, that answers questions like a normal human would, with boundaries, clarity, and plain English.

Basically, when someone interacts with this GPT chatbot, they are essentially interacting with me.

To sum it up, my whole career — music, teaching, UX, and now AI — has really been about making complicated things feel manageable.

That’s awesome. I didn’t know that you created the GPT yourself. So that was a project you took on completely by yourself?

I did, absolutely.

I was able to develop those skills through Pratt. They had some absolutely wonderful, top-notch instructors there.

The GPT I created does exactly what I want it to do and, importantly, what I do not want it to do. I do not want it to give any type of legal advice. I want it to interact with people the same way you are interacting with me.

It has a very human element to it, and it keeps things focused on the services we provide and what we are actually able to do for people.

Would you mind sharing what We The People NYC does?

Sure.

At We The People, we are a legal document preparation company. We are a small company, and that gives me the opportunity to wear a variety of different hats.

As I said, I do our social media, designs, marketing, and many of the creative aspects of the business. I really enjoy that.

We help people through the process of filing for divorce, filing for bankruptcy, creating companies, creating wills, and other legal document preparation services.

These are services where someone does not necessarily need a lawyer because it is paperwork. We are not lawyers, and we never present ourselves as lawyers. We are simply experts in preparing specific types of paperwork.

That means your AI has to be very specific and cannot give legal advice. So there have to be guardrails.

Exactly.

There are many guardrails. I give it instructions and boundaries. I also upload files into the custom GPT, so it is not randomly drawing information from the internet. It is drawing from specific information that I provide.

And, of course, I give it a tone that people can interact with.

That’s really awesome. I love where this is going.

What were the early customer support or CX challenges you observed that made you think there must be a better way to resolve this with AI instead of doing it manually?

When you joined We The People NYC, did you face anything specific that made you think, “Why not try AI for this?”

Yes, in a variety of ways.

When I came to We The People, things were still in a very physical paperwork stage. I helped bring us into the 21st century by creating automated systems.

One of the biggest things that helped from a customer service perspective was breaking down the dense legal barriers that people think they are going to face and making the information simple and understandable.

That was one of the reasons I created the custom GPT.

I also created a number of synthetic customer personas. These personas represented our customers — people from the Bronx, Queens, people with money, people without much money, and different customer situations.

I effectively interviewed our market base through those personas, which helped me better program the custom GPT.

When someone is filing for divorce or bankruptcy, they are often under stress and have a lot of questions. When you are stressed, your brain does not work as clearly as it normally would.

Helping people get their questions answered in plain language really makes a difference.

It also helps when they decide to call us. They always have the option to call us. They do not have to interact with the GPT if they do not want to.

But by the time they call us after interacting with it, we are all speaking the same language because they have already been brought up to speed quickly.

Are you thinking about using a voice AI agent in the future? Since people can reach you through a phone number, is putting a voice AI agent there on your roadmap? Or do you want to take it slow because voice is more technical?

It is not on the roadmap right now, although I do not take anything off the table. Everything is on the table, but it has to be the right time.

For our business, it is just not the right time.

People generally want to speak with another person. They do not want to be shuffled to a call center.

I am not taking anything away from AI call centers. I have used many, and I have been very impressed with the results and how helpful they can be.

But in our line of business right now, I really like the human touch. It is the combination of AI and the human touch.

I do not ever see us going fully automated because we are human. People want that human interaction.

Totally agree. A lot of people feel the same way. The industry will need time, and people really do want to speak to a human in empathetic situations. If someone is going through a divorce, they may not want only a chatbot.

Could you walk us through a specific use case, from problem definition to deployment to results? Have you seen reduced time, improved CSAT, faster response time, or cost savings? What is the biggest impact you have seen since using an AI chatbot?

In terms of impact, I think it is mainly customer satisfaction.

Has it saved us time? Yes, I suppose it has saved us time by getting those initial basic questions out of the way. By the time people do want to reach us, we are talking at a higher level because their basic questions have already been answered.

That is very helpful.

One area where AI has saved a remarkable amount of cost is online marketing and our website.

Because of AI, I have been able to update our website, keep it fresh, and keep the content fresh.

It is not just about asking AI to create keywords for a landing page. It is about working together with AI. The AI has to know our goal, but it also has to know our client base.

With that, I have been able to create a variety of helpful blogs that guide people from step one to step three.

As a bonus, it helps our Google rankings and SEO. It is a win-win.

For me personally, it has also been a big bonus because it has taught me so much. A month or two ago, I would have had to outsource certain website development tasks. Now, if I have a problem, I can use AI to help me troubleshoot it.

Coming back to online content creation, these blogs are resources people can access. Instead of only calling us, which they can still do because we always answer the phone, they now have blogs created with AI and curated by me, along with the chatbot.

We make resources very easy for people to access. That has been a very big part of how we use AI.

Very specific to customer support or customer experience, what data or infrastructure do you think you need in place to make AI in support work effectively?

One of the biggest things to keep in mind is that it is not a “set it and forget it” type of thing.

You are constantly refining it. As you use it, you always find small ways to tweak it and make it a little bit better.

In terms of communicating with people and making it human-like, that refinement matters.

As for infrastructure, our company is on the smaller side, so the infrastructure we need is very simple. We use a paid version of ChatGPT, and I also use a variety of other AI programs, such as Midjourney to create images and Sora to create AI videos.

It really depends on what a person wants to accomplish, but there are more than enough tools available.

How do your customers and customer support representatives feel about AI tools being introduced into their workflow? How does the change management playbook look?

With our customers, they have the option to use the chatbot. They do not have to.

But the smart ones do because it has so much information.

The chatbot is designed with first-time AI users in mind. For many people, this may be their first experience with AI, so it is about building trust.

It is set up in a way that relates to people through familiar language and common scenarios. Someone may ask it something simple, get an obvious answer they agree with, and then slowly begin to trust it more.

With our representatives, there was a little hesitation at first when introducing it. But now, everyone is 100% on board because they realize how much time it saves them from answering basic questions and concerns.

Now we can dedicate more time to helping people at a deeper and higher level.

What ethical or customer trust considerations do you think about when using AI in support, such as data privacy, transparency, or handing off queries from AI to a human support agent?

In a legal context, someone may be discussing money, divorce, or other sensitive topics. How do you make customers comfortable and assure them their data is safe?

To avoid that issue, we never take personal information and feed it to the chatbot. We stop that at the source.

If someone attempts to provide personal information, the chatbot responds and tells them it does not need that information.

That is one of the guardrails I have set up, and it helps create trust with people.

Is there a way you measure whether your chatbot plus human support is working better than your previous support approach? Do customers tell you they like this approach better?

The way I measure it is through feedback from my employees and conversations I have with clients.

When I communicate with clients, I always ask about their experience with the AI. I ask whether it was their first time or whether they are experienced users.

Across the board, people have been very pleased with it.

It is not a bad idea to have a formal questionnaire, but I have not gotten to that point yet. Right now, it is more through conversations, feedback, and employees’ observations.

It has been positive across the board.

What is really positive is that once a person is introduced to AI and has a pleasant, successful experience, it opens them up to using AI in other areas of their personal life.

I cannot tell you how many times someone has gone back and forth with the chatbot, then we speak afterward, and now they are asking things like, “I want to create an image,” or “How do I do this?”

I enjoy AI in general, and it is fun to introduce people to it because it can be a life-changer in so many ways.

Whether you want to use it for a chatbot or as a personal stylist, you can do all of that.

When someone asks, “What can it do?” my response is, “What can’t it do?”

I have always been a huge fan, and I think my enthusiasm tends to rub off on our clients.

Does your chatbot have a personality? Is it very serious and professional, or more relaxed? Did you give it a personality so interactions feel more personalized?

Most definitely. It has a personality.

People do not know it, but it has a name, an age, a photo, a portrait — everything.

From the synthetic personas I created and interviewed, I identified the type of chatbot people most want to talk to, the kind of tone they trust, and the kind of personality they prefer.

Generally, they do not want someone too young. They want someone relatable, perhaps in their mid-40s, with a casual but down-to-earth tone.

They want someone who understands the professional language but also has a little wit and a sense of humor. Nothing over the top, but it is definitely there.

So it feels like you are speaking with a real person, and that person has a personality.

Personality goes a long way.

Do you go back and check the conversations the AI has with customers every week or month? How do you track that?

That is something I am unable to track right now. I wish I could expand on that, but I do not currently have a way to review the conversations.

I can see the number of users who have interacted with it, but as far as the actual conversations, unfortunately, I am not able to do that yet.

If you ever get time, you can check out Kommunicate. It has the same features you mentioned, and you can go back and check the conversations the AI had with customers. We can talk about that later.

Let’s move to the future of AI and customer support.

What are the biggest emerging trends you are seeing in AI and customer support? What is one thing you would say is definitely going to happen?

In the legal field, AI is not going to replace lawyers. That is for sure.

What it will do is level the playing field.

When someone needs to go through a legal procedure, it can make the process much less intimidating. You do not need to be a Harvard-educated lawyer to understand the legal language in a document.

AI will make things more approachable and simplify them.

That is what AI does in general. It levels the playing field.

Whether you want to create art, photography, videos, or music, AI gives access to people who may have ideas but not the technical skills to execute them.

I can create music using AI that I would never have been able to create otherwise because I did not have those specific skills. I had the ideas, but I did not have the tools to put them down.

It is going to make things really wonderful in the future because it gives access to so much for everyone.

Since you talked about music, have you used Suno?

I have, and I absolutely love it.

I think it is a great thing because I am able to create music that I personally relate to. I write my own lyrics with AI. We create lyrics together.

The music I am creating now includes songs about my life, my dog, and the things I love. They are completely relatable to me.

I can refine the feel of it, the singer, and other aspects. Some of my favorite songs are now songs I have created with Suno because they are so personal.

It is a lot of fun.

From what I hear, AI is not just in your job or customer experience. It is also part of your day-to-day life, from blog articles and help center content to music.

You are using AI every day in different ways.

Absolutely.

In my opinion, if you do not stay up to date with AI, you are going to be left behind. It is bigger than the internet was when it came out.

The people who did not jump on the internet or computers became outdated. If you want to stay relevant, whether as an individual or as a company, you have to incorporate AI.

It is such a wonderful thing. I think you would be foolish not to.

From a customer support perspective, AI gives your customers instant support. You cannot be available 24/7, but AI can take over when you are sleeping or on vacation. It can resolve at least 50% to 60% of basic repetitive queries.

We are coming to the end of the session now.

If you were advising a company that is starting to modernize customer support and bring in AI, what would be your top recommendations?

My top recommendation would be to take your time with it.

Do not rush into it.

I do not recommend setting it up yourself unless you have experience, some training, or have worked with people who have set it up before.

You need guidelines and guardrails.

There is a lot of good that can come from AI, but it is not something that should just be thrown together.

If you get help, check the person’s background. Look at other things they have created. Ask whether they act morally and consider different angles, including what could go right and what could go wrong.

Approach it with caution. Slow and steady is the best way to do it.

Are there any support scenarios where you think AI should not be used?

If you are talking about counseling, life-and-death situations, or something similar, those should probably be monitored.

In customer service, as long as AI is set up properly, it can be used safely. In my case, as long as we are not giving legal advice, there is not too much that can go wrong, provided those guardrails are in place.

There are also situations involving financial data or autonomous actions where companies need to be very careful.

For example, if an AI agent is handling refunds, cancellations, or flight bookings, you need to make sure it follows precise rules.

If a customer paid $20, the refund should be $20, not $22. If someone asks AI to book a flight from New York to San Francisco, you do not want it choosing the most expensive option without the right controls.

These are situations where guardrails and testing become very important.

That is where I cannot use AI for right now.

I would hope those situations could be avoided with better testing. Sometimes companies may think something is good enough to roll out, but those are good examples of why careful testing matters.

Fortunately, my company does not deal with situations like that.

For people in customer support and customer experience who are listening to this podcast, how do you think the role of customer support representatives will evolve in an AI-rich environment?

I think people are going to have to become more specialized.

They will need to add something to the conversation that AI cannot.

As AI continues to improve, that challenge will become tougher. My initial reaction is that support representatives will need to know their product or service even better.

But it is hard to one-up AI in that area.

So I think the real differentiator will be relating to people on a human and personal level. Support representatives will need to bring something AI cannot fully provide at this point: empathy.

Very well said. Empathy is something AI still struggles with.

A customer may say, “I really need help. Can you please help me?” That person could be angry, frustrated, or sad.

A human agent can understand the difference based on context. If the message comes at 3 a.m., it may be an emergency. If it comes at noon, it may be a normal request.

That is where human agents can come in and understand the situation better.

Ken, I really loved the conversation we had today. It was very helpful.

Thank you so much for taking the time. Do you have any final words you want to share with the audience before we wrap up?

Sure.

If you have hesitation about getting into AI or using AI, I suggest looking at it as a partner. It is a collaborative partner.

It can be helpful in every aspect of your life. If you have an issue, want to improve something, or want more insight into yourself and your routines, AI can help.

The more you use it, the more you understand it.

And as a customer, the more you understand AI, the better you will be able to relate to it when you come across a chatbot.

You will understand where it is coming from, how to format a question, and how to interact with it.

It is not always the same as speaking with a person right now, but the more experience you have with it, the more you learn the language.

That was a really good session.

Thank you so much, Ken, for joining us today on the AI Customer Service Podcast. We really appreciate it.

And thank you to everyone tuning in. Feel free to reach out to Ken. We will add Ken’s LinkedIn profile in the description section.

Ken, thank you so much for joining us today.

Thank you, man. The pleasure is mine.